Google Cloud Next

Day 2 at Google Cloud Next: A marathon developer keynote

Thu, 23 Apr 2026 13:00:00 +0000

At Google Cloud, every day is Developer Day, but none so much as day 2 of Google Cloud Next, when we hold the developer keynote. This year’s topic? An in-depth look at Gemini Enterprise Agent Platform. This year’s theme? Planning a marathon for 10,000 participants through the Las Vegas Strip.

OK, let’s run with it.

Gemini Enterprise Agent Platform: A warm up

As the evolution of Vertex AI, Agent Platform “allows you to build autonomous agents that proactively help users — and complete tasks independently,” said Brad Calder, President, GCP and SRE. The platform does so with a whole suite of tools and capabilities to build, scale, govern, and optimize your agents.

Brad then passed the baton to keynote emcees Richard Seroter, Chief Evangelist, and Emma Twersky, Developer Relations Engineer, for an in-depth run-through of the agentic marathon simulator.

The system uses three main agents: A planner to determine routes; an evaluator that assesses routes based on business and community requirements; and a simulator that takes the route, adding actors and randomized behaviors to test the impact on the city,

This, Emma said, turns out to be “a great example of how agents can help us plan, simulate, and think about solving a really big challenge.”

First off the blocks: Building the agent

Mofi Rahman, Developer Relations Engineer, came onstage to demo how the Agent Development Kit (ADK), Google Cloud remote Model Context Protocol (MCP) servers, and Agent Runtime provide the planner agent with the Instructions, Skills, and Tools it needs to improve the initial agent. By the end of the demo, the simulator had generated a route.

“The simulated route looks beautiful,” said Mofi. “Looks like the runners are going to get an amazing view of the entire Las Vegas Strip.”

An agent to evaluate the agent

Next, Ivan Nardini, Developer Relations Engineer, and Casey West, Architecture Advocate, showed us how to evaluate the agent, and build a UI for it. “We want to show you how to move from fragile, unpredictable agentic loops, to a rigorously evaluated network of experts that literally build their own UI,” Casey said.

They did so by deploying a separate model to judge the route, checking both deterministic (e.g., route length) and non-deterministic (e.g., community impact) criteria. For UI development, they showed off the Agent-to-User Interface, or A2UI, an open-source standard developed by Google that created an interface in a single shot. They also used the Agent-to-Agent Protocol, or A2A, and Agent Platform’s Agent Registry, to connect and see which agents are deployed.

“Think of Agent Registry as the DNS of your internet of agents,” Casey said.

Agents that never forget

Richard then mused about how to build agents that get better with time, that “take the learnings from the simulation and optimize for the next run.” Because the answer shouldn’t be to “cram raw text in every request we send back to our agents.”

To capture this learned knowledge, Agent Platform offers Agent Platform Sessions and Memory Bank, plus the ability to turn to tools like Spark or a database to retrieve more information, resulting in an even stronger simulator.

When agents go off course

Thus far, everything had gone swimmingly, but then Richard accidentally “broke” the simulator agent. That provided a perfect opportunity for Megan O’Keefe, Senior Staff Developer Advocate, to show off Agent Interoperability and Gemini Cloud Assist, and how to use them to debug agents at scale.

“With these autonomous agents, the production challenge isn’t just scaling the infrastructure, it’s managing the reasoning, the tool calls — all the places in the system where something can go wrong!” Megan said.

Megan used Agent Runtime trace view to see where the problem was, and using natural language, launched a Cloud Assist Investigation to explore logs and events, which pointed to a specific line of code as the offender. Megan then opened up her Antigravity IDE (powered by Gemini 3, and connected via MCP) to find the problem (an insufficiently run “event compaction” run) and to suggest a fix (add a token_threshold parameter to the event compaction config). She approved the fix and committed it to source, triggering a redeployment to Agent Platform. Problem solved!

Scaling the agents

To this point, all of the presenters had been showing off agent services running as Cloud Run services. Bobby Allen, Group Product Manager, then showed how to convert the apps to Google Kubernetes Engine (GKE), which provides greater control, as well as to use a customized Gemma 4 model, all by vibe coding in the Antigravity editor, which is connected to Cloud Assist. Along the way, Bobby also migrated the agents from GCSFuse to a high-performance Lustre file system.

Closely related to scaling is sharing — making agents available for the world to use and build on. Ines Envid, Senior Director, Product Management and Jason Davenport, Area Technical Lead, showed how to build no-code agents from the Gemini Enterprise app, and how to integrate them with other, “high-code” agents.

Shifting down

Last but not least, it was time to talk about security and governance. “Agents give users and other agents new ways to intentionally — or unintentionally — expose data and behavior in ways that we may not want,” mused Emma.

The standard response to that is to “shift left” — move testing, quality, and performance evaluation earlier in the development process — but for developers, that usually means more work, Richard said. “It’s not sustainable for developers to be responsible for all the layers of the stack,” he said. Instead, “we need to shift down.”

To help, there’s Agent Identity and Agent Gateway, demoed by Ankur Kotwal, head of Cloud Developer Relations. Ankur showed how Agent Gateway uses IAM policies to ensure agent actions are only accessible by approved sources, and how Agent Identity provides each agent with a unique and immutable credential. Then, Agent Policies can be configured to provide guardrails for the agents.

Yinon Costica, Co-Founder and VP of Product at Wiz, then went a step further and showed how Wiz can scan your agent code and infrastructure, and Wiz Green Agent can suggest root cause remediations.

“It’s a full architecture for security to easily understand what you built without you having to actually explain it,” Yinon said. Better yet, he also showed using this functionality from Anthropic’s Claude Code with Opus. “With Wiz, we want to enable your choice of tools and models to fix and prevent real risks,” he said.

The finish line

With this, the developer keynote came to an end. But for Google Cloud developers, it’s just the beginning, as the entire solution is available as source code in GitHub, and all the demos are available as Codelabs. Because when it comes to agentic development, these resources will really help you hit the ground running.

Day 1 at Google Cloud Next ‘26 recap

Wed, 22 Apr 2026 23:00:00 +0000

Last year at Google Cloud Next ‘25, we asked you to imagine a new future for AI. At Next ‘26, the question before you is how do you move AI into production across your entire enterprise?

According to Google Cloud CEO Thomas Kurian, the answer is straightforward: You need a unified stack, with “chips that are designed for models, models that are grounded in your data, agents and applications that are built with those models,” and the whole thing “secured by the infrastructure,” Thomas said in his keynote. (This is the same unified stack that Google uses for Search, YouTube, Chrome, and Android. As Alphabet CEO Sundar Pichai said in his opening remarks, “a big focus of ours is to always be customer zero for our own technologies.”)

As AI matures, we’ve laid out a blueprint on how to succeed. Read on for a whirlwind tour of what we announced from the keynote stage.

Gemini Enterprise: The end-to-end system for the agentic era

Throughout this unified stack is Gemini Enterprise — “the connective tissue between your data, your people, and your goals,” Thomas said, providing a combination of intelligence and automation across multiple layers. Here’s what’s new.

1. Gemini Enterprise Agent Platform

Gemini Enterprise Agent Platform is where you go to build, scale, govern, and optimize agents. As the evolution of Vertex AI, it’s built on top of our leading infrastructure, and deeply integrated with our data and security capabilities — the foundation of the Agentic Enterprise. Here’s a sampling of Agent Platform’s new features and capabilities:

Build: Choose the right environment for the job — from the low-code, visual interface of the new Agent Studio, to the code-first logic of the upgraded Agent Development Kit (ADK). We’ve simplified the entire lifecycle with AI-native coding capabilities to help you ship production-grade agents faster.
Scale: Clear the path to production with the re-engineered Agent Runtime. This supports long-running agents that maintain state for days at a time and are backed by Memory Bank for persistent, long-term context.
Govern: Establish centralized control with Agent Identity, Agent Registry, and Agent Gateway. These capabilities help ensure every agent — whether built on Agent Platform or sourced from our partner ecosystem — has a trackable identity and operates within enterprise-grade guardrails.
Optimize: Guarantee quality with Agent Simulation, Agent Evaluation, and Agent Observability. These tools provide full execution traces and a real-time lens into agent reasoning to help ensure your agents always hit their goals.

To dive deep into Agent Platform, read more in our announcement blog here.

2. Gemini Enterprise app

The Gemini Enterprise app is “the primary environment where your business actually operates,” Thomas explained. The app is where many workers, especially non-technical ones, can ask questions of enterprise agents, create generative media, engage with prebuilt agents, and even create their own with conversational interfaces — all with governance, compliance, and security built in. Here’s a sample of what’s new in this foundational interface:

Gemini Enterprise Projects give your agents permanent memory.
Deep Think solves your most complex business challenges without context pollution.
Microsoft 365 interoperability makes it easy to export docs you create with Canvas into Microsoft Office formats.

To illustrate the power of Gemini Enterprise, Shaun White, three-time Olympic gold medalist, entrepreneur, and snowboarding legend, joined us on stage.

“Back when I was training, our tools were camcorders and guesswork. You’d land a trick and watch it back. And you’d be thinking, ‘How can I make that trick better?’” he said.

Alongside Jason Davenport, Google Cloud Tech Lead, they showed a model that Google Cloud built in collaboration with Google DeepMind that tracked Shaun in space from a two-dimensional video, helping him understand what he was doing right and wrong.

“Learning the trick on the mountain is one thing, but actually understanding the physics of a trick is a whole other thing,” Shaun said.

Read more on the Gemini Enterprise app here.

3. AI Hypercomputer

The same technology foundation that athletes like Shaun White use to understand their performance is being used by enterprises to transform their businesses.

Amin Vadhat, SVP and chief technologist, AI and Infrastructure, took to the stage to announce enhancements to AI Hypercomputer, the integrated supercomputing underneath every AI workload on Google Cloud.

First, there’s the eighth-generation Tensor Processing Unit, or TPU — “a thing of beauty,” Amin said. And because “the demands of training and serving have completely diverged,” this TPU family actually consists of two chips:

TPU 8t, optimized for training, uses new Inter-Chip Interconnect (ICI) technology to scale up to 9,600 TPUs and 2 petabytes of shared, high-bandwidth memory in a single superpod. It achieves three times the processing power of Ironwood and delivers up to double the performance per watt.
TPU 8i, optimized for inference, uses the new Boardfly topology to directly connect 1,152 TPUs in a single pod. It features three times more on-chip SRAM compared to previous versions and a specialized Collectives Acceleration Engine offloads resource-heavy tasks. Taken together, TPU 8i delivers 80% better performance per dollar for inference than the prior generation, helping millions of concurrent agents to run cost-effectively.

AI Hypercomputer also supports Arm-based Google Cloud Axion processors, such as the N4A, now generally available, and will be one of the first platforms to offer NVIDIA’s Vera Rubin NVL72 platform when it is released.

Other parts of the network need to keep up with the demands of the agentic era. For example, the new Virgo Network doubles connectivity to scale training beyond AI Hypercomputer superpods, and Google Cloud Managed Lustre now supports an industry-leading 10 terabytes per second of throughput.

Learn more about all of our AI infrastructure innovations here.

4. Agentic Data Cloud

The AI era hinges on data. Lots of data. That data comes with a catch: It needs to be grounded in context. That’s because “reasoning without context is just a guess,” explained Karthik Narain, chief product and business officer.

To that end, we’re totally rethinking our data platform, and giving it a new name: the Agentic Data Cloud.

Here’s a sampling of what you’ll find in the Agentic Data Cloud:

Knowledge Catalog constructs a unified, dynamic context graph of your entire business enabling you to ground agents in all of your business data and semantics. With Smart Storage and the Object Context API, files in Google Cloud Storage are instantly tagged and enriched with metadata before an agent touches them. Knowledge Catalog is also integrated with Gemini’s Deep Research Agent.
Data Agent Kit delivers a Gemini-powered data science authoring experience across your IDEs, Notebooks, and agentic terminals.
Lightning Engine for Apache Spark is a real-time, serverless engine that is up to 4.5 times faster than open-source alternatives and offers up to double price-performance over the leading competitor for large datasets.
Finally, Cross-Cloud Lakehouse, based on Apache Iceberg, lets you query data in Amazon Web Services or Azure without having to copy it.

Learn all about all the innovations in the Agentic Data Cloud here.

5. Agentic Defense

AI makes — and demands — that everything go faster, and security operations are no exception. Increasingly, “human analysts can’t keep up with AI-driven attacks,” said Francis deSouza, COO, and president, Security Products.

“Security must become an autonomous force, responding faster than the threat itself,” he said.

To help, we introduced three new agents in Google Security Operations to help you defend at the speed of AI.

Threat Hunting agent: Helps teams proactively hunt for novel attack patterns and stealthy adversary behaviors that bypass traditional defenses. Now in preview.
Detection Engineering agent: Identifies coverage gaps and creates new detections for threat scenarios. Now in preview.
Third-Party Context agent: Enriches workflows with contextual data from third-party content. Coming soon to preview.

You can also build your own security agents with remote Google Cloud MCP server support for Google Security Operations, now generally available, and access it from the Google Security Operations chat interface, now in preview.

Then there’s Wiz, now a part of Google Cloud, whose AI-Application Protection Platform (AI-APP), Wiz Security Agents, and Wiz Workflow help you identify and respond to risks and threats at machine speed. New in the Wiz family today are:

Secure vibe-coded applications: A new integration runs Wiz security scanning directly inside the Lovable platform so vulnerabilities, secrets, and misconfigurations caught by Wiz surface in Lovable's built-in security view. Generally available in May.
Secure AI-generated code: Wiz can now remove risks from AI-generated code with inline AI security hooks integrated directly into IDEs and agent workflows, injecting security guardrails before code is committed.
Agent-based remediation: Wiz Skills can equip coding agents and AI-native IDEs with full code-to-cloud context and validated attack surface findings from the Wiz Security Graph.
AI-Bill of Materials (AI-BOM): Work towards eliminating shadow IT by automatically inventorying all AI frameworks, models, and IDE extensions across your environment.
Google Cloud Fraud Defense: The evolution of reCAPTCHA, this platform is designed to discern the legitimacy and authorization of bots, humans, and agents. Now generally available.

Read more about these security innovations here.

6. Workspace Intelligence

For a look at AI for end-users, we heard from Yulie Kwon Kim, VP, Product, Google Workspace, who shared new ways that AI is manifesting in our collaboration and productivity suite. Workspace Intelligence is a unifying semantic layer that breaks down information and context silos for you and your agents. It understands your work, your priorities and the people you work with to help you get more done.

“Think of it as a unified intelligence layer that lives inside every Workspace app. It connects the dots and lets AI do the heavy lifting," Yulie said.

Here’s what’s new:

Ask Gemini in Google Chat allows you to instantly synthesize information, surface insights, and query projects from across Workspace directly from your Google Chat window. It provides proactive daily briefings to help you prioritize, and also lets you take immediate action — such as scheduling a meeting on your calendar or creating a Google Doc to develop a pre-meeting brief — turning your conversations into momentum without having to switch tabs.

Reimagined content creation in Docs, Sheets, and Slides uses Workspace Intelligence to synthesize information from across Workspace and the web and creates professionally formatted drafts that match your voice, style, and brand.
AI Inbox and AI Overviews in Gmail creates a personal, proactive inbox assistant with Gemini.
Google Drive Projects instantly organizes your team's files and emails to manage workflows, generate content, and deliver specific answers based on rich project context. In addition to newly added AI Overviews and Ask Gemini, Projects is another way we’re transforming Drive from a storage tool into an active collaborator to provide insights about your data.
Workspace agent in Gemini Enterprise executes complex, multi-step tasks across Google Workspace apps without having to leave Gemini Enterprise.

For more, check out the full Workspace Intelligence blog.

7. Agentic Commerce

For enterprises, AI agents are reshaping how consumers engage with companies and their products.

“In the agentic era, an agent isn’t just a tool; it’s a strategic extension of your business, built to expand your reach, deepen engagement, and personalize service at scale,” said Carrie Tharp, Google Cloud Vice President of Go To Market Strategic Industries.

Gemini Enterprise for Customer Experience offers a suite of tools to enhance the entire customer journey, from the first moment of discovery through on-going interactions that remember the customer like the best shopkeeper would.

Shopping agent and Food Ordering agent bring new conversational sales and ordering capabilities direct to businesses and third-party chat and digital interfaces.
Omnichannel Gateway helps agents maintain context across web, mobile, and voice, so a company’s agents can offer more personalized service.
Agent Assist helps during complex customer service situations, coaching employees to deliver fast and more accurate answers to customer questions by having organizational data readily available through gen AI grounding.

With Omnichannel Gateway in particular, about bridging the physical, digital, and agentic shopping experience, so consumers always have a familiar, brand-aware experience.

“If a customer moves from text chat to a phone call, the agent seamlessly remembers exactly where they left off,” said Carrie. Now that’s progress!

For more customer stories, check out all 1,302 of the latest gen AI use cases from businesses around the globe.

Innovate all the things

From new products, to new solutions, to new ways of working, there are so many other ways that we’re helping organizations take their AI from pilot to production.

Over the following week, we’ll share even more news, helpful how-to guides, and go deeper on today’s announcements. Stay tuned!

Enabling the agentic enterprise: business and industry agents arrive in Gemini Enterprise

Wed, 22 Apr 2026 21:00:00 +0000

We are officially moving past the era of one-size-fits-all AI. Enterprises today require highly specialized, role-specific tools to drive real productivity — but they cannot afford to sacrifice security, oversight, or control to get there.

What's new: Today, we’re bringing partner-built agents from our Agent Marketplace directly into the Agent Gallery inside the Gemini Enterprise app. This creates a powerful, centrally governed hub for discovering and managing specialized AI from partners like Accenture, Adobe, Atlassian, Deloitte, Lovable, Oracle, Palo Alto Networks, Replit, S&P Global, Salesforce, ServiceNow, Workday, and more.

Why this matters for your business: With millions of paid seats across thousands of companies, the Gemini Enterprise app allows organizations to automate complex, end-to-end workflows in a single, unified environment. Five key benefits include:

Move beyond agentic overclaims: The market is flooded with passive chatbots and rigid workflows. Our featured Marketplace partners deliver true autonomous agents capable of executing complex, end-to-end tasks. They achieve this by combining deep context and memory, intelligent orchestration, and safe tool use.
High-caliber agents: Every featured agent earns the "Google Cloud Ready - Gemini Enterprise" designation only after passing a strict four-step evaluation for basic functionality, output accuracy, autonomous execution, and enterprise standards.
Built-in safeguards: Gemini Enterprise Agent Platform assigns every agent a cryptographically secure identity for a clear audit trail, while our Agent Gateway and Model Armor actively screen traffic to guarantee your data is never used for model training.
Strong IT governance: To balance workforce productivity with strict organizational oversight, we've built Agent Gallery around a secure, two-step governance model. Employees can browse the gallery and submit procurement or access requests for the agents that fit their needs, while IT remains firmly in the driver's seat. Administrators review all incoming requests to approve or deny deployments, retaining granular control over exactly who can interact with each agent across the organization.
A unified ecosystem breaking down silos: Typically, purchasing third-party AI means bolting on disconnected platforms. Our approach eliminates that friction. These partner agents are built to natively run within Gemini Enterprise Agent Platform. This means third-party agents and internal custom builds live side-by-side in a single, cohesive ecosystem, allowing your organization to scale its AI capabilities infinitely without ever fracturing your infrastructure.

Go-to-market opportunity for Partners: Today, we announced a $750 million partner fund for agentic development. In addition, partners can tap into a $240 billion backlog of committed enterprise spend and reach millions of Gemini Enterprise users directly in their daily workflows. We are turning everyday user discovery into a procurement flow for IT, standardizing contracts and accelerating purchasing cycles by up to 50%. With Google sales reps incentivized to work with you, and Marketplace vendors closing deals 112% larger, there has never been a better time to build with Google Cloud.

Where you can get started:

For customers: Discover these agents directly via the Agent Gallery in the Gemini Enterprise app. Start your 30-day trial today.

For partners: Take advantage of our rapid agent deployment framework, which provides a clear path to onboard, commercialize, and make agents discoverable to Gemini Enterprise users. Apply for our AI Agents Program now.

In addition, we’re launching the Google for Startups AI Agents Challenge today, including a track dedicated to startups with agents that can be primed for use in Gemini Enterprise.

Meet the agents your organization can start using today

A sampling of featured partner-built agents from Agent Marketplace accessible via the Agent Gallery in the Gemini Enterprise app

Acalvio: ShadowPlex Preemptive Cyberdefense Agent deploys honeytokens in Cloud IAM and in cloud workloads to disrupt agentic AI attacks at the reconnaissance phase. This helps secure cloud workloads from malicious or unsafe agent actions in real time.

Accenture: The Supply Chain Inventory Intelligent Advisor agent is an agentic AI-powered decision support system designed to optimize inventory planning and tracking efficiently. Using real-time Data and AI, it delivers actionable insights to balance service levels and minimize excess inventory.

Adobe: Adobe Marketing Agent for Gemini Enterprise connects natural language queries to Adobe's Customer Experience agentic capabilities. It enables instant insights on campaign performance, target audiences and journey monitoring powered by Adobe, helping teams work smarter without leaving their workflow.

Alteryx: Alteryx AI Insights Agent delivers analyst-curated insights inside Gemini Enterprise. Built on Alteryx One, analysts define trusted datasets while business users simply ask questions to receive governed, repeatable answers. Requires an Alteryx One subscription.

Ambiguous AI: Ambiguous Recruiting Coworker manages candidate communication, multi-round scheduling, and pipeline tracking from application to offer. It works natively across Gmail, Google Calendar, Google Sheets, Google Drive, and more, while looping you in for key decisions.

Amdocs: The Telco Customer Experience Agent autonomously resolves complex customer service issues across digital and contact center channels. Combining Amdocs Cognitive Core with Gemini Enterprise for CX, it reasons, orchestrates, and executes end‑to‑end telco processes, improving resolution speed, cost efficiency, and customer satisfaction.

AODocs: AODocs AI Agent provides instant, trustworthy answers grounded exclusively in the latest validated versions of controlled documents. It supports diverse use cases including standard operating procedures (SOPs), engineering specs, legal policies, and risk management procedures.

Articul8 AI: Articul8 Table Understanding Agent converts complex documents into analytics-ready table data. It reconstructs hierarchical layouts and relationships missed by traditional OCR for accurate analysis. Articul8 Diagram Understanding Agent turns complex diagrams into structured data, interpreting symbols and spatial relationships for automated analysis and seamless workflow integration.

Ascendo: Ascendo AI Agent helps critical infrastructure teams automate workflows, improve diagnostics, and reduce field escalations. By ingesting logs, telemetry, and service manuals, it recommends optimal actions to support faster, accurate decision-making in complex service environments.

Atlassian: Rovo is an AI teammate seamlessly integrated across Jira, Confluence, and your tool stack. Built on Atlassian’s Teamwork Graph, it surfaces knowledge, automates tasks, and takes intelligent actions to help teams pull context, move faster and stay aligned.

AutoCIO: Portfolio managers and investment teams use the AutoCIO Financial Forecasting Agent to design, optimize, and manage portfolios. It automates end-to-end workflows, delivers predictive signals across 50,000+ securities, and enables real-time, hyper-customized strategies via natural language prompts.

Avalara: Avi Agent is an AI-powered gateway for tax and compliance automation. Built on Avalara’s Agentic platform, it smartly coordinates between external systems and compliance agents to instantly and securely calculate taxes, file returns, and manage invoicing.

Backstory: Backstory Revenue Answers Agent tells revenue leaders what's happening on any account, identifies deal risks, and suggests actions. It achieves this by synthesizing CRM data, engagement activity, stakeholder relationships, and public company news into actionable insights.

Botcopy: TrueQ is an AI analyst for government contact centers, enabling supervisors to understand and improve conversational AI performance in the Gemini Enterprise app. It identifies conversation patterns, escalation triggers, and knowledge gaps to improve automation, strengthen governance, and reduce support costs.

Carto: Carto Site Selection AI Agent helps enterprises analyze and visualize prospective commercial locations using spatial analytics. Users ask conversational questions about foot traffic, demographics, or competitors, and results automatically fuel a comprehensive location analysis.

Corvic AI: Corvic AI Agent connects to your enterprise data and delivers production-ready decision intelligence — no pipelines and no data engineering required. Use cases include turning competitive signals and market trend data into actionable intelligence, root cause analysis and customer support automation.

Cotality: The Cotality Payoff Analysis Agent helps mortgage lenders improve customer retention and reduce loan runoff. By analyzing loan originations and payoffs, it delivers actionable intelligence to understand recapture performance, borrower behavior, and competitive insights.

Crescendo Lab Ltd.: DAAC AI Agent transforms complex data into actionable growth strategies for marketing and sales. Using natural language, it delivers instant conversational insights and automated dashboards, effectively eliminating manual reporting and guesswork.

Deloitte: Tariff Management Agentic Suite automates customs regulatory workflows via sub-agents that ingest documentation, extract attributes, and reconcile tariffs against filings. It flags discrepancies to ensure full audit coverage, eliminate manual effort, and recover unnoticed costs.

Devoteam: Demand Sensing Agent helps leaders improve forecast accuracy and optimize inventory. It analyzes sales and real-time market signals to provide immediate demand insights, enabling proactive adjustments. Request for Proposal Agent helps sales teams draft proposals, leveraging centralized knowledge to rapidly generate tailored and compliant proposal drafts.

Dun & Bradstreet: The Dun & Bradstreet Business Verification Agent enables trusted business verification in the Gemini Enterprise app. Powered by Dun & Bradstreet’s trusted data, it resolves businesses to a D-U-N-S® Number and validates identity attributes, enabling front-line teams to move faster and focus on exceptions.

Dynatrace: Dynatrace for Gemini Enterprise integrates AI agents with Dynatrace observability data via Gemini Enterprise Agent Platform and A2A protocol. This enables real-time insights into performance, infrastructure, and user experience, seamlessly enhancing AI workflows while ensuring enterprise-grade security and compliance.

Enigma: Enigma KYB Agent gives compliance teams instant, AI-driven business verification backed by ground-truth business identity from 100M+ registrations and authoritative government records, plus sanctions watchlists and real-time transaction signals.

Eon: Eon AI Agent empowers cloud infrastructure teams to streamline data protection operations. It connects Google Cloud accounts, triggers resource discovery, monitors protected resources, tracks backup jobs, and inspects restore status — all through natural conversation, replacing manual console workflows.

EPAM: Creative Conductor Marketing Agent turns creative briefs into polished promotional videos. It coordinates specialized sub-agents for visuals, voiceover, and music, assembling assets into a final MP4 to deliver fast, scalable, high-quality marketing content.

Exa AI: Exa Agent gives workers a live web view, turning it into a searchable database of code, people, and companies. It searches billions of pages to accelerate tasks like developer tooling, recruiting, and market mapping.

Foxtrot Communications: Forge - AI Agent helps teams eliminate manual ETL pipeline development. Connect any API, and Forge autonomously delivers clean, normalized, analytics-ready datasets, reducing time-to-insight to minutes and freeing engineers to focus on strategic initiatives.

Genpact (UK) Ltd: Genpact Finance One – Revenue Lens and PnL Agents are a suite of agents built on Google ADK and the A2A protocol, enabling finance users to gain actionable insights from revenue and profit & loss data through natural language conversations in Gemini Enterprise.

Genspark: Genspark AI Agent turns multi-step work into finished outputs with one prompt by combining search, web interaction, content generation, and media creation. It helps users execute research, create documents, and automate workflows through a single agent built for real work.

HCLTech: HCLTech ITOps Agent is built on Google Cloud and the HCLTech AI Force platform and enables autonomous IT operations within ServiceNow. Using natural language, it triages incidents, surfaces relevant SOPs, and drives guided remediation to improve overall efficiency, consistency, and operational performance.

Invideo: Invideo empowers marketing teams to scale video production while maintaining consistency. By leveraging brand guidelines and assets, it delivers production-ready ads and short films with built-in quality controls and output in 50+ languages.

Iron Mountain, Inc.: The Iron Mountain InSight® DXP AI Agent bridges Gemini Enterprise and your managed content for unified search and AI-powered Q&A. Its intended purpose is to accelerate results and reduce hours of manual document review to seconds of conversational inquiry.

Lilt: LILT Assist empowers global content managers to transition from manual tasks to autonomous orchestration, managing end-to-end multilingual production and brand governance. Assist scales global content volumes with AI while ensuring enterprise-grade consistency and compliance.

Lovable: Lovable empowers anyone to build real apps and websites by chatting with AI. Teams across product, design, marketing, sales and operations can turn ideas into real, working software.

LTM: LTM Video Intelligence Agent is based on Google Cloud AI technology and converts recorded videos into BDD test cases in a standard format for faster time-to-test and improved productivity.

LumApps: The LumApps AI Agent helps organizations retrieve company knowledge and announcements, trigger workflows, and complete tasks across enterprise systems.

Manhattan: Manhattan Active Agents are AI digital teammates for warehouse management, transportation management, order management, and platform operations. These agents help monitor work, resolve exceptions, guide users, automate tasks, and recommend actions to reduce manual effort; improving speed and accuracy, increasing utilization, and delivering better operational resilience.

Menlo Security: The Menlo Security Orchestrator for Gemini Enterprise reduces mean time to repair (MTTR) from hours to milliseconds. Command global policy and perform deep threat reporting through natural conversation in the Gemini Enterprise app.

monday.com: monday.com agent in Gemini Enterprise turns planning into execution. Teams can quickly build boards, summarize updates, and trigger automations without leaving their workspace—keeping everyone aligned, tracking progress, and driving measurable results.

Nativeorange: LexAI Agent streamlines property and casualty (P&C) underwriting and operations. It ingests submissions, extracts data, and automates risk evaluation, delivering actionable insights on classification and offering next steps to improve speed, accuracy, and decision consistency.

Neo4j: Neo4j Agent translates natural language into Cypher, enabling applications to read from and write to the Neo4j database. Integrated with Gemini Enterprise, it can be used to build agents grounded with GraphRAG and connected enterprise knowledge.

Obin AI: Obin Financial Agent helps private markets and commercial lending teams run financial analysis from Gemini Enterprise. Users submit model spreadsheets and deal-related documents with a scenario description in plain-language. The agent then extracts relevant data and computes financial metrics across modeling assumptions.

Onix: The Onix Risk & Compliance Intelligence Agent automates compliance reviews and risk assessments across regulated industries. It reduces review times from weeks to minutes, proactively identifies regulatory risks, and creates detailed audit trails with citations.

OpenText: OpenText™ Content Aviator™ is a generative AI assistant that enhances content search and productivity. Built on Gemini Enterprise Agent Platform, it allows users to perform natural language searches and efficiently summarize enterprise documents.

Oracle: Oracle AI Database Agent lets you query Oracle data directly from Gemini Enterprise. Available via Google Cloud Marketplace, it delivers fast, enterprise-grade answers without requiring custom chatbots or NL-to-SQL maintenance.

Orion by Gravity: Orion by Gravity is an AI analyst that connects to Looker, BigQuery, and more to proactively hunt for insights. It empowers data teams to identify anomalies and recommend actions that uncover 'unknown-unknowns' in analytics-rich industries.

Palo Alto Networks: Prisma AIRS Model Security agent scans over 35 model formats to eliminate malicious code hidden deep within model layers. This proactive defense goes beyond standard scanning to ensure neural integrity, detecting poisoned weights and backdoors to prevent unauthorized data exfiltration and model hijacking.

Pendo: The Pendo Agent connects your product intelligence to Gemini Enterprise. Query usage trends across SaaS and AI tools, surface adoption gaps, and understand the impact of your AI investments without leaving the Gemini Enterprise app.

Persistent Systems: Ethical Content Auditor Agent helps editors at publishing houses audit political and social content for potential biases and risks. Editorial Assistant Agent enables efficient manuscript evaluation by reading and understanding long manuscripts, analyzing market trends and providing data-backed summaries and recommendations.

Pluto7: Pluto7’s Planning in a Box - Pi Agent orchestrates 50+ sub-agents for supply chain and manufacturing teams. This enables self-healing, autonomous supply chains that optimize demand and production in real-time.

Product Genius: Product Genius brings AI-powered ecommerce intelligence directly into Gemini Enterprise. Share your website URL to get instant insights, actionable policies, and a conversational analyst that helps you optimize your store—no code install required.

Quantum Metric: Felix Agentic for Gemini Enterprise delivers instant insight, contextualized in business objectives, to digital leaders and analytics teams. Leveraging Gemini models, Felix continuously analyzes your digital experience to explain what changed, why it matters, and where it’s costing you.

Red Hat, LLC: Red Hat Lightspeed Agent for Google Cloud helps SREs and admins manage Red Hat Enterprise Linux infrastructure on Google Cloud via natural language. Built on A2A and Gemini, it securely connects with Red Hat Lightspeed services to streamline enterprise IT operations on Google Cloud.

Replit: Replit Agent enables business users to build enterprise-grade applications using natural language. With no coding knowledge required, users turn complex requirements—like a BigQuery-backed customer health dashboard—into functional custom tools in minutes.

S&P Global: The S&P Global Data Retrieval Agent by Kensho enables financial professionals to access cited data from S&P Global's proprietary datasets. It allows analysts and portfolio managers to analyze earnings calls, conduct market research, and instantly retrieve specific financial metrics.

Salesforce: Agentforce Sales for Gemini Enterprise can engage leads, create meeting briefs, surface deal risks and guidance, and manage pipeline and CRM updates — securely and in real-time — without leaving Gemini Enterprise. With the agent handling the grind of manual tasks, sellers can focus on relationship building and strategic selling to win.

Sana from Workday: Sana Self-Service Agent instantly finds and summarizes information from Workday and other knowledge sources, providing personalized answers and assistance with tasks. With over 300 skills across areas like pay, time, and absence, the agent handles everyday HR and finance tasks globally, reducing support tickets and allowing teams to focus on high-value work.

Saviynt: Identity Security for AI enables continuous discovery and inventory of agents, identifying their functions, access, and risks. It treats agents as first-class identities with defined purposes and accountable owners. This blend of posture visibility and enforced accountability converts opaque automation into governable, enterprise-ready identities.

ServiceNow: Now Assist for IT Operations Management operates inside the ServiceNow AI Platform, harnessing the power of GenAI to elevate operational intelligence and automation across IT environments. It introduces a suite of AI-driven capabilities designed to optimize every stage of alert and incident management.

Sheetgo: Sheetgo is an AI-ready data execution platform that transforms spreadsheet-driven operations into governed, reliable pipelines for BigQuery and Gemini Enterprise.

Skyflow: The Skyflow Runtime Data Security Agent helps teams securely deploy agents on Gemini Enterprise Agent Platform. It provides expert guidance and on-demand data protection with fine-grained access controls, enabling enterprises to meet compliance without risking privacy.

Supermetrics: The Supermetrics Marketing Intelligence Agent delivers live, cross-channel performance insights directly in Gemini Enterprise. Connecting to 175+ sources like Google Ads and GA4, it provides fast, hallucination-free analysis using simple natural language prompts.

Synthpop Inc.: The Synthpop Patient Journey Orchestration Agent turns unstructured referrals into validated, payer-ready orders in real time by executing intake and patient-facing workflows. These flows help payers save time as part of Synthpop’s coordinated, production-grade administrative backbone.

Tech Mahindra: Frontline agents use Order Assist to resolve complex orders and service cases faster. By delivering real-time, contextual intelligence through conversation, it reduces handling time, improves first-contact resolution, and enables consistent, data-driven responses.

Teradata: Data Analyst AI Agent enables natural-language analytics on enterprise data in Google Cloud. It translates business questions into governed SQL and Python workflows on Teradata, delivering faster, secure insights without requiring data movement.

Typeface: Typeface Agent acts as a seamless bridge to creative campaigns. It enables instant access to Typeface teams, brand assets, and campaigns via natural language from Gemini Enterprise. Users can trigger high-quality, brand-aligned content creation workflows, streamlining the journey from ideation to production.

UKG: Agentic People Assist — HR Service Delivery elevates the agentic capabilities of the UKG Workforce Operating Platform by empowering employees and managers with knowledge and connected agents to turn intent into action. It orchestrates workflows across time, pay, and HR to deliver real-time, personalized, and measurable outcomes.

WRITER: WRITER Agent is autonomous AI for Fortune 500 enterprises. Built for high-stakes enterprise workflows, it safely plans and executes autonomous work across data and tools based on company context, from brand standards to repeatable playbooks.

XM Cyber: XM Cyber’s PostureAI agent serves as an autonomous posture specialist, providing continuous assessment of Google Workspace applications against security best practices. It is part of a broader Continuous Exposure Management offering that allows organizations to safely adopt AI and secure their critical business processes.

Together, we are creating the world's largest ecosystem of agents — one that will redefine every industry, supercharge every function, and accelerate every business process. I cannot wait to see what we all achieve together.

What’s new with compute: Scaling core and agentic workloads

Wed, 22 Apr 2026 12:00:00 +0000

At Google Cloud Next, we’re announcing a range of compute capabilities to enable your core general purpose and AI workloads for the agentic world with higher performance and lower costs.

Why it matters: IT leaders and builders are faced with balancing compute investments and resources between agentic AI and the general purpose use cases, including the web servers, databases, and enterprise applications that drive everyday customer experiences.

On one side, agents can place unpredictable demand on compute infrastructure, often scaling exponentially. A single user interaction can instantaneously kick off hundreds of concurrent, high-throughput, and low-latency tasks. On the other side, general-purpose workloads generate and hold the data required to fuel the agentic world. Relying on static and siloed infrastructure to run them can risk performance bottlenecks and spiraling costs, leaving your organization unable to respond to surges in demand.

Consider a global travel application where a simple vacation search instantly triggers a massive orchestration of agentic inventory checks, dynamic pricing models, and AI-driven personalized itineraries. Without a modern architecture, this sudden surge in demand can overwhelm the core booking database and bring business to a halt.

We address this with fluid compute, Google Cloud infrastructure that adapts to your general-purpose and agentic workflows, enabling both to win by flexing in performance, capacity, and scale, all in real time. This dynamic flexibility relies directly on the automated orchestration of Google Kubernetes Engine (GKE) and our new Agent Sandboxes to instantly provision secure, isolated execution environments at machine speed.

Let’s take a deeper look at the new compute capabilities announced at Next ‘26.

Run AI and general purpose workloads together

Agentic planning and reinforcement learning depend on highly fluid compute to process unpredictable bursts of autonomous tasks. Relying on static infrastructure to isolate agent-generated code can create severe provisioning delays and heavily inflate your cloud budget. You can remove these bottlenecks by adopting an adaptive cloud foundation. Leveraging GKE Agent Sandboxes empowers your teams to securely launch thousands of execution environments. Pairing these scalable sandboxes with efficient Google Axion processors helps your organization optimize total cost of ownership while fueling artificial intelligence innovation.

Here’s what’s new in Google Cloud compute launches and announcements:

Google Axion N4A is GA: Harness the agility of Google’s custom Arm-based Axion CPUs and achieve up to 2x better price-performance than comparable current-generation x86-based VMs for cost-sensitive workloads such as Java applications, scale-out web servers, and SaaS built by startups, enterprises and partners. Learn more here.
GKE Agent Sandbox, with Axion N4A for price performance, is GA. As the industry’s only native sandbox service among hyperscalers, GKE Agent Sandbox offers scalable and low-latency infrastructure designed for agents to safely execute untrusted code and tool calls without sacrificing performance. With Google Axion, you can build agents on leading infrastructure without compromising on cost or choice. GKE Agent Sandbox running on Google Axion N4A instances provides up to 30% better price-performance than the next leading hyperscale cloud provider. Try GKE Agent Sandbox here.
Google Axion C4A.metal, our first Axion bare metal instance, is in preview: C4A.metal instances power Android development, automotive simulation, CI/CD pipelines, security workloads, and custom hypervisors, without the performance overhead and complexity of nested virtualization. C4A.metal will be GA this summer; learn more here.
C4 instances offer expanded support for Intel Xeon 6 (Granite Rapids) across all shapes: Achieve high-performance for AI workloads like LLM inference and vector search by using Intel AMX with native FP16 support to increase throughput and reduce latency, offering 13% better price-performance versus comparable Intel Xeon 6-based VMs from another leading hyperscaler. C4 VMs are available with Intel Xeon 6 processors across all shapes. Learn more here.
Flexible CUDs expanded support is GA: Shift spend across regions and VM families while optimizing for TCO, with flexible committed use discounts, now with support for a wider range of VM families and services, including memory-optimized (M1-M4) and HPC-optimized (H3, H4D) VM families, as well as Cloud Run. Learn more here.

Here’s what customers are saying:

Unity: Unity is redefining the economics of real-time AI with Unity Vector. By migrating its on-demand feature processor workloads to Google Axion N4A instances, Unity achieved a 20% improvement in cost efficiency without sacrificing latency. As Unity Vector scales to meet increasing demand, the move to N4A instances ensures that Unity continues to deliver industry-leading performance at a sustainable cost.

Deutsche Börse: A leading German market infrastructure provider, Deutsche Börse migrated and modernized dozens of core financial applications onto Google Compute Engine VMs, including latest generation C4 and C4D instances, supporting latency-sensitive Oracle databases and post-trade processing at scale, and boosting release speed, operational agility, and resilience. This delivered the consistent performance they needed to process millions of financial transactions every day and they achieved 58% faster time to market and 33% lower TCO.

WP Engine: WP Engine powers millions of digital experiences where every millisecond matters. By running GKE clusters on C4D and N4D instances, WP Engine has seen up to a 60% reduction in latency for mobile-optimized REST APIs and up to 51% faster processing for data-rich application requests.

eDreams ODIGEO: Operating a high-volume, AI-powered travel platform where every millisecond dictates the customer experience, eDreams ODIGEO migrated its foundational Java-based ecommerce modules on GKE to Axion virtual machines. This immediately eliminated weeks of manual code optimization, delivered a massive 75% improvement in P95 latency with zero code changes, and unlocked price-performance to scale their global services far more cost-effectively than their legacy x86 infrastructure could.

Chainguard: Prioritizing absolute isolation for their foundational software build system, Chainguard deployed the new Axion C4A bare metal instances. This allowed them to establish a strong hypervisor security boundary for package builds, secure their development pipeline with architectural parity, and ensure robust protection, all without compromising build performance.

Run I/O and latency-sensitive workloads together

Both AI and core workloads depend on the ability to store, read, and move data as a single, high-performance operation. Traditionally, these stages are slowed by network and storage limits tethered to vCPU counts, which can starve AI models of the data they need to function. You can remove these constraints by leveraging accelerated Hyperdisk performance for rapid data access and high-performance networking for consistent transit. By allowing your data pipelines to scale independently of compute, your AI training and I/O-sensitive workloads have the dedicated bandwidth they need to remain stable under peak demand.

C4N is in preview: Running high-volume network applications such as concurrent mobile app requests and real-time inventory updates can risk bottlenecks during peak traffic. Maximize your throughput with C4N, featuring Titanium adapters that offload complex packet processing to deliver a market-leading 95 million packets per second — a 40% performance advantage for high-traffic network applications compared to other leading hyperscalers. Designed to rapidly transfer large datasets, C4N provides nearly 400 Gbps of VM-to-VM bandwidth, a 4x improvement in bandwidth-per-vCPU, and achieves an 8x increase in egress network bandwidth through internet gateways compared to C4 VMs. C4N with Hyperdisk Extreme also provides the low-latency, high-speed data access that modern databases and enterprise AI applications need, with 25 GiB/s of block storage throughput and nearly 1M IOPS. Sign-up here for C4N preview access.
M4N is in preview: Running memory-intensive databases can push organizations to overprovision compute cores (vCPU) to meet memory speeds, driving up software licensing fees. We introduced the new M4N series to solve this exact problem. Running Oracle workloads on M4N with Hyperdisk Extreme can reduce TCO by over 20%, enabling you to run Oracle more efficiently, with 26.57 GiB of RAM per vCPU for scale and on far fewer cores. Paired together, M4N with Hyperdisk Extreme delivers the highest per-core IOPS and throughput for high-memory instances across leading hyperscalers. Sign-up for the preview here.
Announcing Z4D: Optimize I/O-intensive workloads and remove network-based storage bottlenecks with new Z4D instances. By securing up to 84 TiB of high-performance local SSD directly on the node, organizations can process massive datasets for SQL, NoSQL, and vector databases. Z4D provides up to 400 Gbps of VM-to-VM bandwidth, matching both C4N and M4N. Z4D virtual machines and bare metal instances will be in preview soon.

Here is what customers are saying:

Ericsson: 5G Core workloads are inherently network-heavy, demanding high-throughput packet processing and deterministic latency that standard public cloud instances often struggle to maintain at scale. By leveraging the Google Cloud C4N, they’ve found the ideal choice for network performance to power Ericsson On-Demand. C4N’s architectural focus on network-optimized compute allows its 5G Core-as-a-Service to reach unprecedented throughput levels, like its recent 1 Tbps milestone, while maintaining the carrier-grade reliability its customers expect.

Teradata: Teradata’s Autonomous Knowledge Cloud enables the world’s largest enterprises to activate enterprise intelligence and turn trusted data into measurable business outcomes. Customers rely on Teradata to run mission‑critical, highly I/O‑intensive analytics at scale where performance and efficiency directly determine value. C4N instances are well suited for these demanding workloads, delivering strong price‑performance and supporting more efficient, optimized deployments. With C4N, Teradata can help customers accelerate insights, scale with confidence, and drive greater impact from their data and AI investments.

Handle demanding storage requirements

Foundational workloads such as web servers, applications, and databases hold the data required to fuel the agentic world. Siloing this critical information on rigid hardware creates bottlenecks that can completely stall enterprise modernization. Imagine a global retail brand running a holiday promotion, but the inventory database times out and drops customer requests because the legacy hardware couldn’t process the sudden flood of agentic queries.

Organizations require the highest performing database hosts backed by high performance IOPS and throughput per vCPU to ensure non-blocking data delivery. Moving these applications to modern cloud infrastructure dramatically improves total cost of ownership and operational throughput. Through strategic cloud migrations, customers can eliminate the architectural walls that stall modernization and unlock their data for AI. Here is what is new in fluid compute for throughput- and capacity-sensitive workloads:

Hyperdisk Balanced improvements. Hyperdisk Balanced enables fast and efficient block storage for general purpose workloads, including applications and relational databases. With Hyperdisk Balanced you can drive up to 2.4 GiB/s and 160K IOPS per volume, higher than general-purpose block storage offerings from other hyperscalers, all while achieving lower mean latency than alternatives. With Hyperdisk Balanced High Availability you can now achieve a 4x performance improvement for high availability databases like SQL Server or PostgreSQL by dynamically routing full disk performance to the active VM, removing the need to overprovision storage. Leverage zero-downtime encryption key rotation and consistency groups for instant snapshots, making it easier to stay more secure. With these capabilities, you can deliver lower TCO, higher performance, and workload resilience for your general-purpose workloads. Learn more here.
Hyperdisk ML performance improvements and Hyperdisk Exapools are GA: With 2 TiB/s of aggregate throughput (up from 1.2 TiB/s), Hyperdisk ML helps eliminate AI storage bottlenecks, offering more than 200x higher throughput per disk than competitive offerings, so your valuable accelerator clusters never sit idle. This allows you to maximize AI compute ROI while powering the next generation of intelligent agents. In addition, for large-scale training needs, Hyperdisk Exapools offer the highest aggregate block storage performance and capacity, per AI cluster, of any hyperscaler. Learn more about Hyperdisk ML and Exapools here and here.
Announcing Z4M: Access up to 168 TiB of local SSD coupled with up to 400 Gbps of network bandwidth, support for RDMA, and bare-metal shapes to run distributed parallel file systems and large-scale AI/ML workloads. Z4M will be integrated with Cluster Director with the option to be colocated with accelerators to provide fast and low-latency access to data. Z4M VMs and bare metal instances are expected to be in preview in Q3 2026.

Here is what customers are saying:

Shopify: During Black Friday weekend sales, Shopify processed over $14.6B and tracked 136 million packages for 81M buyers using its Shop App built on Compute Engine’s Z-series backed storage — without compromising speed or reliability.

HubX: Operating a massive portfolio of AI-powered mobile applications where rapid model loading dictates the user experience, HubX deployed Hyperdisk ML on GKE to eliminate severe I/O bottlenecks. Leveraging this specialized storage layer allowed HubX to support hundreds of concurrent readers and accelerate pod initialization times by 30x during peak traffic surges, drastically reducing idle accelerator costs and helping ensure their complex inference workloads scaled as expected.

Fluid infrastructure for the agentic era

Now, your foundational workloads and agents no longer need to compete for capacity or performance. With Google Cloud’s fluid compute, you get adaptive cloud infrastructure that prevents bottlenecks and enables both your foundational and AI workloads to collaborate and thrive.

Ready to get started? Head straight to the Google Cloud console to spin up a VM for your next big project. Or start planning your migration by checking out Migration Center's AI-powered toolsets to perform cost estimates, create a business case, and evaluate your modernization options.

What’s next in Google AI infrastructure: Scaling for the agentic era

Wed, 22 Apr 2026 12:00:00 +0000

AI is evolving from answering questions to reasoning and taking action. Companies who want to lead in today’s agentic era require computing infrastructure designed and optimized for these new requirements. Today at Google Cloud Next, we are introducing new AI infrastructure capabilities that help you innovate faster, deliver compelling user and customer experiences, and optimize for cost and energy efficiency — all at massive scale.

The shift to agentic intelligence

In the agentic era, a single intent triggers a chain reaction. Unlike chat, a primary AI agent decomposes goals into specific tasks for a fleet of specialized agents that then collaborate, preserve state, and use reinforcement learning to deliver outcomes in real-time.

This process scales intelligence per interaction, but also creates complexity that yesterday’s architectures cannot support without spiraling costs or performance bottlenecks. To scale efficiently and effectively, you must move beyond manually integrating fragmented components and technologies. To deliver agentic experiences that are smart, fast, scalable, and cost-effective, you need a unified infrastructure stack that spans purpose-built hardware, open software, and flexible consumption models.

Google’s AI Hypercomputer is AI-optimized infrastructure built for the agentic era, engineered to deliver on these new requirements. This is the same foundation that powers Google’s flagship Gemini models, consumer AI services, and enterprise AI offerings. Today, we are announcing a significant expansion of our AI infrastructure portfolio, including:

TPU 8t and TPU 8i, our eighth generation TPUs
A5X bare metal instances, powered by NVIDIA Vera Rubin NVL72
Axion N4A VMs, powered by our custom Axion Arm-based CPUs
Google Compute Engine 4th generation VMs, powered by Intel and AMD x86-based CPUs
Virgo Network, our breakthrough data center fabric for AI workloads
Google Cloud Managed Lustre, a high-performance parallel file system
Z4M VMs with high-capacity local SSD storage and RDMA for open parallel file systems
Dedicated KV Cache scalable storage subsystem
Native PyTorch support for TPUs
New Google Kubernetes Engine (GKE) capabilities for agent-native workload orchestration

Taken together, these capabilities will help you accelerate the development of models and complex agentic workflows to accelerate innovation, and deliver useful, responsive services to customers, all while reducing costs and using energy responsibly at scale.

Let’s take a closer look.

Introducing our eighth-generation TPU systems, purpose-built for agentic AI

Today, we’re pleased to announce the eighth generation of our Tensor Processing Units (TPUs), which for the first time includes two distinct chips and specialized systems, engineered specifically for the agentic era.

TPU 8t is our training powerhouse, specifically designed for high-throughput AI workloads. It redefines the scale of AI development, delivering nearly 3x higher compute performance than previous generations to shrink training timelines for massive models. It packs 9,600 chips in a single superpod to provide 121 exaflops of compute and two petabytes of shared memory connected through high-speed inter-chip interconnects (ICI). This massive pool of compute, unified memory, and doubled ICI bandwidth helps ensure that even the most complex models achieve near-linear scaling and maximum system utilization. We can now turn months of training into weeks with the power of 1 million+ TPU chips in a single cluster, orchestrated by Pathways and JAX.

TPU 8i is our breakthrough reasoning system for inference and reinforcement learning (RL), engineered to deliver the ultra-low latency required by agentic workflows and Mixture of Experts (MoE) models. By tripling on-chip SRAM to 384 MB and increasing high-bandwidth memory (HBM) to 288 GB, it breaks the memory wall, hosting massive KV Caches entirely on silicon. Additionally, it doubles ICI bandwidth to 19.2 Tb/s, reduces the ICI network diameter by over 50%, and introduces a dedicated Collectives Acceleration Engine (CAE), which reduces on-chip latency by up to 5x to minimize lag during high-concurrency requests. With this design, TPU 8i delivers 80% better performance per dollar for inference than the prior generation, enabling fast, interactive user experiences, cost-effectively.

TPU 8t and TPU 8i will be available to Cloud customers soon. To learn more, check out this deep dive on the architecture.

A5X with NVIDIA Vera Rubin platform

We know that one size doesn't fit all. Different customers have different workloads, different requirements, and different use cases. So, we also partner deeply with NVIDIA to deliver the latest GPU platforms as highly reliable and scalable services in Google Cloud. We will be among the first to deliver instances based on the next-generation Vera Rubin platform when it becomes available later this year.

We are also co-engineering the open-source Falcon networking protocol with NVIDIA via the Open Compute Project, pushing the frontiers of reliable transport protocols. A5X will implement a variety of innovative concepts from Falcon.

Thinking Machine Labs, for example, uses our NVIDIA-based infrastructure to power Tinker, an open platform for reinforcement learning and fine-tuning of frontier models for specialized use cases, achieving over 2x faster training and serving with Google’s AI Hypercomputer.

Fueling agentic logic and reinforcement learning with Axion, Intel, and AMD

While GPUs and TPUs are great for training and serving AI models, they need to be complemented with high-performance CPU-based services to handle the complex logic, tool-calls, and feedback loops that surround the core AI model. Our new Axion-powered N4A CPU instances deliver outstanding price-performance for these agent runtimes. In fact, GKE Agent Sandbox with Google Axion N4A offers up to 30% better price-performance than agent workloads on other hyperscalers. This efficiency extends across our entire portfolio, including our 4th generation Compute Engine VM families, powered by the latest x86 instances from Intel and AMD. These are specifically optimized for the broadest range of RL tasks, such as RL reward calculation, agent orchestration, and nested visualization, providing the optimal capabilities for every AI workload.

Virgo Network for data center scale-out fabric

As part of AI Hypercomputer, the Virgo Network is designed to meet the demanding requirements of modern large-scale AI workloads. Its collapsed fabric architecture with 4x the bandwidth of previous generations eliminates the "scaling tax" to deliver staggering peak computing power. This capacity helps the most ambitious AI workloads scale with near-linear efficiency.

With Virgo Network and TPU 8t, we can connect 134,000 TPUs into a single fabric in a single data center, and connect more than one million TPUs across multiple data center sites into a training cluster — essentially transforming globally distributed infrastructure into one seamless supercomputer.

We are also making Virgo Network available for A5X (powered by NVIDIA Vera Rubin NVL72), supporting up to 80,000 GPUs in a single data center, and up to 960,000 GPUs across multiple sites.

Storage: Minimizing data bottlenecks

A massive compute cluster is only as effective as the storage system feeding it data. To ensure storage is not a bottleneck while making compute faster, we are delivering four key storage advancements that let you:

Accelerate training and inference: Google Cloud Managed Lustre now delivers 10 TB/s of bandwidth — a 10x improvement over last year and up to 20x faster than other hyperscalers. We’ve also increased its capacity to 80 petabytes. These advancements are powered by our new C4NX instances and Hyperdisk Exapools.
Minimize latency: Managed Lustre can leverage new TPUDirect and RDMA to allow data to bypass the host, moving directly to the accelerators. By removing this processing overhead, your AI agents can respond with the near-instant speed users need.
Maintain peak utilization for training: Rapid Buckets on Google Cloud Storage transforms object storage with sub-millisecond latency and 20 million operations per second. This helps ensure large-scale training checkpoints and recoveries happen near-instantly, allowing your accelerators to maintain 95% utilization or higher, accelerating training cycles, while also providing cost-effective utilization of valuable TPUs and GPUs.
Build custom solutions: For ISVs and organizations that want to build storage solutions, we are launching the Z4M instance, specifically engineered for customers who want to integrate trusted parallel file systems like Vast Data or Sycomp. Each Z4M instance scales to a massive 168 TiB of local SSD capacity and can be deployed in RDMA clusters of thousands of machines.

These new storage options provide a comprehensive storage portfolio, giving you the raw power of the AI Hypercomputer stack with optimal storage services for each use-case.

GKE: Orchestration for agent-native workloads

In the agentic era, intelligence is only as effective as the speed at which it can be scaled. So, we’ve transformed GKE to serve as the premier orchestration engine for agent-native workloads.

Reducing latency across the stack
To support responsive agentic responses, we optimize every millisecond of the start-up and scale-out process. By streamlining how infrastructure responds to surges in demand, GKE ensures that your agents are ready the moment a user engages with the system. New in GKE are:

Accelerated node and pod startup: GKE nodes now start up to 4x faster, while pod startup times have been slashed by up to 80%.
Rapid model loading: Leveraging the run:AI Model Streamer and Rapid Cache in Google Cloud Storage, models now load 5x faster, removing a traditional storage bottleneck.

Intelligent routing with AI-powered Inference Gateway
Building on last year's introduction of GKE Inference Gateway, we are using "AI for AI" to solve the complexities of serving at scale.

Inference Gateway’s new predictive latency boost replaces heuristic guesswork with machine learning-driven, real-time capacity-aware routing. This intelligent orchestration cuts time-to-first-token (TTFT) latency by more than 70% without manual tuning. For businesses, this translates directly into more natural voice conversations and smooth, real-time interactions across a range of use cases.

Inference Gateway can be deployed alongside llm-d, a Kubernetes-native high-performance distributed LLM inference framework, which was recently accepted as a Cloud Native Computing Foundation (CNCF) Sandbox project. Google Cloud is proud to be a founding contributor to llm-d alongside Red Hat, IBM Research, CoreWeave, and NVIDIA, uniting around a clear, industry-defining vision: any model, any accelerator, any cloud.

Open software ecosystem for the full AI lifecycle

Hardware reaches its full potential through co-designed software. AI Hypercomputer enables engineers to move faster by providing native, optimized support for the industry’s most popular frameworks, including JAX, PyTorch, and vLLM. This open software layer reduces friction between development and deployment, translating to faster time-to-market and better resource efficiency.

We are now in preview with select customers with native PyTorch support for TPU, which we call TorchTPU. With TorchTPU, you can run models on TPUs as they are, with full support for native PyTorch features like Eager Mode. When you combine this with our robust support of vLLM on TPU, our message is clear: we always focus on building for openness and customer choice.

Your foundation for agentic growth

To innovate quickly and cost-effectively in the agentic era, you need a unified system that doesn’t compromise on performance or choice. That is exactly what AI Hypercomputer delivers. By co-designing every layer — from the silicon to the software — we remove the integration burden so your teams can focus on driving your business forward.

AI Hypercomputer also serves as the powerful foundation for Google’s entire ecosystem of high-level services. This integrated stack powers everything from Gemini Enterprise to the Gemini Enterprise Agent Platform, ensuring that all these infrastructure innovations translate directly into business value. By leveraging our fully managed services, such as our serverless training service and our new Managed RL API, you can apply AI Hypercomputer’s massive performance gains to customize Gemini with your own business logic, delivering sophisticated, agent-based solutions.

We’re looking forward to seeing what you build next with this updated and expanded AI platform.

What’s new in GKE at Next ‘26

Wed, 22 Apr 2026 12:00:00 +0000

This week at Google Cloud Next ‘26, we are sharing the evolution of Google Kubernetes Engine (GKE), delivering leading performance, efficiency, security, and scale for your most demanding and complex workloads, and the next generation of AI and agentic applications.

Why it matters: Kubernetes has rapidly become the operating system for the AI era, with GKE now powering AI workloads for all of our top 50 customers on the platform, including the largest frontier model builders. We are witnessing a massive acceleration in enterprise AI. In just a few months, the number of multi-agent AI workflows has surged by 327%. At the same time, 66% of organizations rely on Kubernetes to power generative AI apps and agents.

This new era of autonomous agents operating at massive scale requires a foundational change in how we manage infrastructure — a change that is more demanding than the shift from stateless to stateful applications.

What’s new:

GKE Agent Sandbox: Secure, highly scalable, low-latency agent infrastructure
GKE hypercluster: A single, conformant GKE control plane to manage millions of accelerators across Google Cloud regions
Improved inference performance: Foundational enhancements to GKE Inference Gateway and KV Cache management
Reinforcement learning (RL) enhancers: Native capabilities to relieve bottlenecks that throttle accelerator utilization
Scaling on custom metrics: Support for intent-based autoscaling on triggers besides CPU and memory

Read on for details about these GKE announcements.

GKE Agent Sandbox: Accelerating the agentic era

As AI evolves from simple conversational chatbots to entire ecosystems of proactive, autonomous agents, the underlying infrastructure must adapt to handle hundreds or thousands of agents collaborating with workers to plan, evaluate, and execute complex tasks. At scale, infrastructure performance, responsiveness, and rigorous security are essential.

We are excited to announce GKE Agent Sandbox, the industry’s most scalable and low-latency agent infrastructure. Built with gVisor kernel isolation — the same technology securing Gemini — Agent Sandbox allows you to safely execute untrusted code, tools, and entire agents without sacrificing performance. GKE provides leading speed and efficiency for fully isolated agents with 300 sandboxes per second at sub-second latency and up to 30% better price-performance when running on Axion compared to other hyperscale clouds.

Lovable empowers anyone to build apps and websites — with builders creating 200,000+ new projects daily. Lovable runs these AI-generated applications in GKE Agent Sandboxes because of the fast startup, fast scaling and secure isolation.

GKE's cutting-edge sandboxing capabilities allow us to reliably scale to hundreds of secure sandboxes per second, ensuring we can seamlessly empower builders, even during massive, unpredictable demand." - Fabian Hedin, Co-founder, Lovable

GKE hypercluster redefines the scalability ceiling

As foundational AI models grow exponentially and accelerators remain in high demand, organizations resort to fracturing Kubernetes compute infrastructure into hundreds of disconnected clusters, which can create a massive operational burden. To help, we’re announcing the private GA of GKE hypercluster, which allows a single, Kubernetes conformant GKE control plane to manage a million chips distributed across 256,000 nodes — spanning multiple Google Cloud regions. With the GKE hypercluster, widely distributed infrastructure becomes a single, unified capacity reserve that spans geographical locations.

To scale globally without compromising security, GKE hypercluster relies on Google’s Titanium Intelligence Enclave, a software-hardened security engine that delivers private AI compute. This "no-admin-access" model provides hardware-attested, pod-level isolation, so that proprietary model weights and prompts remain cryptographically sealed from platform administrators and infrastructure layers.

Supercharging state-of-the-art inference

Achieving frontier inference requires months of complex performance tuning. To reduce this heavy lifting, GKE now slashes your "time to SOTA" across TPUs and GPUs to mere minutes. We do this with new capabilities:

ML-driven Predictive Latency Boost in GKE Inference Gateway, which can reduce time-to-first-token latency by up 70% by replacing heuristic guesswork with real-time capacity-aware routing — no manual tuning required.
Automatic KV Cache storage tiering across RAM, Local SSD, and GCS/Lustre solves long-context memory bottlenecks. Offloading KV Cache to RAM yielded a more than 40% TTFT reduction and a 50% throughput gain for a 10K system prompt length. Offloading KV Cache to Local SSD yielded an almost 70% throughput improvement for a 50K system prompt length. Learn more about these benchmarks in the llm-d Offloading Prefix Cache to Shared Storage guide.

Built as part of a layered composable suite, these new GKE capabilities leverage llm-d, now an official CNCF Sandbox project. To give you maximum flexibility, we’ve partnered closely with NVIDIA to seamlessly integrate Dynamo for scaling massive Mixture-of-Experts (MoE) models. Whichever tools you choose, GKE provides the highly-optimized, flexible infrastructure you need to safely run any frontier AI workload — including the advanced agentic capabilities of the newly announced Gemma 4.

Eliminating RL compute bottlenecks

Reinforcement learning (RL) is a key driver of AI compute demand and RL jobs involve sequential processing for sampling, reward, and training that can leave GPU and TPU accelerators idle between these RL steps. To streamline RL, we are adding new GKE capabilities in preview:

RL Scheduler solves for the "straggler effect" and inter-batch tail latency, maximizing throughput via intelligent routing.
RL Sandbox provides kernel-level isolation for tool-calling and reward evaluation with millisecond-scale provisioning. Easy integration with RL sampling and reward steps.
RL Observability and Reliability dashboards offer the deep visibility required to troubleshoot and optimize the entire RL loop instantly, out of the box.

Review the RL on GKE recipe, specifically the implementations for Verl and NeMo RL.

Intent-based autoscaling on custom metrics

Traditionally, scaling AI workloads based on application health has imposed a "custom metric tax." To scale the system on anything but basic compute or memory utilization, organizations have to manage complex monitoring systems and IAM roles. This creates operational risk: if your external observability stack fails, your autoscaling breaks along with it.

Intent-based autoscaling eliminates this overhead via native custom metrics support for GKE’s Horizontal Pod Autoscaler (HPA). This agentless architecture bypasses external dependencies by sourcing metrics directly from Pods, hardening reliability while cutting costs. Crucially, it drops reaction times from 25 seconds to just 5 seconds—a 5x performance gain for near-instantaneous infrastructure elasticity.

New workloads, same mission

For over a decade, GKE has set the standard for scalable infrastructure. As we enter the era of agentic and autonomous AI, our mission remains the same: eliminating operational friction so you can focus on innovation. The capabilities we are announcing at Next ‘26 — from GKE hypercluster and the Agent Sandbox, to ultra-fast inference and intent-based autoscaling — give you the secure, efficient, and powerful engine you need to succeed with your ambitious AI workloads. To learn more about using GKE for your AI workloads, check out GKE Inference Quickstart.

What’s new in Cloud Run at Next ‘26

Wed, 22 Apr 2026 12:00:00 +0000

From vibe-coded and large-scale apps to AI models and agents, Cloud Run delivers on-demand compute with zero overhead and pay-per-use pricing for all of your workloads. Last year, the number of external active developers and applications on Cloud Run doubled, with more new customers and apps coming to Cloud Run in 2025 than in its first 6 years combined!

Today, we’re announcing new features and improvements to Cloud Run to help you run your workloads:

Build and deploy full-stack apps in Google AI Studio with Cloud Run, Firestore, and user authentication.
Build, scale, govern, and optimize reliable AI agents with the all new Gemini Enterprise Agent Platform and Cloud Run.
Enable easy deployments from developers and agents with Cloud Run's fully managed remote MCP server.
Combine high-performance inference and serverless compute with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs on Cloud Run.

Empowering the new era of developers

For decades, software development has had an inaccessible learning curve, but thanks to AI, anyone can be a digital builder. With Cloud Run, you can go from prototype to deployed app in seconds.

Build full-stack apps in Google AI Studio
AI Studio now supports full-stack applications that can run server-side code, a Firestore database, and user authentication. Deploy your vibe-coded apps with a single click to Cloud Run, now generally available.

Cloud Run's fully managed remote MCP server
To make it even easier for developers or agents to deploy code, we are launching an official remote Cloud Run MCP (Model Context Protocol) server, giving you the tools to manage and deploy apps. Now GA.

Billing caps
Soon, you’ll be able to define your maximum spend per month. If your bill reaches this amount, your Cloud Run resources will be de-activated.

"Cloud Run has been one of the best technical choices we made in our deployments platform. It is our primary target, powering us to over 1 million live projects being hosted on Replit." - Scott Kennedy, VP of Engineering, Replit

Embracing the agentic era

AI agents are just like people in that they need access to a compute environment to perform their tasks. For a cloud-based AI agent to take complex actions, it can use Cloud Run’s on-demand compute service.

Cloud Run integration with Gemini Enterprise Agent Platform
Through its integration with Cloud Run, Agent Platform helps agents transition from experimental environments into fully managed, production-grade systems without having to rebuild them. Now in preview with select customers.

Cloud Run instances
Traditionally, Cloud Run services, jobs, or worker pools have been opinionated ways to manage Cloud Run infrastructure. Now, we are giving you access to the underlying primitive: you can create individual Cloud Run instances. Coupled with Cloud Storage volume mounts, these instances are ideal for hosting long-running background agents like OpenClaw in one simple command:

code_block: <ListValue: [StructValue([('code', 'gcloud run instances create \\\r\n --image alpine/openclaw:latest \\\r\n --port 18789 \\\r\n --memory 4Gi \\\r\n --default-url \\\r\n --add-volume mount-path=/home/node/.openclaw,type=cloud-storage,bucket=$BUCKET_NAME'), ('language', ''), ('caption', <wagtail.rich_text.RichText object at 0x7f9d6374c1f0>)])]>

This functionality is available in preview to select customers.

Cloud Run sandboxes
Agents often need a safe place to execute code or other commands as quickly as possible. Coming soon, while processing a single request, you will be able to very quickly spin up an ephemeral sandbox that’s strictly isolated from your agent code using a built-in sandbox tool:

code_block: <ListValue: [StructValue([('code', 'app.post(\'/execute\', (req, res) => {\r\n const escapedCode = req.body.code.replace(/"/g, \'\\\\"\');\r\n\r\n exec(`sandbox do -- /usr/bin/python3 -c "${escapedCode}"`, (e, stdout, stderr) => {\r\n res.send({ stdout, stderr });\r\n });\r\n\r\n});'), ('language', ''), ('caption', <wagtail.rich_text.RichText object at 0x7f9d6374ca90>)])]>

"Cloud Run’s concurrency model has been instrumental in simplifying our AI workloads for our customer service AI tool, Lumi. Using Cloud Run alongside Gemini and AlloyDB, we’ve created a unified action layer that enables real-time call summarization and flow guidance, leading to improved first-call resolution rates and faster onboarding for our contact center team." - Edward Wright, Head of Engineering, VirginMedia O2 UK

Automatic scaling for high-demand applications

Cloud Run automatically scales to meet your demand, making it a great fit for large customers that need to serve and respond instantly to heavy traffic spikes.

SSH support for Cloud Run
Developers can now gain secure shell access (SSH) directly into a running Cloud Run container, enabling advanced troubleshooting and inspecting the container's file system on the fly, now in preview with select customers. Open a secure interactive shell session with simple command:

code_block: <ListValue: [StructValue([('code', 'gcloud run services ssh SERVICE'), ('language', ''), ('caption', <wagtail.rich_text.RichText object at 0x7f9d6374cbe0>)])]>

Cloud Run service bindings
Coming soon, unlock seamless service-to-service communication for your scalable microservices architectures with Cloud Run service bindings.

“Cloud Run’s serverless architecture empowers us to meet exponentially growing demand through near-instant scaling, while its streamlined developer experience simplifies building and running our applications.” - Mimi Chen, Member of Technical Staff, Anthropic

Running AI models

From serving custom models with GPUs to training and fine-tuning models with jobs, you can use Cloud Run for your most AI-intensive workloads.

Support for NVIDIA RTX PRO 6000 Blackwell GPU on Cloud Run
We’re bringing the serverless experience to high-end inference with support for NVIDIA RTX PRO™ 6000 Blackwell GPUs on Cloud Run, now GA. This means you can serve up to 70B+ parameter models without having to manage any underlying infrastructure, including scaling to zero when the resource is not in use.

Ephemeral disk
With per-instance temporary disk storage, workloads can process large files or use scratch space without eating up your container memory. Now in preview, ephemeral disk storage is created when an instance starts and deleted when it stops.

"Cloud Run has fundamentally changed how we manage our model deployments. By moving to a usage-based, scale-to-zero model, we’ve eliminated idle GPU costs for low-traffic models. We are now running over 17 model variants in production across multiple regions, each independently deployable and isolated, without the burden of capacity planning or fleet management." - Ajay Nair, Global VP, Elastic

On-demand compute for every workload

Whether you’re a seasoned software developer or a vibe coder looking to deploy the next viral app, Cloud Run delivers on-demand compute for everyone and every workload. Get started today.

Welcome to the agentic BI era with Looker

Wed, 22 Apr 2026 12:00:00 +0000

By combining the analytical depth of Looker with Google’s Agentic Data Cloud, the potential to transform how we model, interact with, and act on our data appears limitless. This week at Google Cloud Next, we are thrilled to share how we are reimagining the business intelligence (BI) stack with Looker for the AI era, starting with a deeper integration with BigQuery.

Our product strategy is focused on delivering trusted data and prescriptive insights for everyone to take actions through Agentic BI.

We’ve rebuilt Looker with Gemini to power a new era of AI agents. With Looker providing the trusted data foundation, your team can innovate rapidly without losing accuracy.

Conversational BI and agents: From conversations to actions

At Next ‘26, we are announcing Looker BI Agents that don’t just provide static answers, but trigger downstream business actions grounded in the Looker semantic layer and your existing enterprise governance framework.

Today, we introduced a number of new agents in Looker:

Dashboard Agents - getting answers to your data directly within a dashboard.

Conversational Agents upgrades (Preview): We are significantly upgrading our core Conversational Analytics agent, which is already GA, with superior reasoning and semantic grounding, helping to eliminate ambiguity. New visibility tools provide admins with end-to-end observability, so that they can monitor performance trends and refine model accuracy at scale. Teams can now also publish Conversational Analytics in Looker directly into Gemini Enterprise to incorporate trusted data into broader workflows.
Dashboard Agents (Preview): Dashboard Agents bring conversational capabilities directly into your BI workflows, providing instant summaries and the ability to ask questions directly within a dashboard.
Embedded Conversational Experiences (GA): You can now embed Conversational Analytics agents directly into your custom applications and internal workflows, allowing users to query data via the UI or API without leaving their specialized tools.
Agentic Workflows in Looker (Preview): Agents can now monitor critical metrics for irregularities and identify hidden correlations and "what’s next" recommendations, so you can address shifts in the business — before they impact the bottom line.

“At YouTube, we're focused on helping creators succeed and bring their creativity to the world. We've been testing Conversational Analytics in Looker to give our partner managers instant, actionable data that lets them quickly guide creators and optimize creator support." - Thomas Seyller, Senior Director, Technology & Insights, YouTube Business

Reimagining self-service: AI power, UI control

AI powered Google-easy exploration for ease and simplicity.

We are making Looker more intuitive and easy for both business users and analysts.

Additionally, we are introducing new AI assistants in Looker Explores to erase the hard parts of BI and transform complex workflows into natural language interactions that feel as familiar as a Google search. New today are:

The new and modernized Looker interface (Preview): A modernized, Google-easy drag-and-drop interface for AI-driven self-service Explores, featuring tabbed dashboards (GA) and paginated reporting (GA).
Visualization Assistant (GA): Use natural language to create and refine charts. Simply ask to change drafted visuals to a stacked bar chart or color-code by region, and Gemini handles the rest.
Expression Assistant (Preview): Generate custom dimensions, measures, and filters just by describing your desired logic — no more having to remember syntax!
Insight Assistant (Preview): Beyond just showing data, Insight Assistant automatically generates summaries and highlights key trends within your reports, helping you identify the so what in seconds.
Self-service Explores (GA): Blend personal data from CSV and Excel files with your enterprise data, while a new interface makes it easier to build models visually. Behind the scenes, Looker handles the LookML code. Together, they provide the agility of a spreadsheet with the rigorous version control of an enterprise platform, so that even ad-hoc data analysis remains governed.

“Our users (or customers) love the flexibility of Excel but fear the lack of governance. Looker’s Self-service Explores give them the best of both worlds. They can upload ad-hoc CSVs for a quick 'what-if' analysis, but because that data is analyzed alongside our LookML dimensions, we know the underlying business logic is still accurate.” - John Pettit, Chief Technology Officer, Promevo

A trusted and open platform

Having a rock-solid, governed semantic layer is a strong way to prevent AI hallucinations and ensure a single source of truth, and we have some exciting announcements in this area:

Open BI and MCP (Preview): Looker supports open-source Model Context Protocol (MCP) via MCP Toolbox, and now we’re offering an all new managed MCP server that’s native to Looker, for ease of customer management. Through this open BI approach you have the flexibility to choose how you deliver AI transformation to your users.
Looker Extension for VS Code (Preview): We are accelerating the developer lifecycle with a new intelligent setup wizard and a specialized LookML AI Agent that translates business intent directly into production-ready code. This new plugin for VS Code and Agentic IDEs lets you vibe-code with LookML, delivering an AI-powered authoring experience for building LookML models simply by describing your desired outcome.

Updated LookML with Gemini support in the Gemini-powered IDEs

Continuous Integration (CI/CD) (GA): Now generally available, this fully integrated advanced CI capability in Looker automates SQL validation and content testing within your development workflow. By identifying potentially breaking changes before they reach production, Looker helps ensure model updates are reliable and accurate.
Knowledge Catalog integration (Preview): With Knowledge Catalog, Looker transforms metadata into a semantic graph, providing the essential context AI agents need to complete tasks autonomously.

The future of data isn't just about seeing insights — it's about acting on them. By combining the power of Gemini with Looker’s trustworthy semantic layer, we are making data-driven actions a reality for every user, in every workflow, every day.

Welcome to the era of Agentic BI.

Get started in your journey to transform your business through the agentic era with Looker at cloud.google.com/looker or start your developer experience at https://docs.cloud.google.com/looker/docs/api-sdk.

What’s new in BigQuery: Powering the Agentic Era

Wed, 22 Apr 2026 12:00:00 +0000

Succeeding in the agentic era requires a transformation in your data strategy: moving from human-scale to agent-first workloads, evolving from reactive intelligence to proactive action, and shifting from raw data to semantic knowledge that agents can use to reason accurately.

For over a decade, BigQuery’s continuous innovations have helped tens of thousands of organizations build a scalable data and AI foundation and navigate several industry and technological transformations. BigQuery has evolved into an autonomous data-to-AI platform, growing over 30x in data processed with Gemini, 25x in AI functions processing unstructured data, and 20x in agent-building tools with Model Context Protocol (MCP).

Customers like Definity are building data platforms to enhance their customers’ experience, improve back-office operations, and boost data team productivity.

“We stood up our data platform in Google Cloud and ingested all critical insurance data in 10 months, which is about half of the time that people see in the industry. The technology that BigQuery provides, processing large amounts of data very quickly, is giving our practitioners and engineers tools that are advanced and a platform that has AI and ML built in. We have doubled the number of users [in a very short period of time].” — Tatjana Lalkovic, Chief Technology Officer, Definity

Today, we are announcing new BigQuery capabilities in lakehouse, built-in AI processing and reasoning, and agentic experiences, all anchored by our commitment to industry-leading price-performance and enterprise readiness.

Open, cross-cloud lakehouse

Enterprise data is often scattered across applications, multiple cloud environments and on-premises. While early lakehouse solutions reduced data duplication, the agentic era demands a foundation that is natively multimodal, cross-cloud, and AI-ready. Our approach blends Apache Iceberg’s interoperability and Google’s differentiated infrastructure with new capabilities, including:

Managed Iceberg tables in Lakehouse (GA, formerly BigLake) enables the openness of Iceberg with advanced BigQuery capabilities, including automatic table management, Iceberg partitioning, multi-table transactions, change data capture, enhanced vectorization, and history-based optimizations.
Iceberg REST catalog enables read/write interoperability (preview) on Iceberg tables between BigQuery, Spark, and other OSS and third-party engines, so you don't have to make complex engine trade-offs.
Cross-cloud Lakehouse (preview) brings BigQuery AI and analytics to other clouds, starting with AWS and Azure. Using open standards like the Iceberg REST Catalog, high-bandwidth networking via Cross-Cloud Interconnect, and transparent caching, BigQuery achieves performance and total cost of ownership comparable to native warehouses, enabling true cross-cloud for enterprises.
Catalog federation (preview) enables easy discovery, analysis, and zero-copy sharing of data across AWS Glue, Databricks, SAP, Salesforce, Snowflake, and Confluent Tableflow (coming later this year).
Real-time data replication that closes the loop between raw data and operational action by allowing you to replicate data from Spanner, AlloyDB, and Cloud SQL instantly into BigQuery tables (GA) and Iceberg (preview).

Analyze data across BigQuery and Iceberg table on AWS using SQL or natural language

Refer to this blog to learn more about our latest lakehouse innovations.

Graph-based reasoning for enterprise agents

For AI to solve truly complex, multi-hop operational problems, business logic must extend all the way down to the data platform layer. Defining logic at this layer ensures that definitions remain consistent and governed from ingestion to consumption. BigQuery Graph (preview) provides the foundation to activate this context, allowing data practitioners to map entities, relationships, and business logic directly within the data platform. This anchors AI agents in a governed reality, enabling them to solve sophisticated challenges at scale with consistent accuracy.

Today, we are announcing new graph capabilities to enhance AI reasoning:

Native support for measures in BigQuery Graph (preview) enables you to unify analytical metrics and relationships into a single, governed entity. It transforms data into a "business map" for multi-hop structural reasoning, allowing agents to move beyond simple searches to trace the ripple effects of business events.
Graph support in BigQuery Conversational Analytics (preview) allows conversational analytics agents to navigate a deterministic business map instead of raw tables to provide answers with higher accuracy. Graphs enable dual reasoning: agents can instantly calculate precise KPIs using measures while simultaneously traversing complex relationships to uncover the "why" behind the numbers.
BigQuery Graph and Looker integration (preview) allows you to reuse measures defined in BigQuery across your data stack by exposing graphs as Looker views. You can also define BigQuery Graphs using Looker with source control and validation. This interoperability ensures that metrics, such as Churn Rate, remain identical across both dashboards and AI agents.
Visual modeling experience in BigQuery Studio (preview) provides an intuitive interface to easily build and manage the entities, relationships, and business logic that power your agentic context.

Analyze data in natural language using Conversational Analytics Agents with BigQuery Graph

Native AI processing to unlock structured and unstructured data

Your data is no longer confined to rows and columns; agents demand a platform that can work across structured and unstructured data at scale without requiring data copies or movement. BigQuery AI makes it easy to perform predictive machine learning tasks (like forecasting, anomaly detection, and recommendations) and generative AI tasks (like entity extraction from images and text, content generation, and data enrichment), with built-in access to over 170 foundational models. New capabilities in BigQuery AI include:

AI.PARSE_DOCUMENT (preview) simplifies complex document processing workflows with a single SQL function that automates Optical Character Recognition (OCR), layout parsing, and chunking.
TabularFM model (preview) brings high-quality regression and classification to BigQuery without the need for extensive feature selection, tuning, training or model management.
ObjectRef (GA) allows you to process unstructured data alongside structured data using SQL and Python. This establishes the foundation for building rich, multimodal context directly on your Knowledge Catalog.
Optimized mode (preview) for SQL-first, AI co-processing-managed functions like AI.CLASSIFY and AI.IF trains task-specific models on the fly, delivering a 230x reduction in tokens consumed compared to row-by-row gen AI processing.
BigQuery-native Gemma embeddings (preview) allow you to generate high-quality embeddings at scale on standard CPUs.
Autonomous embedding generation (GA) fully manages the pipeline for unstructured data, automatically, keeping vector indexes in sync as new data gets ingested.
BigQuery hybrid search (preview) unifies retrieval by integrating semantic and full-text search into a single function, delivering superior precision for RAG and complex exploration.
Python UDF (GA) allows you to enrich, transform or clean data with fully managed Python scalar functions. You can bring your own code or libraries and the functions will autoscale to millions of rows with serverless, scale-out execution. This will be rolled out in phases over the next few weeks.
Connected Sheets brings the scale of BigQuery to the familiar Google Sheets interface, now supporting forecasting with TimesFM model (GA) and anomaly detection (preview).
Geospatial analytics datasets (preview) can now be accessed directly from BigQuery, allowing you to combine geospatial data — such as infrastructure assets and road management insights — with your enterprise data for deeper analysis, without complex manual data wrangling.

Access, process and activate unstructured data in BigQuery

Agentic experiences

AI enables highly skilled data teams to focus on high-impact strategic initiatives by offloading tedious and time-consuming tasks. We are pioneering this space by integrating Google’s foundational research, capable models, and specialized tools directly into BigQuery, providing automation and assistance at every stage of the data lifecycle. New agentic experiences in BigQuery include:

Conversational Analytics in BigQuery (GA) allows teams to query complex datasets using natural language. It provides a secure and transparent path to insights while supporting advanced features like predictive analytics and reasoning across structured and unstructured data. In addition to BigQuery Studio, BigQuery Conversational Analytics agents can be published to and consumed from Data Studio (Preview), Gemini Enterprise (Preview), or via an API for custom applications (GA).

Gain insights from your data using natural language with Conversational Analytics in BigQuery

Proactive agentic workflows (preview) go beyond simple questions to detect metric shifts, perform root-cause analysis to explain why a change occurred, and deliver scheduled research briefings directly to your inbox.

Proactive agentic workflows in BigQuery

BigQuery Agent Analytics offers plugins for ADK (GA) and LangGraph (Preview) frameworks to record agent activity to BigQuery and analyze it for troubleshooting, optimization, and agent evaluation.
BigQuery Studio gets new productivity tools, including a contextually aware assistant (preview) for resource discovery and troubleshooting, SQL Cells (GA) for blending SQL and DataFrames, and Visualization Cells (GA) to create visuals directly within notebooks. The Files Explorer (GA) allows developers to organize, share, and manage code assets in folders. Additionally, we are introducing Git integration and workflows (preview) in BigQuery Notebooks, providing complete SCM coverage across GitHub, GitLab, Bitbucket, and Azure DevOps for data science workflows.
Data Science Agent (GA) in BigQuery notebooks allows you to simply state goals in plain English to automatically execute plans for loading, cleaning, and visualizing data using BigQuery ML, DataFrames, or Spark.
Colab Data Apps (preview) bridge the gap between analysis and action by transforming notebook analyses into shareable, fully managed interactive Python applications that business teams can access from Data Studio.
BigQuery remote MCP server (GA) and BigQuery ADK toolset (GA) reduce the need for manual database connectors for agents.
Google Cloud Data Agent Kit (Preview) provides a portable suite of skills, Model Context Protocol (MCP) tools, environment-specific extensions, and native plugins. By meeting you where you build, like VS Code, Gemini CLI, Codex, and Claude Code, the Data Agent Kit turns your IDE, notebook, or terminal into a native data environment. This enables your environment to orchestrate a massive range of data workflows, automatically selecting the right frameworks (like BigQuery, dbt, Apache Spark, or Apache Airflow) and generating production-ready code.

Unparalleled performance and scale

Modern analytical workloads are increasingly unpredictable and distributed. We continue to invest in enhancing BigQuery's core engine to ensure that as data grows in your environments, your costs and operational overhead do not.

Fluid scaling (GA) enables you to execute highly variable workloads with a premier autoscaling model that does not require a cost-and-performance trade-off. Fluid scaling in BigQuery enables true per-second billing, offering up to 34% cost savings.
Advanced runtime, small query, and history-based optimizations (GA) accelerate native and Iceberg workloads without code or schema changes. BigQuery has improved query speed by 35% year over year while reducing query processing costs by 40% year over year.
New workload management features — including reservation groups (GA), flexible dynamic assignments (preview), and project-level slot and concurrency controls (preview) — provide granular cost attribution and price-performance control, all simplified by declarative, rules-based workload management (preview).
Enhanced observability (GA) and intersection routing (preview) provide flexible disaster recovery capabilities for mission-critical workloads. Agent-powered observability (preview) for turnkey troubleshooting further simplifies operations, while the agent-ready security center (preview) provides a unified, fine-grained access control experience.

BigQuery fluid scaling for unpredictable workloads

In the agentic era, BigQuery isn’t just where your data lives — it’s where your data thinks, reasons, and acts. The future is here, and it’s more open, autonomous, and capable than ever.

Get started on your data and AI journey by taking advantage of our BigQuery data migration offer. We can’t wait to hear how you’re innovating with data.

Cross-cloud infrastructure innovation for the agentic enterprise

Wed, 22 Apr 2026 12:00:00 +0000

The era of agentic AI is accelerating from human- to machine-speed operations, while also creating profound stress on legacy technology infrastructure. This new reality pushes foundational systems to their limits: agents generate thousands of internal messages and complex queries, spawning more agents, all of which can rapidly overwhelm traditional networks and databases, and expose new security vulnerabilities.

Unlocking AI's full potential in the era of agents requires a secure, adaptive foundation. We call it cross-cloud infrastructure for the agentic enterprise – and at Google Cloud Next ‘26, we’re launching a powerful set of new innovations across four areas:

What’s new:

Fluid compute: Google Compute Engine and Kubernetes services work together to enable cost-effective, high-speed AI agents and enterprise workloads with new compute and orchestration capabilities.
Secure cross-cloud connectivity: Agent Gateway, Cloud Armor, and other tools deliver a secure, governed, and simplified networking foundation for AI agents, including observability of agentic traffic across clouds.
Unified data layer: Smart Storage, Knowledge Catalog, and other innovations transform passive data archives into dynamic reasoning engines, giving AI agents the context they need to execute.
Digital sovereignty: Confidential External Key Management and new features in Google Distributed Cloud bring Google’s leading models and AI enablers wherever your data lives.

Let’s take a closer look at all the news for each of these four areas.

Fluid compute

Agentic workloads are dynamic and unpredictable, impacting both traditional enterprise applications and the AI agents themselves. Fluid compute is enabled by Google Compute Engine and Google Kubernetes services working together to dynamically adapt and shift weight in real-time, enabling cost-effective, high-speed AI agents and operational enterprise workloads for all customers.

While our AI Hypercomputer delivers raw power for large-scale AI model training, fluid compute addresses the needs of operational workloads and agents. As agents move toward reasoning and reinforcement learning, CPUs are reclaiming a central role, excelling at the "branchy" logic, complex control flows, and secure execution sandboxes (like those for agentic orchestration, RL, SLM inference, and RAG) that agent workflows demand. CPUs also provide the critical isolation needed for secure agent execution, complementing the parallel processing strength of GPUs and TPUs used in training.

We are introducing new CPU families, GKE capabilities, and Hyperdisk block storage capabilities to run traditional workloads and AI agents securely at scale, including:

Google C4N Series: These VMs help ensure your enterprise workloads don't slow down under the demands of agentic AI by processing up to 95 million packets per second, up to 40% faster than other leading hyperscalers. This eliminates I/O bottlenecks for demanding workloads like security appliances, streaming media, and open source databases, even when utilizing smaller instance sizes.
Google M4N Series with Hyperdisk Extreme: M4N removes data pipeline bottlenecks and eliminates overprovisioning to deliver industry leading per-core IOPS and throughput required to handle massive data I/O from agents, analytics, and mission-critical databases. M4N provides 26.57 GB of RAM per vCPU, allowing you to scale mission-critical workloads cost-effectively on fewer cores. For example, M4N with Hyperdisk Extreme reduces Oracle workload total cost of ownership by over 20% compared to leading hyperscale clouds.
GKE Agent Sandbox: This solution secures agents with trusted gVisor isolation and handles demand spikes, launching up to 300 sandboxes per second, per cluster. Backed by the only managed sandbox technology available among leading hyperscale clouds, it achieves up to 30% better price-performance than competitors when running AI agents on GKE Agent Sandbox with Google Axion N4A.

“Wayfair's AI strategy is built on years of systematic infrastructure modernization on Google Cloud — migrating our core eCommerce engine and databases off legacy systems, decomposing monolithic services into cloud-native architecture, and unifying our data and analytics platform. That foundation is what makes everything else possible. Today, Gemini Enterprise Agent Platform is powering everything from catalog enrichment to generative shopping experiences that help customers create a home that's just right for them — and it's the same foundation preparing us for the agentic era, where AI doesn't just assist but actively drives discovery, personalization, and commerce across every customer touchpoint and across our business.” - Fiona Tan, Chief Technology Officer, Wayfair

Explore all our latest compute innovations in this blog.

Secure cross-cloud connectivity

Agentic AI replaces predictable human requests with autonomous “reasoning loops,” in which agents call other agents that, in turn, call LLMs, triggering massive, sudden surges in compute and machine-to-machine traffic. This shift creates unique challenges for network predictability and security of non-human identities. Optimized for agentic AI, our Cross-Cloud Network moves data across diverse environments, connecting employees, customers, and agents with visibility and security. New in Cross-Cloud Network are:

Agent Gateway: Governs and orchestrates your enterprise agentic traffic as the “air traffic controller” in Gemini Enterprise Agent Platform. It natively understands agent protocols like MCP and A2A to inspect and govern every agent interaction. By integrating with Google and third-party identity and AI safety services, it enables deep inspection to verify access, block attacks, and protect sensitive data, maintaining compliance across your core business.
Cloud Network Insights: Delivers broad visibility across your hybrid and multi-cloud infrastructure to drive faster troubleshooting and network resolutions. Continuously monitor your end-to-end agent, network and web performance across Google Cloud, AWS, Azure, data centers, internet applications, and agentic workloads. Using synthetic traffic analytics, Cloud Network Insights provides hop-by-hop network path visibility to help you pinpoint the source of degradations and is coupled with AI-powered insights from Gemini Cloud Assist to deliver more autonomous operations.
Enhanced Cloud Next Generation Firewall (NGFW) and Cloud Armor: Provides machine-speed, AI-powered protection to combat the rapid explosion of AI-generated polymorphic malware and zero-day exploits. Cloud NGFW advanced malware sandbox delivers real-time inline prevention of AI-generated threats, while Cloud Armor managed rules provides automated protection against both known and unknown Common Vulnerabilities and Exposures (CVEs). Together with Model Armor, these services analyze the intent and content of AI agent communications.

Discover more about how we optimized networking for AI in and outside of the data center.

Unified data layer

AI agents are only as powerful as the data they can access and the context they’re given. More applications and platforms are using structured and unstructured data, but it can be difficult to catalog, find, and act on that data at scale, leading to less effective agent interactions. To close the gap, your agents need all of your data brought together into a cohesive, queryable knowledge engine, or unified data layer. This way, your agents can identify and access accurate sources. At Next ‘26, we’re enhancing the unified data layer with:

Smart Storage: This solution transforms dark data into a powerful knowledge asset for AI agents and training by embedding new semantic intelligence directly into your data objects. With new Google Cloud Storage capabilities like automated annotation, entity extraction, and semantic search, your agents can instantly find and use the specific data they need — whether it's hidden in spreadsheets, PDFs, or other unstructured formats across your entire organization. This significantly speeds up the development and deployment of your AI solutions. Learn more about storage innovations to accelerate your AI workloads.
Knowledge Catalog: Knowledge Catalog maps business meaning across your entire data estate, providing a grounded source of truth so agents can deliver the most accurate results. This foundation enables AI training and inferencing and doesn’t require you to migrate your data; your agents interact with it directly, wherever it lives, with full context and governance, making modernization easier.

Part of our Agentic Data Cloud, Smart Storage and Knowledge Catalog can take your data from a passive archive into a dynamic reasoning engine.

“AI is critical to making our customers’ smart home and security solutions more intelligent and convenient. By leveraging Google Cloud’s Smart Storage, we auto-annotate rich metadata delivered in BigQuery. We’ve scaled and accelerated our data discovery and curation efforts, speeding up our AI development process from months to weeks, continuously delivering innovations that build trust and enhance the overall home experience.” - Brandon Bunker, VP of Product, AI, Vivint

Digital sovereignty

In the agentic era, digital sovereignty is a fundamental requirement for public sector and enterprise customers looking to accelerate innovation — without sacrificing control. There’s no one-size-fits-all solution, which is why we’ve designed a comprehensive set of offerings to meet different sovereign AI needs anywhere: public cloud, on-premises, or hybrid. New capabilities in our sovereign AI portfolio include:

Confidential External Key Management: Organizations can use Confidential External Key Management to maintain complete possession, custody, and control of your encryption keys and the policies that govern them. Confidential External Key Management leverages Confidential Compute to host the key management endpoint in a tamper-proof environment within Google Cloud. You are in control and determine where your keys are stored, who can access them, and under what circumstances. Even highly privileged Google administrators cannot access your keys without authorization, which you can revoke at any time. Your data, your control.
Gemini on Google Distributed Cloud: With Gemini on GDC, companies can securely deploy Gemini in sensitive environments, while meeting data sovereignty needs. Your choice of deployment models includes managed software on your connected hardware or a fully disconnected, air-gapped solution. You can now scale with Google's leading AI capabilities even in the most restricted, high-security environments — from powerful Gemini models to advanced coding, search, and other agentic capabilities.

In addition, Google Distributed Cloud supports an end-to-end AI stack, combining our latest-generation AI infrastructure with Gemini models to accelerate and enhance all your sovereign AI workloads. This stack includes:

NVIDIA Blackwell GPUs: NVIDIA Blackwell (NVIDIA HGX B200) and NVIDIA Blackwell Ultra platforms (NVIDIA HGX B300) GPUs accelerate AI performance, leveraging fifth-gen NVIDIA NVLink to deliver data-center scale bandwidth directly to your environment.
New VM families: New A4 family offerings provide the ability to handle the most demanding inference tasks, delivering a 2.25x increase in peak compute. Memory-Optimized M2 and M3 brings the high memory-to-vCPU ratios needed for massive ERP and data analytics workloads on-premises.
Enhanced storage: Eliminate storage bottlenecks with 6x storage capacity per zone and a 10x performance boost, giving you the ability to do AI reasoning on-premises. Now, your data infrastructure moves at the speed of AI reasoning.

"Our customers demand high-performance, private AI inference without the risks of multi-tenancy. Google Distributed Cloud allows us to provide dedicated, low-latency environments that meet strict sensitive data requirements. With the ability to run Gemini on B200s and B300s, we can significantly increase inference speeds and provide the token throughput our clients need to scale." - Dave Driggers, CEO & Co-founder, Cirrascale Cloud Services

Transforming vision into reality

When these product areas converge, your infrastructure evolves into a high-performing, secure, adaptive foundation for the agentic era. We're not just offering tools; we're providing the architectural blueprint to enable enterprises and the public sector to rapidly embrace the full power of AI and agents with confidence.

To learn more about key industry trends for AI Infrastructure, read our State of Infrastructure in the Agentic AI Era report.

Announcing Spanner Omni: Your infrastructure, Google’s innovation

Wed, 22 Apr 2026 12:00:00 +0000

Today, we announced the preview of Spanner Omni, a downloadable version of Spanner, that expands its industry-leading distributed database capabilities beyond Google Cloud. This enables enterprises to run Spanner in their own data centers, across clouds, or on a laptop — bringing its virtually unlimited scalability, high availability, strong consistency, enterprise-grade security, and multi-model capabilities to AI-enabled applications wherever they operate.

Why Spanner Omni is a big deal

Over a decade ago, Google pioneered the distributed SQL market with Spanner. It offered the "holy grail" combination: the horizontal scalability of NoSQL with the ACID (atomicity, consistency, isolation, durability) compliance and strong consistency of a traditional relational database. Since then, Spanner has evolved into an interoperable multimodal database, incorporating SQL, graph, key-value, full-text search, vector search, and analytical processing with columnar engine. This evolution offers the capabilities needed for the AI era and promises an opportunity to simplify and consolidate customers’ workloads — an advancement that has delivered significant value to our customers.

Customers often navigate across a range of IT environments, from public cloud and hybrid cloud to air-gapped deployments. Spanner Omni meets them where they operate to address the evolving needs of modern enterprises, including:

Business continuity: Mission-critical workloads often need truly resilient, high-availability architectures that extend beyond the edge of the cloud. These workloads need to keep running even when business conditions or world events compel changes in their cloud posture.
Regulatory compliance: Customers in highly-regulated industries, like financial services, must meet regulatory requirements like data sovereignty. While many still maintain large on-premises footprints to stay compliant, they’re eager to modernize for the agentic era.
Application portability: For SaaS providers and other independent software vendors (ISVs), growth depends on meeting customers wherever they are — whether in different clouds or a private data center. They want a consistent technology stack across those environments to minimize development and operational overhead without having to compromise on advanced features for innovation.

Introducing Spanner Omni

Spanner Omni delivers the same core capabilities as our fully-managed Spanner service, with the added freedom to deploy it wherever needed. It provides flexible configuration options ranging from virtual machines (VMs) and Linux containers to Kubernetes clusters. Customers can also run Spanner Omni in a wide range of configurations — on-premises, multicloud, multi-region and hybrid (within reasonable latency topologies), air-gapped, or connected — scaling from a single machine to clusters of thousands of servers. Our internal benchmark tests confirm that Spanner Omni processes millions of queries per second (QPS) across petabytes of data in a single regional deployment.

What makes Spanner Omni unique

Spanner Omni opens the door to new possibilities. Through our work with early adopters, we have observed three popular architectures for deploying Spanner Omni:

Hybrid and multicloud resilience: Customers are running their Spanner managed service in Google Cloud as their primary database while deploying Spanner Omni in a secondary cloud or on-premises data center as a hot-cold failover site. This primary-secondary architecture provides a critical safety net and helps organizations meet disaster recovery and business continuity requirements, especially in regulated industries. Our customers can also deploy multi-regional HA in jurisdictions where Google Cloud only has one data center, but data sovereignty is a requirement.
Unified tech stack across environments: For organizations with a multicloud or hybrid strategy, Spanner Omni provides a "write once, run anywhere" application building experience across environments. By standardizing on a single database layer, teams can drastically reduce their operational overhead and achieve true application portability across any environment, while tapping into cutting-edge database technologies.
On-premises modernization: Organizations with substantial on-premises commitment or preference to self-manage are no longer locked out of cloud-native innovation. With Spanner Omni, they can build or rewrite next-generation AI applications with Spanner’s multi-model capabilities on existing hardware infrastructure using Spanner’s innovative technology to achieve scale insurance, high availability, global consistency, and much more.

The tech behind Spanner Omni

Our vision is simple: If you have compute and a database-grade local file system, you should be able to run Spanner Omni. To make this a reality, we have made a series of innovations to remove Spanner’s dependency on Google. We took the core technologies of Spanner, such as the Paxos consensus, automatic sharding, synchronous replication, and replaced Google-dependent components like Colossus and TrueTime with new creative and innovative solutions.

For distributed storage everywhere, Spanner Omni replaces Spanner’s reliance on Colossus, Google’s distributed file system, by introducing a "Colossus-like" abstraction layer. This layer writes to attached local file systems and makes them available across the network to other nodes. The software automatically handles shard splitting and rebalancing to balance storage across all available servers to ensure peak performance. While Spanner Omni’s file layer isn’t Colossus, it’s a sufficient stand-in to allow Spanner Omni to perform comparably to the Spanner managed service for most workloads.

We also reimagined TrueTime, one of Spanner’s most well known technologies, which uses atomic clocks and GPS to synchronize time globally. For Spanner Omni, we developed a software-based TrueTime alternative, which — like TrueTime in Google Cloud — provides highly reliable, error-bounded time synchronization across servers in a Spanner Omni deployment. While Spanner depends on time synchronization to achieve strong external consistency, its ability to overlap time uncertainty waits with other database work allows Spanner to tolerate weaker uncertainty bounds than TrueTime provides in practice. TrueTime uses this flexibility to provide timekeeping across diverse and heterogeneous environments without limiting Spanner’s availability or performance.

aside_block: <ListValue: [StructValue([('title', 'What customers are saying'), ('body', <wagtail.rich_text.RichText object at 0x7f9d31548190>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Where you can get access

Spanner Omni’s developer edition is available today in preview (download). This edition includes Spanner’s core capabilities — suited for developing and testing in non-commercial, non-production environments. While it excludes enterprise security features, it’s the ideal sandbox for your next project. For early access to the commercial edition with full features, please reach out to us at https://cloud.google.com/consulting/spanner-omni

Small and midsize businesses jumpstart their AI transformations with Gemini Enterprise

Wed, 22 Apr 2026 12:00:00 +0000

Small businesses are the backbone of the global economy. With 400 million SMBs worldwide and 36 million in the U.S. alone, they provide 50% of global employment. Now, with Google Cloud AI, they’re scaling faster, operating more efficiently, and delivering better results for their customers.

For years SMBs have been working with cutting edge products like Google Workspace, Google Ads, Search, YouTube, Maps, Google Wallet, and more to grow their business. Now, more SMBs are taking the first big steps in their AI journeys with our leading models and our agentic platform, Gemini Enterprise.

By making AI accessible to every employee, Gemini Enterprise serves as a launchpad for SMBs ready to develop an AI strategy and build new solutions to work smarter than ever before.Businesses are using AI to automate manual tasks and build custom solutions that extend their reach and help them better support their own customers.

We are proud to support these early milestones, and at Next we’re excited to showcase dozens of new customer partnerships with organizations taking their first steps toward an AI-driven future. Read on to explore their success stories and find our guide to the best SMB resources at Google Cloud Next ’26.

How Gemini Enterprise helps SMBs solve complex challenges

Across the globe, small and medium businesses are using AI to help grow their business. Here’s just a few of the many exciting examples of how SMBs are already benefiting from the secure, scalable, enterprise-grade tools in Gemini Enterprise:

Apex Leaders, a US-based business services company, is using Gemini Enterprise to power an internal search engine, providing easy access to internal data sources and automated information summarization and content drafting for consultant teams.
ARCHINERGY, a US based architecture firm, utilized Gemini Enterprise to create an AI agent that answers questions about architectural design and internal documentation, generating faster insights in just a fraction of the time compared to traditional search methods.
Constructora Las Galias, a Colombia-based construction firm, is using Gemini Enterprise to power a centralized internal knowledge platform, providing streamlined sales and legal workflows and immediate employee access to critical business information.
Eficode, which operates as a software development and DevOps firm within Finland, is using Gemini Enterprise to create a multi-platform support engine that quickly indexes complex internal wikis, so employees can provide customers with faster technical support
Grant Thornton Chile, a leading audit and consulting firm, is using Gemini Enterprise to enhance its agile management by powering secure financial data validation, accelerating HR and sales questions, and enabling unified search across disparate document types.
Grupo Ruiz, a leading Spanish company in sustainable mobility, uses Gemini Enterprise to improve team productivity and process efficiency, achieving an optimized workflow and greater data visibility
Huge, an American business services agency, utilized Gemini Enterprise to create AI agents that automate market research and contract analysis, generating new business intake in just minutes compared to several days.
Incrementa, a Chilean marketing company, is using Gemini Enterprise to automate client reporting and data analysis, helping it scale personalized service and boost campaign reach by 40%.
Koufu, one of Singapore's largest food and beverage companies, which operates and manages a large network of food courts and coffee shops, is using Gemini Enterprise to automate sales reporting and BigQuery data analysis. This reduces spreadsheet bottlenecks so shop managers can focus on front-of-house service quality.
La Maison du Whisky (whisky.fr), a specialized spirits and beverages retailer in France and internationally, chose Gemini Enterprise to create a 'Digital Sommelier' that transforms technical product data into high-end storytelling and marketing copy in seconds
SAI360, a US-based software company, is using Gemini Enterprise to power the design, production, and localization of ethics training, achieving 99% faster course creation and a 95% reduction in training media production costs.
Tauá Resorts group is using Gemini Enterprise to create an intelligent agent that generates real-time insights and dialogue suggestions. This technology allows the sales team to answer complex questions instantly, leveraging data from previous events to craft the perfect proposal for every client.
Tirol, a leading dairy producer in Brazil, is using Gemini Enterprise to create an interactive knowledge base, providing democratized data access and enhanced visibility across its entire value chain.
Wimberly Allison Tong & Goo leverages Gemini Enterprise to streamline digital planning, landscape, architecture and interior design workflows and visual queries, enabling the International Hospitality Design firm to accelerate the transitions from idea generation to design concepts and high-fidelity visualizations.

Practical Resources for Lean Teams at Next ‘26

At Google Cloud Next 2026, we are showcasing how SMBs can enhance their impact by integrating AI into their daily workflows. For attendees focused on maximizing impact with focused resources, we have curated a dedicated track to help you navigate the AI era.

When browsing the Session Library, use the Small IT Team filter to find deep dives into:

Cost Optimization: How to deploy AI effectively and manage resources on a budget that’s right for your goals.
Low-Code AI: Training on how to accelerate your business goals by building custom apps and agents.
Adaptive Security: Proactive defense strategies designed to provide enterprise-grade protection regardless of your security team's size.

Get Started Today

If you can’t make every session, we’ve made it easy to keep the momentum going:

The AI Starter Kit: A streamlined, four-step guide to deploying Cloud AI across your organization.
Monthly AI Newsletter: Sign up to receive the latest industry highlights and Cloud AI updates all year long.

Check out the full SMB guide here and start planning your itinerary. See you there!

New innovations in Google Distributed Cloud

Wed, 22 Apr 2026 12:00:00 +0000

Today at Google Cloud Next, we’re announcing new capabilities in Google Distributed Cloud (GDC) that bring Gemini and our advanced AI stack to wherever your data is, so you don’t need to compromise between AI innovation and sovereignty. This will serve as a catalyst for a sovereign neocloud architecture.

GDC brings Google Cloud to wherever you need it — in your own data center or at the edge. It is offered as two distinct models to meet your specific security and hardware requirements: GDC air-gapped, a fully disconnected deployment that runs on purpose-built, Google-supplied hardware designed for maximum security and compliance; and GDC connected, where you benefit from an integrated, Google-managed software lifecycle on your own hardware.

Traditionally, enterprises and governments with strict data regulatory and sovereignty requirements, were locked out of the latest AI capabilities. Their only choice was to build their own systems, which is slow, complicated, and expensive. GDC ends that struggle. You get world-class AI innovation in your own premises without the toil.

GDC delivers a complete, on-premises AI solution: managed infrastructure optimized for AI workloads, a choice of Gemini or open models for flexibility and efficient Inference services that are cost effective. This foundation allows you to build and run secure AI agents and applications while maintaining total control over your data.

Let’s take you through how the new innovations in GDC come together to support your sovereign AI workloads.

Managed AI infrastructure

To support sovereign AI needs on-premises, organizations require managed infrastructure that can handle the massive performance demands of compute, storage, and networking. Because on-premises AI workloads are dynamic and unpredictable, we are introducing new infrastructure innovations that deliver peak performance across a variety of requirements:

NVIDIA Blackwell GPUs: Accelerate AI performance with NVIDIA Blackwell (NVIDIA HGX B200) and NVIDIA Blackwell Ultra platforms (NVIDIA HGX B300) GPUs, leveraging 5th-gen NVIDIA NVLink to deliver data-center scale bandwidth directly to your environment
Google Cloud machine families: GDC already supports the N2 and N3 machine families for general-purpose workloads, and now it supports the new A4 machine family delivering a 2.25x increase in peak compute to handle demanding inference tasks. We’re also bringing the memory-optimized M2 and M3 machine families to GDC for workloads like ERP and data analytics that require higher memory-to-vCPU ratios.
Enhanced storage scale and performance: GDC now supports 6PB object storage per zone (as compared to 1PB earlier) — 6x the previous storage capacity. In addition, it now offers 30 IOPS/GB (as compared to 3 IOPS/GB earlier) per zone, a 10x performance boost, minimizing storage bottlenecks.

Foundational models in your data center

With GDC, you can bring the power of Google’s flagship Gemini models directly into your environment, bridging the gap between world-class generative AI and strict data sovereignty by enabling native deployment within your own perimeter, now powered by the latest generation NVIDIA Blackwell GPUs.

Today, we are excited to announce that the latest Gemini Flash models are now available (in preview) on the NVIDIA Blackwell and Blackwell Ultra Platforms for GDC connected customers, joining our existing support for GDC air-gapped customers.

"Deploying Gemini on Google Distributed Cloud has significantly improved our global manufacturing. Running frontier AI locally allows us to analyze IoT data for real-time predictive maintenance and quality control, avoiding cloud latency. We maintain strict data sovereignty over our IP while retaining cloud-like agility." - Junhee Lee, CEO, Samsung SDS

AI Inferencing services: Introducing Google Distributed Cloud AI gateway

To optimize performance and abstract infrastructure complexity, we are introducing the AI gateway for sovereign environments. This intelligent middleware acts as the control plane for your models. This provides:

Dynamic request routing: Automatically routes inference requests to the right AI model based on cost, latency, and accuracy, rather than on hard-coded logic.
Intelligent load balancing: Routes requests for optimized inference efficiency, picking GPUs based on their utilization.
Quota management: Prioritizes requests to ensure high-priority applications receive required throughput, and meet quota management goals.
Observability: Built-in tracing and logging for every inference call, helping ensure auditability for compliance-heavy environments.

Agentic AI applications and agents

To truly operationalize AI at the edge, organizations need more than just foundational models. They need autonomous, secure agents built on an agentic AI architecture that can take action. We are thrilled to announce a new sovereign agentic AI architecture for Google Distributed Cloud. Built with 3rd party providers on Kubernetes, this architecture helps to ensure that your agentic workflows execute entirely within your secure Customer Organization boundary.

Using this agentic architecture, you can build and deploy powerful AI agents for agentic tasks like development, coding or data analysis all within your secure perimeter.

AI anywhere with Google Distributed Cloud

We believe GDC is the best platform to serve Google and other models on-prem, connected and air-gapped, enabling all customers to leverage AI and agentic solutions, without compromising on sovereignty. To learn more about these product offerings, visit our website. The innovations we discussed here deliver the flexibility and security required for the sovereign AI era. To see them in action, join our GDC breakout sessions or the Showcase at Next ’26.

Storage innovations to accelerate your AI workloads at Next ‘26

Wed, 22 Apr 2026 12:00:00 +0000

At Google Cloud Next, we are announcing innovations across every layer of our storage stacks — performance, intelligence, and management — to ensure your data is as fast and as useful as the AI models, apps and agents you are building.

Why it matters: Storage is no longer just a place to keep data. When training AI models, storage is the engine that feeds data-hungry accelerators. During AI inference, it’s the access layer that makes it responsive, acting as the source for the context that AI agents need to be effective. When storage performance falls short, accelerators sit idle, agents respond slowly, and data remains invisible to AI models.

But storage performance is only half the battle; you also need storage that’s smart. With the help of Google’s AI models integrated directly into the storage layer, you’re no longer just storing bits, but data that has full context about its content. In this new era of smart storage, raw data becomes a valuable asset that’s ready to use by a variety of downstream AI and enterprise applications.

What’s new:

High-performance storage infrastructure: New Rapid family of features in Cloud Storage for high-performance object storage; delivering 10x performance enhancements plus a new cost-effective Dynamic tier for Google Cloud Managed Lustre.
Smart Storage: Unlocking unstructured data with automated metadata annotation, and AI agent connectivity via MCP.
Storage Intelligence: Streamlined data management through zero-configuration dashboards, aggregated activity views, and enhanced batch operations.
Enhanced ecosystem: Expanded capabilities across Google Cloud NetApp Volumes, Filestore for GKE, and our backup and data protection portfolio.

Let’s take a deeper look at the storage enhancements we are unveiling this week.

Storage infrastructure that keeps up with AI

As AI models scale, getting data from the storage to the compute layer fast enough can be a bottleneck. New storage capabilities move performance directly into the storage layer, reducing total cost of ownership (TCO) and keeping accelerators fully utilized.

Cloud Storage Rapid
Cloud-based object storage like our Cloud Storage is scalable and cost-effective, but bottlenecks can stall AI jobs and waste expensive compute cycles. Every time a training cluster waits on a read or a checkpoint write stalls, you're paying for accelerators that aren't doing useful work.

Cloud Storage Rapid marks a fundamental shift in designing AI infrastructure: you no longer have to choose between the reliability of object storage and the high performance of a specialized AI storage system. Cloud Storage Rapid lets you leverage the industry-leading durability, massive distributed scale, and cost-effective auto-tiering of object storage, while simultaneously achieving extreme throughput, frequent I/Os, and ultra-low latency. With native integrations into PyTorch and JAX, Cloud Storage Rapid is optimized out-of-box for the most popular AI/ML ecosystem frameworks, so that your data preparation, training, and inference workloads run on a high-performance and reliable foundation.

The Cloud Storage Rapid family consists of two offerings: Rapid Bucket and Rapid Cache.

Rapid Bucket, now generally available, leverages Colossus, the Google distributed storage system that powers Gemini and YouTube, to deliver more than 15 TB/s of bandwidth, 20 million requests per second, and sub-millisecond latency in a single zonal bucket. With access via high-performance gRPC and S3-compatible APIs, Rapid Bucket increases accelerator utilization for multi-modal training with 50% reduced GPU blocked time and 2.5x faster data loading. Checkpoint restores are 5x faster and checkpoint writes are 3.2x faster compared to traditional object storage, reducing workload interruptions and wasted GPU time.

Checkpoint writes are 3.2x faster and restores are 5x faster with Rapid Bucket

Rapid Cache, formerly Anywhere Cache, accelerates bandwidth for bursty workloads like model loading for inference, delivering an aggregate read throughput of 2.5 TB/s for existing buckets, with no code changes. The new ingest-on-write feature provides up to 2.2x faster checkpoint restores, allowing training clusters to recover faster from interruptions.

Rapid Cache’s combination of simplicity and performance has resulted in strong adoption, including from cutting-edge AI/ML customers like Thinking Machines Lab.

“Rapid Cache has become a core foundation of our AI/ML data infrastructure, supporting our critical workflows, from data prep and pretraining to training and model loading. By acting as a crucial bandwidth shield and booster, it enables us to scale our data-intensive workloads across our entire fleet without compromise, providing us with the on-demand high bandwidth and consistent stability that we need to innovate at speed.” - James Sun, Member of Technical Staff, Thinking Machines Lab

Google Cloud Managed Lustre
The Lustre parallel file system is the industry standard for organizations whose AI training and inference workloads require high throughput and sub-millisecond latency, and is trusted by AI labs and HPC centers worldwide to feed thousands of accelerators simultaneously and keep them saturated under pressure. Google Cloud Managed Lustre brings that capability as a fully managed service, and with today's announcements, it is the most performant managed Lustre offering available in any cloud.

Managed Lustre now delivers up to 10 TB/s of throughput — a 10x increase since last year and 4–20x higher than managed Lustre offerings from other hyperscalers for a single instance. Powered by C4NX VMs and Hyperdisk Exapools, Managed Lustre writes and restores checkpoints 2.6x faster when compared to other Google Cloud storage solutions.

The new Dynamic tier ($0.06/GB-month) delivers the low-latency performance required for intense AI workloads like training and checkpointing. By serving data from persistent disk rather than relying on object-based caching, we eliminate a performance cliff — helping ensure your data remains responsive and your accelerators stay productive. A single SKU provides simple, predictable billing without the hidden complexity of traditional data tiering.

“By integrating Managed Lustre we eliminated the typical onboarding bottlenecks, allowing us to hit the ground running with the inferencing workload. This high-throughput, low-latency storage keeps our B200 GPUs fully saturated, driving a substantial performance gain in LLM inference over the H200. For our customers, this translates directly into faster, more responsive AI agents that can handle complex reasoning at a fraction of the previous latency.” - Lavnaya Karanam, Software Engineering PMTS, Salesforce

Smart Storage: Context for the AI era

The beauty of an object storage system like Cloud Storage has long been its simplicity: the system knows the object’s name, its size, and when it was created. But if you want to understand the object’s content — what entities it references, whether it contains sensitive PII, or whether it’s relevant to a pending query — you need to use custom pipelines, separate databases, and bespoke enrichment systems.

AI has changed the equation. To fine-tune a model, you need to select the right objects from the get-go, from a corpus of millions. Building an agent requires retrieving the right context for each decision. To meet a compliance obligation, you need to know what every file contains up front, before it becomes a liability. In each case, the bottleneck isn’t compute or model quality — it’s the inability to describe, find, and act on objects at scale.

To bridge that gap between stored and usable data, last year we introduced Smart Storage, a set of capabilities built directly into Cloud Storage that makes every object self-describing. New Smart Storage capabilities include:

Automated annotations, which eliminates the need to build and maintain custom annotation pipelines. With Smart Storage enabled, Cloud Storage can now automatically generate context — including image annotations — so your data is discoverable and usable from the moment it lands. You pay to annotate the data once at write time, and every downstream system can use those annotations immediately for the life of the object.
Cloud Storage MCP server lets you read, write, and analyze Cloud Storage data using the standard MCP protocol.

Smart Storage enables these capabilities, and others, thanks to its object context, now generally available. This metadata substrate adds structured, mutable, IAM-governed context to every object. Customers write their own tags and classifications; Google's annotation pipelines automatically attach labels, extracted entities, and compliance signals.

With Smart Storage, ML teams can select training datasets from semantic criteria without building retrieval pipelines. AI agents can ground their reasoning in enterprise data without a separate retrieval layer.

Storage Intelligence: Data management at AI scale

As data estates grow to hundreds of petabytes, storage costs can spike without warning, and security blind spots can multiply across billions of objects. To manage this, teams have to stitch together multiple tools just to answer basic questions about their own data.

Last year we launched Storage Intelligence to give enterprises a unified management experience built directly into Cloud Storage. Today, 70% of our largest customers use Storage Intelligence, each of whom manage over 50 billion objects.

Storage Intelligence provides a single view across your entire project or organization, with unique capabilities like bucket relocations across regions. Today, we're making it significantly more powerful with:

New zero-configuration dashboards instantly surface cost anomalies and integrate Security Command Center’s Data Security Posture Management (DSPM) data governance feature, to detect critical security vulnerabilities across Cloud Storage — no setup required.
New object events and bucket activity tables in Insights Datasets now drive deeper cost analysis and accelerate operational tasks. You can use these insights to perform a wide range of analyses, from optimizing bucket placement based on egress patterns to quickly troubleshooting 429 errors by finding the impacted objects.
Enhanced batch operations make it even simpler to act on billions of objects with new change ACL and storage class operations, and support for multi-bucket operations.

Enhancing the storage ecosystem

Beyond our core storage offerings, we are streamlining how enterprises migrate to and protect data in the cloud.

Google Cloud NetApp Volumes: With the launch of Flex Unified, NetApp Volumes now provides a unified enterprise storage platform that bridges the data center and the cloud, provisioning both block (iSCSI, NVMe/TCP) and file (NFS/SMB) on the same storage pool. New ONTAP-mode lets you bring your existing automation (Terraform, Ansible) and ONTAP APIs directly to NetApp Volumes.
Filestore for GKE: Developers building AI workloads on Google Kubernetes Engine (GKE) can start small, with shares as small as 100 GiB, and scale capacity and IOPS independently. At the same time, tighter integration to the Colossus distributed file system provides more scale and enterprise capabilities.
Data protection: Google Cloud Backup and DR now features agentic AI capabilities that can autonomously audit your backup estate and remediate coverage gaps, with new GA integrations for AlloyDB and Filestore.

Where to start

As you navigate today’s generational AI shift, you need a storage foundation to support ever-larger, more intelligent, and autonomous models. With new high-performance and intelligent storage layers, plus enhanced storage management tools and a deeper data protection bench, Google Cloud’s storage platforms lets you understand and use your enterprise data in ways that weren’t previously possible, allowing you to:

Reduce the AI data bottleneck: Saturate compute and accelerate ROI. Keep your expensive GPUs and TPUs fully productive with high-throughput storage that delivers the extreme performance required for large-scale training and inference.
Build agent-ready data foundations: Shift from building custom pipelines to an active knowledge base where self-describing objects let AI agents instantly reason over data without manual prep.
Minimize blind spots across exabytes: Replace fragmented management tools with zero-configuration dashboards and datasets to instantly surface cost anomalies and security risks across billions of objects.
Embrace the storage ecosystem: Streamline migration and protection. Bridge your data center to the cloud, scale containerized apps, and automate data resilience with agentic AI.

Visit the Google Cloud Storage console to explore these new features, read more about Cloud Storage Rapid, or explore our Next '26 storage sessions.

What’s new in the Agentic Data Cloud: Powering the System of Action

Wed, 22 Apr 2026 12:00:00 +0000

Companies are shifting from gen AI that simply answers questions to autonomous agents that perceive, reason, and act on their behalf. Attempting to scale these agents on legacy stacks exposes structural failures that can lead to fractured governance, a persistent trust gap, and broken reasoning loops, all while causing costs to spiral.

To solve this, we’re introducing the Agentic Data Cloud: an AI-native architecture that evolves the enterprise data platform from a static repository into a dynamic reasoning engine. It closes the gap between thinking and doing, allowing AI agents to act on your business data and context. While last-generation systems of intelligence were built only for human scale, the Agentic Data Cloud is a System of Action, built for agent scale.

Leading organizations are already using the Agentic Data Cloud to deliver tangible value:

Vodafone has launched hundreds of agents to deliver uninterrupted service to their customers, which is expected to save them millions of euros every year.
American Express is moving a core on-premises data warehouse and hundreds of production applications to BigQuery to power trusted agentic commerce at scale.
Virgin Voyages is using over 1,000 specialized AI agents, including one that slashes mass itinerary rebooking from six hours to just 11 minutes.

Today, we’re announcing three new innovation areas powering our Agentic Data Cloud:

A universal context engine that provides agents with trusted business context to drive higher accuracy.
Agentic-first practitioner experiences to evolve the role of data practitioners and developers as orchestrators of agents.
An AI-native, cross-cloud lakehouse that eliminates data silos by connecting your entire data estate.

Enabling agents with a universal context engine

AI is only as smart as its context. If an agent doesn't understand your definition of "margin" or the intricate relationships in your supply chain, it’s forced to guess. In the age of the agentic enterprise, data alone is not enough, and the old model of governance is insufficient. This is why we evolved the Dataplex Universal Catalog into the Knowledge Catalog, which maps and infers business meaning across your entire data estate, using a rigorous framework of aggregation, continuous enrichment, and search.

Here’s how it works:

Aggregation: To build true context, you must bring it together from everywhere. We are aggregating native context across your Google Cloud and partner data platforms. This includes third-party catalogs, applications, operating systems, and AI platforms like Palantir, Salesforce Data360, SAP, ServiceNow, and Workday (Preview). By using our Lakehouse, your third-party data assets are automatically mapped to the Knowledge Catalog. For Google Cloud sources, we are automating business logic with the new LookML Agent (Preview), which autonomously generates semantics from strategy docs, and BigQuery measures (Preview), which embeds that business logic natively into the platform.
Continuous enrichment: The Knowledge Catalog delivers continuous enrichment by analyzing usage logs across your organization and profiling data behind the scenes. It learns how your enterprise actually uses data, not just what it is. This extends to unstructured data. The moment a file lands in Google Cloud Storage, our Smart Storage (Preview) instantly tags and enriches images and, soon, PDF objects. The Knowledge Catalog also identifies useful collections of unstructured data and uses Gemini to automatically generate missing schemas, mapping complex relationships so your AI is no longer flying blind.
Search and retrieval: Creating a massive context layer is great, but in the agentic era, search has evolved to be the new query path. The hardest problems at enterprise scale are speed, relevance, global reach, and security. To solve this, the Knowledge Catalog uses a sophisticated hybrid search stack built on Google Search innovation. To deliver relevance, it combines semantic and lexical matching with intelligent, machine-learning-based re-ranking. To deliver trust, we are enforcing your security permissions natively with access-control-aware search, so agents can only retrieve and act on the assets they are authorized to see. This high-precision infrastructure instantly identifies trusted context and feeds it to specialized agents.

The Knowledge Catalog now powers the Deep Research Agent (Preview). Part of the suite of Google-made agents available in Gemini Enterprise, this agent can perform multi-step reasoning across Google Cloud data platforms, such as BigQuery, as well as internal documents and web assets to answer complex questions with citations and precision that previously required weeks of manual effort.

Agentic-first practitioner experiences

As we move to this new architecture, the data practitioner role shifts from writing manual pipelines to orchestrating intent-driven engineering.

We’re accelerating this transition with the Google Cloud Data Agent Kit (Preview). Rather than introducing a new interface, we are launching a portable suite of skills, tools, environment-specific extensions, and built-in plugins, that drop into the environments developers love. By meeting practitioners where they already build — including VS Code, Gemini CLI, Codex, and Claude Code — we turn your IDE, notebook, or agentic terminal into a native data environment. This enables your environment to autonomously orchestrate a wide range of business outcomes, automatically selecting the right frameworks (e.g., dbt, Apache Spark, or Apache Airflow) and generating production-ready code based on Google’s gold standards.

This kit doesn't just connect tools — it injects high-performance capabilities directly into the developer's flow, scaling to petabytes without moving data. In fact, the Data Agent Kit features the same skills and tools that power our own out-of-the-box agents, including:

Data Engineering Agent (GA): Builds complex pipeline transformations from scratch and enforces governance rules to keep bad data out of production.
Data Science Agent (GA): Automates the model lifecycle — from wrangling to training — scaling across BigQuery Dataframes and Serverless Apache Spark.
Database Observability Agent (Preview): Acts as a 24/7 guardian for your infrastructure, diagnosing root causes and executing database remediations.

To help ensure the smooth execution of agents, Google Cloud has fully embraced Model Context Protocol (MCP), which provides a secure, universal interface that allows any agent to discover and use your data assets across our core engines, including: BigQuery, Spanner (Preview), AlloyDB, Cloud SQL (GA), and Looker MCP (Preview). MCP for Google Cloud uses our security stack, governing agent interactions based on your existing IAM policies, VPC Service Controls, and data residency requirements.

We’re also reimagining the business user experience with Conversational Analytics, now supported across BigQuery (GA), Cloud SQL, Spanner, AlloyDB (Preview), and Looker (GA). Organizations can simply publish these custom analytical agents in Gemini Enterprise, enabling employees to chat with live data in a familiar interface. By removing the technical barriers, we’ve eliminated the weeks spent waiting for manual reports, allowing businesses to move at the speed of thought.

A cross-cloud foundation built for agentic scale

For an agent to act, it must have a fundamentally open foundation. If an agent is blocked by cross-cloud latency or trapped in a proprietary walled garden, its autonomy is broken. That’s why we are introducing a truly borderless, cross-cloud Lakehouse that liberates your data wherever it resides by:

Connecting analytical estates: We’re integrating Cross-Cloud Interconnect (CCI) directly into our data plane. By combining CCI’s dedicated, high-speed private networking with Apache Iceberg REST Catalog, we’re enabling connectivity across clouds that is low latency and eliminates massive egress fees. As a result, agents can use data across AWS and Azure as if it were local to Google Cloud with seamless cross-cloud access.
Ending proprietary silos: We’re championing open federation to end the era of proprietary catalogs by launching bi-directional federation (Preview). Powered by the Iceberg REST Catalog, engines can now read directly from Databricks Unity Catalog on Amazon S3 (Preview), Snowflake Polaris (Preview), and the AWS Glue Data Catalog on Amazon S3 (Preview). This is reinforced by enhanced Lakehouse Governance (Preview), which ensures your security policies and access controls apply instantly across this borderless environment.
Unlocking operational data: We’re announcing Spanner Omni (Preview), unchaining the most scalable, globally consistent database on the planet. For the first time, you can run the Spanner engine anywhere — across clouds, on-premises, or on your laptop — with the same capabilities used to run Google.
Bridging the insight to action gap: We’re also closing the gap between insight and action. Most "unified" data platforms force the creation of complex ETL pipelines that block agents from accessing your real-time data. With Lakehouse federation for AlloyDB (Preview), we’re removing these pipelines by providing protocol-level, zero-ETL synchronization to give agents access to deep analytical history with low latency in operational transactions.

Automating the future

Moving to agent scale generates orders of magnitude more workloads. To support this, we are announcing four major performance breakthroughs:

Lightning Engine for Apache Spark delivers up to 2x the price-performance over the proprietary market alternative.
Managed Lustre delivers up to 10 terabytes-per-second of throughput to make sure data moves quickly enough for demanding models.
Bigtable now supports an in-memory tier that delivers sub-millisecond read latency for real-time applications. This means you can finally eliminate separate, side-by-side caching layers.
BigQuery fluid scaling helps lower costs by up to 34% on average for autoscaling workloads, scaling up resources instantly when agents act, and scaling back when they don't.

Build your success on a System of Action

The era of passive observation is over. The future belongs to the System of Action, made possible by Google’s Agentic Data Cloud.

Ready to see what an Agentic Data Cloud can do for your business?
Sign up for a strategy workshop today on how to get your data ready to fuel autonomous agents through a System of Action.

Want to learn more about an Agentic Data Cloud?
Read our blueprint for turning passive data into proactive action.

The future of data lakehouse: Open and interoperable for the agentic era

Wed, 22 Apr 2026 12:00:00 +0000

Traditional lakehouses were engineered for the era of reporting, not the high-velocity, multimodal demands of AI agents. To bridge this gap, architecture must evolve into an AI-native foundation — one that replaces batch processing with continuous feedback loops and live data streams. This shift gives agents the reliable context they need to transform raw data into action and unlock all enterprise data (structured and unstructured) across cloud boundaries.

Today, we announced our next-generation cross-cloud Lakehouse that delivers four core breakthroughs:

Fully managed Iceberg storage with enterprise-grade features, giving you the benefits of open-source flexibility plus performance, scale, governance, and multimodal processing.
New cross-cloud interoperability, bringing Google’s high-performance, scalable foundation and AI capabilities to your data, supporting an expansive data ecosystem.
A high-performance Apache Spark experience, accelerating your data science workloads with exceptional performance and your choice of developer environments.
AI-powered, always-on context, enabling AI agents to reason in real time across operational and analytical data.

This agentic-first lakehouse approach can deliver an estimated 117% ROI with payback in under six months. Spotify is already unlocking innovation with Google Cloud’s Lakehouse.

“Spotify is leveraging Google Cloud’s Apache Iceberg products as part of our efforts to build a truly modern data lakehouse that removes the silos between our data lakes and warehouses. This architecture provides us with an interoperable and abstracted storage interface, allowing our teams to process the same data across BigQuery, Dataflow, and other open-source engines without duplication. It will simplify our governance and unlock the ability to innovate at a scale that was previously impossible.”
— Ed Byne, Product Manager, Spotify

Accenture, a key partner in this journey, sees this as a fundamental shift in how enterprises operate:

“To reinvent enterprise operations, organizations must collapse the data boundaries that fragment their intelligence. By utilizing the Google Cloud Lakehouse and 'zero-copy' innovation, we can help customers activate agentic AI with surgical precision. Whether leveraging high-performance Apache Spark for complex data science or delivering scale for industries like retail and life sciences, this AI-native foundation transforms trapped data into real-time action."
— Scott Alfieri, Global Lead, Accenture Google Business Group

Openness without compromise

With Google Cloud’s unique, vertically integrated infrastructure, you get the open-source flexibility of Apache Iceberg backed by a fully integrated, managed data-to-AI experience. You can manage all your multimodal data with unified governance, and get your data estate ready for agents with always-on context. By connecting your Iceberg tables directly to engines like BigQuery and Managed Service for Apache Spark, you can accelerate your AI workloads in real time.

Today, we announced four new innovations to make your Iceberg experience on Google Cloud even stronger:

Fully managed Iceberg storage with read/write interoperability: Experience unified Apache Iceberg tables, managed via the Lakehouse runtime catalog (formerly BigLake metastore). This provides read and write interoperability between BigQuery and Managed Service for Apache Spark including Iceberg-compatible OSS engines like Spark, Trino, and Flink, and third-party engines like Databricks and Snowflake (Preview).
The power of BigQuery with Iceberg: Access advanced runtimes, automatic table management, partitioning, multi-table transactions, and history-based optimization for REST catalog tables (Preview) and Iceberg tables managed by BigQuery Catalog (GA).
A unified multimodal foundation: Use BigQuery ObjectRefs (GA) to merge unstructured data in Cloud Storage with structured data in Iceberg. This simplifies multimodal analysis and manages conversational insights through BigQuery AI.
Unified management and governance: Enhance enterprise trust with open lakehouse governance (Preview) via Knowledge Catalog (formerly Dataplex). Securely ground your agents with business context using end-to-end data lineage, search, quality profiling, and table-level access controls for your Iceberg estate.

Cross-cloud power without the friction

Cross-cloud is today’s enterprise reality, and your agents need a scalable solution to work across data, no matter where it is. Production cross-cloud data access often fails to scale due to high egress overheads and performance bottlenecks. So we are introducing a new high-performance, scalable cross-cloud experience by bringing Google’s AI capabilities to your AWS and Azure data that delivers similar price-performance characteristics to cloud-native solutions:

A new AI-native cross-cloud lakehouse: Lakehouse cross-cloud interconnect and cross-cloud caching (Preview) provides BigQuery and Managed Service for Apache Spark (formerly Dataproc) with high-performance access to AWS Iceberg data at scale. This new capability, powered by high-throughput, low-latency, cross-cloud connectivity and advanced cross-cloud query processing innovations, delivers price-performance characteristics similar to AWS-native data platform solutions. You can run Gemini-powered use cases, such as building agents in Gemini Enterprise and BigQuery AI functions, over your Amazon S3 Iceberg data.
An interoperable ecosystem powered by open standards: To help you easily discover and analyze all your enterprise data across any engine or cloud, we are launching Lakehouse catalog federation (Preview) for AWS Glue, Databricks, SAP, and Snowflake, with Confluent Tableflow coming later this year. This foundation enables simple access to data across clouds through BigQuery and Managed Spark, supported by an expanding partner ecosystem that includes bi-directional access for Databricks, Oracle Autonomous Database, and Snowflake pipeline support for dbt, serving with Clickhouse, and catalog integrations with Atlan and Datahub (preview). And, advanced Lakehouse Governance (Preview), guarantees that security protocols and access permissions are immediately enforced throughout this unified environment.

High-performance Spark, built for enterprise scale

Managed Service for Apache Spark offers a unified, high-performance experience that accelerates everything from data engineering to agentic AI development. It empowers data teams to extract maximum value from their enterprise data without friction by delivering key advantages:

Frictionless, agentic data science: Customers can gain a highly flexible data science environment, integrating Colab Enterprise, Gemini Enterprise, and local IDEs with BigQuery and managed Spark. This allows developers to run Python, Spark, and SQL on a single, unified copy of data, eliminating movement and optimizing engine choice while the Lakehouse runtime catalog Iceberg REST catalog endpoint (Preview) automates table management.
Better Spark processing: Lightning Engine for Apache Spark delivers up to 2x the price-performance over the leading high-speed Spark alternative. This engine uses vectorized execution, intelligent caching, and optimized I/O to provide industry-leading performance on Iceberg, Parquet, and Delta formats without requiring any code changes.

Built for the scale and speed required by agentic AI

Google Cloud’s Lakehouse is a high-performance, real-time foundation businesses can use to scale in the agentic era. We use AI to discover hidden relationships within your enterprise data to provide 24/7 curated context for your agents, making it easier to activate them instantly using databases with new integrations. This includes:

Always-on context for agents: Knowledge Catalog (formerly Dataplex) builds a unified foundation by aggregating business context from your entire data landscape, including Iceberg. It delivers continuous enrichment by learning how your enterprise actually uses data, using Smart Storage to automatically map complex relationships within unstructured files. To guarantee trust and relevance, it uses access-control-aware, high-precision search powered by Google Search innovations. This instantly identifies trusted context and feeds it to AI agents, ensuring they deliver reliable, grounded results.
Ready-to-use BigQuery and Looker agents: BigQuery provides built-in Conversational Analytics, a Data Engineering Agent, and a Data Science Agent (Preview) that work with your cross-cloud, multimodal data. Looker Conversational Analytics provides insights in natural language to your business users. You can build your own agents with your lakehouse as the foundation with Google-native tools like Agent Developer Kit (ADK) and Model Context Protocol (MCP).
Real-time analysis and agentic activation of operational data: Integrate operational data into your lakehouse. Spanner, AlloyDB, and Cloud SQL support real-time change replication into BigQuery (GA) and Iceberg (Preview). Analytical data in Iceberg can also be served with low latency using AlloyDB and Spanner (Preview).

Get started today

Built for the AI era, our cross-cloud Lakehouse delivers uncompromising open Iceberg storage, enables multi-cloud interoperability, and equips your AI agents with always-on context for closed-loop activation. Learn more and start building today.

Introducing the Google Cloud Knowledge Catalog

Wed, 22 Apr 2026 12:00:00 +0000

Traditional data catalogs were built as manual inventories for technical users, focusing on table structures rather than the deep context that AI agents need. When agents lack business semantics and data relationships, this triggers hallucinations, high latency, and stale insights.

To address this problem, we are evolving Dataplex into a dynamic, always-on Knowledge Catalog that serves as the universal context engine for your enterprise, helping agents execute complex tasks with accuracy.

Customers like Bloomberg Media are already using the Knowledge Catalog to power agents with trusted context:

“By unifying Bloomberg Media’s enterprise metadata and business context through the Knowledge Catalog, we successfully launched our Data Access AI Agent. This internal solution empowers stakeholders across the organization to intuitively explore our data lake, translating complex business inquiries into instant, AI-driven narratives. Crucially, by grounding our AI in trusted institutional context, we ensure confidence in the accuracy and quality of every insight generated."

— William Anderson, CTO, Bloomberg Media

The Knowledge Catalog operates on three foundational pillars:

Aggregation: Unifying context and resolving conflicting definitions
Enrichment: Generating continuous meaning and mapping relationships
Search: Empowering agents with high-precision retrieval

Aggregation: Unifying context across your data estate

To build true context, you must bring it together from everywhere. The Knowledge Catalog aggregates native context across your Google and partner data platforms, semantic models, and third-party catalogs, unifying them into a single, governed source of truth.

Broad metadata aggregation (GA): To build a truly comprehensive context engine, you must leave no silo behind. The Knowledge Catalog automatically harvests technical metadata across your foundational systems — including BigQuery, AlloyDB, Spanner, Cloud SQL, Firestore (Preview), and Looker (Preview). It also supports integrations with third-party databases and partner catalogs like Atlan, Collibra, Datahub, Ab Initio, and Anomalo, ensuring that even your legacy metadata is brought into the agentic fold.
Enterprise connectivity (Preview): To truly understand your operations, semantic context must cover all the key systems in your enterprise. Using Google Cloud Lakehouse, we are interconnecting these systems with context federation, and Knowledge Catalog gains full and immediate visibility to applications, operating systems, and AI platforms — including Palantir, Salesforce Data360, SAP, ServiceNow, and Workday. For example, your SAP data products are automatically mapped to the Knowledge Catalog.
LookML Agent: We are automating how you define business logic. The new LookML agent autonomously reads your strategy documents to instantly generate business-ready semantics. By aggregating these semantic models into the Knowledge Catalog, we federate your core business logic across the enterprise, ensuring agents reason using the same definitions as your analysts. We are giving developers our new VS Code Extension for LookML semantic models, and the entire semantic layer lifecycle from any agentic IDE.
BigQuery measures (Preview): We are redefining data consistency by embedding programmatic business logic directly into the SQL engine. BigQuery measures ensure every calculation is universally reusable and mathematically accurate. The Knowledge Catalog acts as the ultimate aggregator, pulling BigQuery measures and LookML together into a single, governed semantic foundation.
Data products (GA): Data products package data assets and context that grounds agents and makes them reliable in production. These self-contained blocks include built-in intent, SLAs, and governance constraints, providing the essential building blocks to scale complex AI use cases.

Enrichment: Generating meaning through continuous learning

The Knowledge Catalog provides continuous data enrichment — going beyond manual curation to actively mine structured schemas, query logs, and BI semantic models while extracting entity relationships from unstructured data. We are delivering this continuous enrichment where your teams work:

Smart Storage and Object Context API (Preview): Built natively into Google Cloud Storage (GCS), Smart Storage automatically tags, embeds, and enriches files with metadata as soon as they land in your buckets. By integrating this intelligence feature with the Knowledge Catalog, unstructured data is instantly discoverable by agents.
Deep multimodal metadata extraction (Preview): For collections of complex unstructured data, the Knowledge Catalog natively integrates with Gemini to identify useful business information and automatically build pipelines that extract entities and map complex business relationships directly from unstructured content.
Automated context curation (Preview): The Knowledge Catalog automatically generates natural language descriptions, including business glossaries, for datasets, data products, relationships, and verified SQL patterns that allow both humans and agents to interact with data without guesswork. By inferring these hidden relationships and intent-based patterns, it constructs a dynamic, evolving map of how data actually relates to the business.
Verified queries and semantic guardrails (Preview): One leading cause of AI failure is hallucinated logic and guessed SQL joins. To prevent this, the catalog provides verified SQL patterns and pre-generated natural language questions.

Search: Unleashing agents with high-precision, secure retrieval

Creating a massive context layer is great, but in the agentic era, search has evolved to be the new query path. When autonomous agents are working on your behalf, they are iterating incredibly fast. The hardest problems at enterprise scale are speed, relevance, global reach, and security.

High-precision semantic search (GA): The Knowledge Catalog uses a hybrid search stack leveraging decades of Google innovation. Built on the same advanced query-rewriting and machine-learning technologies that power Google Search, it delivers the sub-second latency and pinpoint relevance that agents need. When an agent receives a prompt, the catalog instantly ranks and returns the right context to agents in real-time.
Access control-aware search: The ability to find the right data and its corresponding context is critical in the agentic era; if an agent retrieves the wrong context, it hallucinates. To gain trust, our global search respects metadata access permissions as defined in the source systems, ensuring agents can only retrieve and act on the assets they are explicitly authorized to see.
Measurable context evaluation: To ensure long-term accuracy, we are augmenting our search capabilities with a robust evaluation framework. This transforms context construction from a guessing game into a measurable engineering discipline. It allows your teams to quantitatively test and iterate on various context construction strategies, ensuring continuous optimization of the relevance and quality of the context feeding your agents.

With foundational data products, high-precision search, and guardrails in place, we can deploy advanced AI reliably. A prime example is the Deep Research Agent in Gemini Enterprise, powered by the Knowledge Catalog (Preview). Now natively powered by the Knowledge Catalog, this agent synthesizes live business data, internal documents, and web research to answer highly complex questions. It delivers deterministic precision and deep citations, executing tasks in minutes that previously required weeks of manual effort.

Stop forcing your agents to guess the unwritten rules of your business. Build the context once, and unleash your agents to do the rest.

Get started today with Knowledge Catalog.

What’s new with the Cross-Cloud Network at Next ‘26

Wed, 22 Apr 2026 12:00:00 +0000

While generative AI sparked a revolution, the true paradigm shift is the rapid evolution from standalone AI models to multi-agent autonomous systems. In this new era, the network transcends basic connectivity to become the critical integration layer for your agentic enterprise.

As AI agents and services surge, your core applications remain as vital as ever. To thrive in this rapidly evolving landscape, you need a planet-scale network to connect, protect, govern, deliver, and secure all your users, data, agents, AI services, and core applications across clouds and on-premises.

Google Cloud's Cross-Cloud Network provides this unified foundation, and is now used by 65% of the Fortune 100 and handles up to 27 exabytes of data per month. At Google Cloud Next, we are introducing networking innovations to accelerate your AI infrastructure, strengthen security, and simplify operations.

Optimized networking infrastructure for AI

As we move toward an agentic world, the network must support massive-scale inference paired with reinforcement learning. At Google, we’ve spent years refining this cycle to power our own global AI services. Today, we’re announcing AI infrastructure network innovations that bring this same architecture directly to your workloads, across agents, inference, training, and beyond.

Networking for agents

The Gemini Enterprise Agent Platform is a comprehensive enterprise environment designed to build, scale, govern, and optimize the next generation of autonomous agents. Key innovations being announced in preview include:

Agent Gateway: Air-traffic control for agentic traffic

Agent Gateway understands MCP and A2A agentic protocols and provides an open, extensible, scalable way to enforce centralized governance policies to securely connect agents, models, and tools across runtimes.

Ambient networking: A seismic shift in service-to-service connectivity

Ambient networking, a new integrated data plane for Google Kubernetes Engine (GKE) and Cloud Run, provides service discovery, zero-trust access, and traffic management without the need for complex and resource-heavy sidecar proxies. It reduces operational overhead and enables up to a 10x reduction in GKE resource usage for layer 4 (L4) mesh capabilities

Ambient networking underpins two new capabilities:

Service bindings automatically establish service-to-service connectivity, allowing developers to move faster to build and scale their agentic applications and services.
Network Services Monitoring bridges application and network observability gaps resulting in faster root-cause analysis and simplified troubleshooting.

Rich partner integrations and customizations

With the help of Service Extensions, we are developing solutions for identity, governance, and AI security for agent-to-anywhere traffic. Coming soon in preview to Agent Gateway are:

Identity and governance administration: Offering delegated authorization to Cloud IAM and partner services from Okta, Ping, Saviynt, and Silverfort to enforce real-time, contextual governance policies based on application and business context.
Runtime security: As a universal enforcement point by integrating with Google Cloud’s Model Armor and partner solutions from Broadcom, Check Point, Cisco, CrowdStrike, Exabeam, F5, Netskope, Palo Alto Networks, Thales, and Zscaler. Together, these can help to secure agentic communications against emerging AI attack vectors.

These innovations are built on an open foundation including Envoy and Kubernetes, providing strong, integrated governance in multicloud environments using standard Kubernetes Gateway APIs.

Networking for inference

At Google we run inference at scale with optimized use of distributed GPU and TPU resources, automatic failover between regions for high availability, and optimized global request routing for fast end-user performance. GKE Inference Gateway delivers these capabilities to our cloud customers including the following new innovations:

Multi-region support allows scaling inference services across regions, enabling cross-regional failover, optimized utilization, and reduced global latency (preview).
Predictive latency boost improves utilization with intelligent request routing based on predefined performance targets (preview).
Disaggregated serving leverages llm-d’s SGLang support, offering the flexibility to choose between vLLM and SGLang for model serving (GA).

Gemini Enterprise Agent Platform reduced Time to First Token (TTFT) latency by over 35% for Qwen3-Coder by using GKE Inference Gateway.

“Before GKE Inference Gateway, managing our inference stack with Ray Serve created a complex, dual-orchestration layer that was a significant burden on our small operations team. Moving to the Inference Gateway and native Kubernetes deployments was the 'North Star' architecture we needed to simplify management and achieve robust production stability with a GKE-native batteries-included solution.” - Mikhail Lubinets, Lead HPC Engineer, Technology Innovation Institute

Networking for training

At Google, we build and run the largest AI models in the world — and we built a network to support that. Some of the new enhancements are:

Massive scale with Virgo Network

This new non-blocking data center fabric removes latency barriers:

Virgo can link up-to 134,000 chips with 47 Petabits/sec of non-blocking bi-sectional bandwidth in a single fabric. This delivers a staggering 1.6M Exaflops of FP4 compute.
With enhancements in Pathways and JAX, you can further connect these Virgo fabrics to scale to over 1 million TPU chips in a single training cluster.
We are also making Virgo Network available on NVIDIA Vera Rubin NVL72, supporting up to 960,000 GPUs.

For more on Virgo Network, check out this blog.

Accelerator network profiles

It’s easier than ever to handle the complex networking prerequisites for accelerator-equipped GKE node pools with DRANET, which improves bandwidth for distributed AI/ML workloads by up to 60% (GA).

AI-native Cloud Interconnect

SLA-backed, and optimized for efficiency, Cloud Interconnect supports petabit-scale data transfers and is available with a fixed price option. Cloud Interconnect now supports:

400 Gbps circuits with up to 3.2 Tbps in a single connection (GA)
Partner Cross-Cloud Interconnect for AWS (GA), CoreWeave (in preview soon), and Lumen (in preview soon)

Cross-Cloud Network for AI and core applications

The Cross-Cloud Network helps ensure you can securely connect users, data, locations, applications, services, and infrastructure anywhere in the world, at planetary scale. We designed our global multi-shard network to scale horizontally to meet the demands of the AI era and enable us to accommodate our 10x WAN traffic growth from 2020 to 2025.

These are some of the improvements we’re making to the Cross-Cloud Network:

Ultra Low Latency Solution for financial exchanges

In partnership with CME Group, we are bringing the world's leading derivatives marketplace to Google Cloud. To support CME Group’s performance requirements, we developed an ultra low latency (ULL) networking and compute solution. This fully managed cloud environment will allow CME Group and its clients to migrate its core trading systems to Google Cloud.

Now in preview, the solution is designed to meet the unique and exacting requirements of running financial exchanges in the cloud. It includes several new technologies:

Deterministic high-performance compute powered by ULL networking, with bare metal and VM form factors, delivers a comprehensive portfolio for your trading compute needs.
Scalable multicast data distribution with hardware-based ultra-low latency enables reliable one-to-many market data sharing.
Nanosecond-level clock sync enabled by Firefly, a novel clock synchronization system. Firefly achieves sub-10ns NIC-to-NIC synchronization to support high-frequency trading.
Advanced network observability with 64-bit nanosecond timestamps, support for multiple traffic-mirroring destinations and multicast traffic, and support for auditing and regulatory requirements.
Low-latency inference allowing exchange participants to connect their AI-driven services to the exchange’s infrastructure.

“The Google Cloud Ultra Low Latency Solution provides the level of performance necessary for CME Group futures and options markets to run in the cloud, expanding access to clients worldwide.” - Sunil Cutinho, CIO, CME Group

Cross-cloud observability for networks, applications, and agents

Whether you’re running core applications or new AI agents, you need visibility into your network infrastructure. Cloud Network Insights, now in preview, offers network performance monitoring (NPM) and digital experience monitoring (DEM) to dramatically reduce the mean time to detect and mitigate network-related agent, application, and API issues.

Cloud Network Insights is enabled by technologies from Broadcom’s AppNeta and powered by AI-enabling natural language queries through Gemini Cloud Assist.

"In an environment as complex and high-scale as Sabre’s, total visibility isn't just a luxury — it's a requirement for operational resilience. Cloud Network Insights will enable us to further shift our posture from reactive troubleshooting to proactive optimization. By providing granular, real-time telemetry across our global cloud footprint, it helps eliminate the traditional 'black box' of the network, allowing our teams to resolve bottlenecks before they impact the traveler experience." - Alfredo Rodriguez, VP Cloud Platform Infrastructure, Sabre Corporation

“Cloud Network Insights closes the 'visibility gap' between the private corporate network and the public cloud, empowering our joint customers to pinpoint performance bottlenecks in seconds rather than hours.” - Alan Davidson, CIO, Broadcom

Cross-Cloud Network for distributed applications

Multicloud and hybrid networks require secure, reliable, and high-performance connectivity. New enhancements for our foundational networking services and tools include:

Private Service Connect

Private Service Connect traffic volume grew 4x in 2025 and it now supports 40+ Google and third-party published services, enabling secure private global access to your managed services.
Private Service Connect endpoint-based security allows for granular authorization policies for producer-to-consumer service communications (preview).
Gemini Cloud Assist for Private Service Connect provides for automated troubleshooting (preview).

Cloud-native IP address management (IPAM)

Cloud Number Registry is an IPAM solution powered by agentic technologies. Network admins can easily find free IP ranges, track utilization, and allocate resources (preview). It also integrates with Infoblox Universal DDI for Cross-Cloud Network IPAM discovery and enforcement.
Hybrid Subnets allow you to migrate legacy workloads from on-premises to a VPC without needing to change hard-coded IP addresses (GA).
Cloud NAT allows you to connect your IPv6-only workloads to private IPv4 destinations using the combined power of DNS64 and private NAT64 (in preview soon).

Network Connectivity Center (NCC)

Partner Cross-Cloud Interconnect for AWS is available as a connectivity type in NCC (preview).
Support for static routes using an internal load balancer as the next hop allows the integration of Secure Web Proxy and third-party network security virtual appliances (GA).
Support for privately used public IP (PUPI) allows the exchange of PUPI IPv4 addresses with VPC spokes and producer VPC spokes (GA).

Granular networking charge visibility

Cost Explorer and the new App Optimize API now provide attribution of associated Data Transfer costs to the originating resources for Google Cloud products (in preview soon).

Cross-Cloud Network for internet-facing services

As part of Cross-Cloud Network, the Global Front End simplifies how you deliver, scale, and protect web, API, and AI workloads. New capabilities include:

Global Front End Enterprise delivers simplified consumption by combining capabilities from global Cloud Load Balancing, Google Cloud Armor, Cloud CDN, and Service Extensions with up to 15% lower TCO (in preview soon).
Post quantum cryptography (PQC) helps secure your workloads with industry-standard algorithms that provide a layered defense against both classical and quantum adversaries.
Google tag gateway, enabling advertisers to serve tags from their own domain, which can significantly improve the accuracy and resilience of measurement signals (GA soon).

In addition, Cloud CDN, an important part of the Global Front End, now offers:

Built-in image optimization to help you deliver content that best fits your end users’ screens and saves on bandwidth costs (in preview soon).
GKE Gateway support so you can enable and manage caching services using GKE APIs (GA).

Cross-Cloud Network’s Cloud WAN for global enterprises

Cloud WAN is a fully managed, reliable global backbone to connect your enterprise. New capabilities include:

Expanded geographic reach: Our network spans more than 10 million kilometers of terrestrial and subsea fiber, and Network Connectivity Center’s site-to-site data transfer is now available in over 25 countries.
NCC Gateway enables third-party secure service edge (SSE) integrations from Palo Alto Networks (GA soon) and Symantec (preview).
The Verified Peering Provider program, which offers highly reliable internet connectivity to Google, now has dramatically expanded availability through 175+ providers worldwide.
Last mile connectivity: Provision site-to-cloud private connectivity in minutes with preferred partners from the Google Cloud console (in preview soon).

“Cloud WAN enables Dun & Bradstreet to evolve our global network via composable, cloud-native constructs. Leveraging NCC, we’ve built a resilient, high-performance platform that simplifies operations and optimizes costs. This foundation supports continued modernization and AI-driven workloads. We expect to extend this architecture as new patterns emerge, maintaining our blueprints-first approach.” - Josh Barry, VP, Network Engineering, Dun & Bradstreet

AI-powered security against evolving threats

The threat landscape is evolving faster than ever, with AI-driven attacks. Staying ahead requires the latest defenses. Cross-Cloud Network relies on Cloud NGFW and Cloud Armor for advanced security capabilities. Here’s the latest on those offerings.

Cloud NGFW

Advanced malware sandbox uses AI models trained on data from 70k+ customers to stop 99% of known and unknown malware, including evasive zero-days. Advanced malware sandbox is powered by Palo Alto Networks Advanced Wildfire (in preview soon).
Internal Application and proxy Network Load Balancer support helps to enforce consistent, service-centric security for abstracted services like GKE, Cloud Run, and Private Service Connect traffic (preview).
Project-level policies allow for creating and managing Cloud NGFW endpoints, security profiles, and security profile groups at the project level (in preview soon).

Cloud Armor

Managed rules, built-in rulesets across 15 threat categories, deliver automated threat protection against a broad set of attacks and zero-day CVEs. This is powered by Thales Imperva based on visibility to 1.5 trillion web requests each month (in preview soon).
Google Cloud Fraud Defense integration helps to discern the legitimacy and authorization of bots, humans, and agents. Fraud Defense is the evolution of reCAPTCHA, which protects over 14 million domains from fraud and abuse.
Adaptive protection for Network Load Balancers & VMs brings advanced machine learning to L3/L4 traffic, to detect and mitigate volumetric DDoS attacks (in preview soon).
A simplified user experience with a visual rule builder makes custom rule creation easier (in preview soon).

AI-powered network operations

Finally, new AI-powered technologies in Gemini Cloud Assist can help automate manual tasks, ease troubleshooting, predict reliability issues, improve security, and help optimize your network to reduce toil and improve reliability with new specialist agents. These include:

A network security agent that streamlines network security operations by assisting with policy generation, recommendations, and impact analysis (in preview soon).
A network agent that optimizes workload placement for performance and reliability, and also provides advanced cost estimation for observability services (in preview soon).

Additionally, to enable customers and partners to build their own agents, we are releasing Network observability MCP tools and agent skills. This will allow their agents to leverage connectivity tests, and allows for natural language querying of VPC Flow Logs (both in preview).

The network that scales with you

We built our Cross-Cloud Network on the same global infrastructure that powers Google’s largest AI and internet services. This provides you with a blazing-fast, planet-scale foundation that is both secure by design and open by principle, allowing you to integrate your trusted partners across any environment.

As we move into the agentic era, our flexible, future-proof solutions ensure you can quickly adopt the latest AI technologies while maintaining the reliability of your core applications.

Whatever comes next, we’ve built the network to help you lead it. Attend our networking sessions at Next ’26 to learn more, or learn more about the Cross-Cloud Network!

Converging operational and analytical data for AI transformation

Wed, 22 Apr 2026 12:00:00 +0000

To act at the speed of business, AI agents must operate in fast and trusted reasoning loops. They need to “think” by reasoning across both your historical context and your live operational reality. Only by understanding this complete, real-time picture can they “do” — taking immediate action.

For decades, data architectures have been built with a structural wall that breaks this loop, separating the platforms that generate insights from the platforms that manage actions. This latency means insights are gleaned after the critical window for an agent to take action has closed. Achieving true AI transformation requires organizations to move from a passive system of record to a proactive System of Action, built on a closed-loop architecture that converges operational and analytical data.

At Google Cloud Next, we announced new unifying capabilities that drive our Agentic Data Cloud, eliminating silos and enabling 98% of our largest data cloud customers to run operational and analytical workloads in a unified data platform. By operating AlloyDB, BigQuery and Spanner together, we are delivering an AI-native architecture that unlocks the full potential of your data for real-time, agentic applications.

Flexible, real-time data agents

To act effectively, agents require both operational and historical signals for sound decision-making. Our Agentic Data Cloud bridges the gap between the operational "now" and analytical history by handling the complex plumbing for you. We provide diverse integration models across data federation, reverse ETL, and real-time ingestion to the lakehouse, empowering your agents to make high-stakes decisions with both live context and historical depth.

For example, sometimes an agent driving a live operational application needs to pull historical context on demand. Through Lakehouse federation for AlloyDB (Preview), agents can access Lakehouse data directly from AlloyDB itself. This allows frontline systems to instantly query extensive historical data without relying on brittle data movement pipelines.

In other scenarios, the challenge is reversed: deeply complex historical insights have already been calculated in the data warehouse, but an agent needs to deliver them to millions of users at conversational speeds. Reverse ETL for BigQuery (Preview) provides a one-click solution to push these heavy analytical insights back into AlloyDB, Bigtable, or Spanner, enabling agents to serve them with sub-millisecond latency.

One-click reverse ETL

Teams running real-time analytics on live operational data typically have to move that data into analytical systems — an error-prone process that introduces lag. With Spanner Columnar Engine (GA) users can perform analytical queries that run up to 200 times faster with zero impact on production transactional workloads.

Finally, the reasoning loop is not complete until an agent’s real-time action is captured for downstream analysis. To close this loop, Datastream for Lakehouse Apache Iceberg tables provides real-time Change Data Capture (CDC) from AlloyDB, Cloud SQL, Spanner, and Oracle directly into the open Lakehouse. This process streams every operational change as an append-only event into Lakehouse tables, making that data immediately available in BigQuery for ML model training, feature engineering, and real-time analytics.

"AlloyDB, along with other Google Cloud products like BigQuery, provides the agility and performance needed to continually enhance our platform's capabilities and help us anticipate emerging trends rather than merely reacting.” - Javi Fernández, CTO, Loyal Guru

Grounding agents in a unified governance foundation

Inconsistent definitions and unclear data ownership across operational and analytical systems can cause agents to hallucinate. To address this, we are extending Knowledge Catalog (Preview), formerly Dataplex, with new integrations for AlloyDB, BigQuery, Bigtable, Cloud SQL, and Spanner to provide a unified map of your data landscape. Integrations with Oracle AI Database@Google Cloud and Firestore are coming soon. The Knowledge Catalog works by aggregating native context across your Google and partner data platforms, semantic models, and third-party catalogs, unifying them into a single, governed source of truth needed to build and scale reliable agents.

"Seven-Eleven Japan created “Seven Central,” a scalable data platform that uses Spanner and BigQuery to provide real-time insights and support the company’s digital innovation strategies. We collect data from all 21,000+ stores, and in anticipation of a future expansion in business operations, we have designed a system that can scale up and run without issue, even if we were to have 30,000 stores, with 1,000 customers per store per day."
-Izuru Nishimura, Executive Officer and Head of ICT Department, Seven-Eleven Japan

Unified engines for deep reasoning

To move beyond simple Q&A chatbots to autonomous agents, AI must reason across every dimension of your data estate. Historically, combining keyword search, semantic understanding, and relationship mapping meant moving data out of operational databases and into specialized, siloed search engines — introducing latency and complexity.

Google’s Agentic Data Cloud eliminates these silos. By embedding native vector and full-text search directly into operational databases like AlloyDB, Bigtable, Cloud SQL, Firestore, and Spanner, agents can execute highly accurate hybrid searches combining keyword relevance and semantic intent.

We’re also bringing together graph and vector support across BigQuery and Spanner. With graph federation, an agent can match live user intent in Spanner and immediately trace that intent through historical graph relationships in BigQuery Graph — accelerating autonomous decision-making without moving the data. This multi-model approach powers advanced GraphRAG patterns, equipping agents with the rich, interconnected context required to accelerate autonomous decision-making.

“To deliver AI that actually works across HR, payroll, and workforce operations, you need a consistent, real-time data layer. With the power of Google’s Agentic Data Cloud, People Fabric is the backbone of UKG’s Workforce Operating Platform — turning fragmented systems into a single source of truth that powers intelligent, agent-driven experiences.”
-Radhi Chagarlamudi, Group Vice President, Product Engineering, UKG

Built for performance at agent scale

Our Agentic Data Cloud delivers the closed-loop architecture required for the AI era without compromising operational performance. Built on open standards like Iceberg and PostgreSQL, and governed by universal semantics, Google Cloud provides the speed, throughput, and trusted context needed to build the next generation of conversational and autonomous applications.

Build: Explore the AlloyDB AI documentation to start grounding your agents.
Connect: Visit the BigQuery Console to set up your first federated query to Spanner.
Govern: Opt-in to the Knowledge Catalog for unified visibility.

Gemini Cloud Assist: Proactive cloud operations that work for you, even before you ask

Wed, 22 Apr 2026 12:00:00 +0000

Today at Google Cloud Next, we are unveiling a more proactive Gemini Cloud Assist, our AI-assisted cloud operations platform. This update shifts your Google Cloud operations from manual workflows to a proactive, intelligent experience supported by a powerful ecosystem of agents.

Why it matters: A new agentic architecture enables Gemini Cloud Assist to handle the heavy lifting of your cloud management. By embedding intelligence, your enterprise context, and the power of Gemini directly into the operational layer, Gemini Cloud Assist proactively executes complex tasks such as designing applications, troubleshooting issues, and preemptively optimizing costs, that previously required constant human oversight. In enterprise-scale systems, this approach accelerates development velocity and reduces resolution times.

What’s new:

Using natural language and the power of Gemini, reduce the time from design to deployment of new or existing multi-resource deployments via a redesigned Application Design Center.
Automate infrastructure operations via gcloud, kubectl, and Terraform while using proactive multi-turn agents to troubleshoot and resolve incidents.
Identify your cost anomalies 24/7 using a proactive FinOps agent that analyzes spending spikes and generates granular cost reports on demand.
Assistance wherever you are. Powered by Google Cloud MCP servers and our proactive agents under the hood, Gemini Cloud Assist also exposes its own design, operation, troubleshooting and optimization capabilities as published MCP servers, bringing them straight to your IDE.

“Gemini Cloud Assist has significantly helped our dev teams. It reduced the number of outreach and touch points I have with them regarding Google Cloud questions by 60%. This allows our cloud team to scale more effectively and focus on more complex tasks.” - Oscar Aldana Assad, Senior Cloud Engineer, Petco

Let’s take a deeper look at how the agentic Gemini Cloud Assist can help your operations.

Accelerate production-readiness with Application Design Center

Gemini Cloud Assist serves as the intelligent reasoning engine for Application Design Center, acting as the bridge between natural-language intent, and a visual, production-ready architecture. By describing your infrastructure goals in plain language, Gemini Cloud Assist leverages Application Design Center to automatically lay out a visual design, including deployable Terraform. These templates are based on best-practice architecture guidance from Google Cloud and help bring security, reliability and compliance by design. Integrated with Security Command Center, quickly go from idea to deployment that conforms to your organizational policies.

Platform teams can then curate shared catalogs of pre-approved templates and integrate their own custom Terraform modules directly into the design process, providing a governed framework. This established, well-lit path helps developers adhere to organizational security and compliance guardrails from the first day of deployment. Beyond initial deployment, Gemini supports the full application lifecycle with interactive, multi-turn problem solving to update cloud resources.

Move from reactive to proactive remediation

In production, Gemini Cloud Assist helps you shift operations from reactive troubleshooting to quickly analyzing hypotheses to drive a lower time to resolution. Triggered by alerts, Gemini Cloud Assist proactively clusters and analyzes signals to initiate investigations before issues escalate. Now with Gemini 3, Gemini Cloud Assist correlates logs and metrics and identifies root causes from infrastructure signals down to the application code. Gemini Cloud Assist explores parallel hypotheses via tool calls and presents a technical breakdown of observations in a centralized UI. If intervention is required to address an underlying Google Cloud issue, users can hand off complete context to Google support, minimizing the iterations required for sharing configuration and context data.

Identify cost anomalies 24/7

To maintain economic health, Gemini Cloud Assist now acts as an proactive optimization agent for your projects. Running in the background 24/7, it monitors for cost anomalies and provides root-cause analysis, correlating spending spikes with specific engineering triggers like new resource creation, auto-scaling events, or pricing changes. You can query resource utilization via natural language to generate on-demand, tabulated reports, by project and applications registered in AppHub, providing granular visibility into "who, what, when, and how" — without manual data aggregation. For example, you can ask ‘Why did the cost of my application increase yesterday?’ or ‘How much did my project cost last month?’ and Gemini Cloud Assist answers by correlating cost data with infrastructure change, audit, and monitoring logs to get you an accurate answer.

Assistance everywhere

We are meeting teams where they work by expanding the surfaces where Gemini Cloud Assist is available. A Gemini Cloud Assist agent is already accessible through the console and mobile interfaces. And with new support for the Model Context Protocol (MCP), Gemini Cloud Assist is now available in Gemini CLI, your favorite agentic IDE or CLI, and third-party toolchains like ServiceNow and Slack. Integrating proactive assistance within existing workflows helps teams to avoid context switching and maintain flow-state.

Proactive capabilities at your fingertips

We designed Gemini Cloud Assist to help manage the end-to-end lifecycle of your applications, providing a multi-agent approach from deploying new applications to managing existing applications in the cloud. With the help of Gemini 3, Gemini Cloud Assist can now:

Increase your development velocity: Accelerate production-readiness using intent-driven architectures that unify best practices, security policies, and enterprise compliance.
Streamline production operations: Triage, diagnose and resolve production issues faster, through Gemini-based troubleshooting, recommendations and remediations.
Automate cost optimization: Automatically detect, analyze, root-cause, and alert you about cost anomalies for your projects on a daily basis.
Meet your teams where they are: Through proactive agents and MCP tools, engage with functionality using surfaces that range from the Google Cloud console to your CLI and IDE, so teams can stay in a flow-state.

The future of operations is agentic. You can begin your journey with our proactive cloud by enabling Gemini Cloud Assist in your project settings today.