Infrastructure Modernization

Building the agentic future: A spotlight on Google Cloud’s media & entertainment partner ecosystem

Thu, 16 Apr 2026 13:15:00 +0000

As we gather in Las Vegas for NAB Show 2026, the industry conversation has shifted. We are no longer asking if AI works; we’re now focused on how it scales. The era of AI experimentation is over — production-grade execution is here.

At Google Cloud, we believe no studio or broadcaster should have to build this future in isolation. Our mission is to provide the agentic platform and AI and cloud tools that allow our partners to innovate at the speed of ideas — from the tools used in the edit suite, to the technology that delivers video to millions of viewers worldwide.

Enhancing production: From manual tasks to intelligent assistants

Modern creative workflows are often slowed down by manual technical tasks. Google Cloud is working with ecosystem leaders to integrate advanced AI capabilities directly into the core of production software, so creators can focus on their artistry, not tedious tasks. With AI acting as a proactive assistant within the creative suite, production teams can significantly reduce the time between a raw idea and a finished frame.

Avid: With the launch of Content Core on Google Cloud, Avid is delivering a truly cloud-native studio. And by integrating multimodal AI search into Media Composer, editors can find the exact frame they need using natural language, turning hours of logging into seconds of discovery.
Backlight: Backlight makes complex media workflows simple for teams of all sizes, from production through monetization. Built on Google Cloud with the Video Intelligence API, Backlight's Iconik platform automatically adds searchable metadata upon upload. Customers see up to 50% shorter production cycles and save up to 60% on storage by deeply understanding their media libraries, reducing duplications, and optimizing asset placement.
Brahma.ai: Brahma AI, an enterprise AI content platform, is powering high-fidelity digital likenesses across retail, entertainment, and healthcare, making them interactive and intelligence-driven within a secure and governed framework.

Unlocking content value: From static archives to active assets

Data is only as valuable as the insights you can extract from it. Our partners, listed on the Google Cloud Marketplace, are using generative media models to transform massive, static archives into searchable, revenue-generating engines. By making every frame discoverable, we’re helping media companies turn decades of history into immediate opportunities.

Ateme: Ateme helps simplify global distribution with its new generative AI-powered subtitling solution, which can significantly reduce the manual labor of localizing different media types.
Perfect Memory: Perfect Memory helps customers turn traditional storage into a context-aware knowledge engine. The platform understands the relationships between athletes, historical events, and emotional nuances — transforming massive media archives into an intelligent library that lets creative teams instantly surface the perfect content for any story.
VionLabs: Working with companies like Cineverse, Plex, and Crunchyroll, Vionlabs uses AI to analyze and index content libraries — making video assets more accessible and enabling metadata generation. By understanding the specific mood and aesthetic of each scene, Vionlabs helps streaming platforms move beyond basic genre tags toward more nuanced, sentiment-driven content discovery and marketing.

Scaling global reach: From simple streams to audience growth

To succeed today, media companies must provide a smooth viewing experience and easy payment options. Our ecosystem provides the tools to grow a company’s reach and maximize the value of every subscriber through reliable, high-quality delivery.

Bending Spoons: By leveraging the global scale of Google Cloud, Bending Spoons’ properties such as Brightcove and Vimeo are delivering professional-grade tools for large enterprises, SMBs, the next generation of creators, and more. Its platforms ensure that high-quality video production and distribution are accessible to everyone, from global brands to independent storytellers.
Bitmovin: Bitmovin enables streaming services to scale efficiently while delivering a premium experience across the widest range of devices. By combining real-time observability with AI-driven insights, media teams can proactively optimize engagement and monetization. Furthermore, Bitmovin’s advanced encoding ensures superior visual quality at lower bitrates, supporting everything from high-demand Video on Demand (VOD) to massive, 24/7 live events.
Evergent: Evergent automates complex billing and monetization workflows for AI-powered revenue management. Media and telecommunications companies can use Evergent’s tools to maximize subscription growth and improve long-term customer retention through personalized and agile payment offers.
Harmonic: Harmonic is helping major broadcasters like Grupo Globo modernize their operations. By integrating new digital broadcast capabilities into their cloud-based streaming solutions, Harmonic provides leaders with a faster, more efficient path to manage video processing and delivery at a massive scale.

Ensuring reliability: From infrastructure to a foundation of trust

High-quality content requires a high-performance foundation. We are partnering with infrastructure leaders to ensure that even the most complex global broadcasts remain stable, secure, and responsive.

Zixi provides the broadcast-grade transport and workflow automation needed to move professional video across any network. By offering centralized control and complete visibility into the delivery process, Zixi ensures that leaders like Fubo can manage high-stakes, broadcast-quality live events without the risk of a signal drop.

Visit the ecosystem in action

The strength of our ecosystem is its integration across all aspects of the media and entertainment landscape. From the cameras, to the cloud, to the viewers' screens, these partners represent the future of a more creative, efficient, and agentic media industry.

Visit the Google Cloud Booth (West Hall, #W2731) at NAB Show from April 19-22 to see many of these partners in action through live demonstrations and theater sessions.

Cool stuff Google Cloud customers built, April edition: BMW big on SLMs, MLB’s Scout Insights AI, personalized resort experiences

Wed, 15 Apr 2026 16:00:00 +0000

AI and cloud technology are reshaping every corner of every industry around the world. Without our customers, who are building the future on our platform, there would be no Google Cloud. In this regular round-up, we dive into some of the exciting projects redefining businesses, shaping industries, and creating new categories.

For our latest edition, we learn why BMW Group is experimenting with small language models (SLMs); catch AI-powered commentary from Major League Baseball; hit the slopes with Vail Resort’s AI concierge; build an intelligent grid with CTC Global; witness how ID.me created secure global scale; and see how Manhattan Associates supply chain tools now handle 1 billion daily API calls.

Be sure to check back next month to see how more industry leaders and exciting startups are putting Google Cloud technologies to use. And if you haven’t already, please peruse our list of 1,001 real-world gen AI use cases from our customers.

BMW tests the big potential of small models

Who: As one of the world’s leading providers of premium cars and motorcycles, BMW Group is always at the forefront of automotive technology. This ethos pushed the company to test what type of AI language models are ideally suited to driving situations, where access to cloud-based LLMs isn’t always possible.

What they did: BMW Group wanted to explore the potential of small language models (SLMs), which could run within the limited hardware on a vehicle. Finding the right trade-off between size and capability requires careful optimization, and the sheer volume of viable combinations renders manual searches for the optimal configuration an incredibly tedious, if not impossible, undertaking. To overcome this challenge, BMW and Google built automated, reproducible workflows through executable pipelines using Vertex AI.

Why it matters: The path from a general-purpose LLM to a specialized SLM isn’t straightforward. Every choice — from type of quantization to characteristics and contents of the fine-tuning domain-specific dataset — affects the quality and efficiency of the final model. This creates an exponential range of configurations, each with different trade-offs. It’s a great example of using AI to scale an optimization problem for other AI.

Learn from us: “With automated pipelines, we can rapidly adapt models to our domain and rigorously test and evaluate them against domain-specific benchmarks. This allows us to iterate and optimize models in hours rather than days.” – Dr. Céline Laurent-Winter, vice president, Connected Vehicle Platforms at BMW Group

MLB Scout Insights: AI-powered color commentary

Who: Major League Baseball is famous for its colorful announcers. Now, MLB is bringing more baseball color straight to your pocket, and Gemini is helping give it a voice.

What they did: Each season, millions of baseball fans use the MLB app and tap over to the Gameday feature for live, up-to-the-pitch action across more than a dozen games. Starting this season, the league launched MLB Scout Insights in Gameday, which uses Gemini models to quickly scan decades of game and player data, cross-references it with situational game scenarios, and then delivers game-relevant context during key matchups.

Why it matters: Given the sport’s storied history, 162-game regular season, and global reach, baseball fans are among the most sophisticated and passionate out there. To keep them engaged with Gameday and the MLB app, the league wanted to deliver insights that truly felt meaningful and interesting. Building the tool meant answering a rather squishy question: What makes an insight actually insightful, not just an accurate fact, and how can an AI learn that distinction? The answer came from some clever “surprisal” analysis.

Learn from us: "With Scout Insights, every fan can feel like the smartest person in the stands, at the water cooler, or on the couch. It’s about deepening connections to the game, and sharing that passion with others. That’s the magic of sports, and we’re making more of it possible with the magic of AI." – Josh Frost, senior vice president of product & Matt Graser, director of engineering, Major League Baseball

Vail Resorts makes personalized AI assistance easy

Who: Vail Resorts operates some of the most iconic and beloved mountain destinations in the world, including Whistler Blackcomb, Park City Mountain, Stowe, and Crested Butte.

What they did: Vail Resorts launched My Epic Assistant during the 2024-2025 snow season, and expanded it this year to add even more AI-powered chat features powered by Google’s powerful Gemini models. The result is an agentic guide to the slopes that can help skiers and snowboarders decide on the right season pass, share the latest snow report, check on lesson preparations, or suggest a good stop for cocoa.

Why it matters: Vail Resorts wanted more than a chatbot; they wanted a digital concierge that understands the nuance between a powder day at Whistler and a family trip to Beaver Creek. As the company implemented and refined personalization, improved search, summary capabilities, and conversational flow within My Epic Assistant, the app has delivered a 45% reduction in escalation to human agents since launch.

Learn from us: "Utilizing tooling from Google Cloud, we could lean into agentic design patterns that gave us a way to unlock natural, personalized conversations. These boosted customer satisfaction, while reducing the need for manual intent design. These tools also let us combine flexibility and control to enable the assistant to respond fluidly but always within the boundaries of our brand, policies, and product strategy.”— The Vail Resorts technical team

CTC Global turns the smart grid into an intelligent one

Who: CTC Global is a leading manufacturer of advanced transmission conductors and power lines. While many nodes in the grid contain IoT sensors, it recognized a literal gap in the transmission lines themselves.

What they did: CTC’s new GridVista platform threads fiber-optic cable into its high-strength carbon fiber composite core, and connects these to monitoring technology built with AI and monitoring technology from Google Cloud and Tapestry. With GridVista, CTC can turn every inch of transmission into a smart sensor.

Why it matters: GridVista gives CTC grid operators an accurate and reliable view of what’s happening across the entire line — based on actual, real-time data from the entire length of the conductor, not point estimates from a static model of line conditions or the occasional clamped-on sensor. This means they can significantly improve safety, manage costs, increase the line’s capacity to transmit power, and enhance reliability with more precise insights about events that might trigger an outage.

Learn from us: “This awareness allows for a grid that can truly sense its own health in real time and provide unprecedented awareness of conditions on the entire line. Whether that’s real time storm impacts, ice load, wind load, branches on the wire, or temperatures on or under the line. The GridVista system truly represents next generation capabilities. ” — J.D. Sitton, CEO, CTC Global

ID.me reduces risk while scaling past 160 million users

Who: ID.me is transforming digital identity security for the modern era, offering a single login that lets you easily prove you’re you across a wide range of platforms and wallets.

What they did: ID.me currently serves more than 160 million users, including as many as 40,000 at any time, so they can prove their identity online as easily as flashing their driver’s license in person. Over the last two years, ID.me migrated more than 50 terabytes of data across 15 database instances to Google Cloud with minimal downtime. They also introduced a two-tier architecture with Cloud SQL supporting its smaller and more standard services, while AlloyDB runs heavier workflows that form the backbone of the ID.me platform.

Why it matters: AlloyDB AI has allowed ID.me to scale its systems to handle 10X-20X of what was possible before — and at a lower price to boot. That responsiveness and reliability led the U.S. federal government to recognize ID.me for its role in preventing large-scale fraud within national systems.

Learn from us: "We’ve been able to scale both our infrastructure and trust. With a platform that’s faster, smarter, and built to handle portable identity at massive scale, we’re one step closer to our goal: a secure, digital way to prove who you are, wherever you need it, that works everywhere you need it." — Kevin Liu, Cloud Platform Architect, ID.me

Manhattan Associates powers more than a billion daily API calls

Who: Manhattan Associates is a global leader in supply chain and omnichannel commerce solutions, offering tools and platforms that reach more than 2 billion people across 20 billion consumer touchpoints.

What they did: Manhattan Associates modernized its Manhattan Active SaaS platform by migrating from legacy Oracle and DB2 systems to Google Cloud databases. Each capability of Manhattan Active now runs as an independent, containerized service orchestrated by Google Kubernetes Engine (GKE). Data flows through Pub/Sub into BigQuery for real-time analytics, while Cloud Logging and Cloud Monitoring deliver observability at scale.

Why it matters: With its new microservices-first design, Manhattan gained the agility to evolve faster and the confidence that mission-critical operations would remain resilient across regions. With Cloud SQL and BigQuery, the company now processes more than a billion daily API calls with average response times of less than 150 milliseconds. This evolution supports hundreds of thousands of monthly active users across tens of thousands of stores and distribution centers. The new platform also created the foundation for Manhattan’s Agentic AI suite, which includes prebuilt agents — like the Intelligent Store Manager and Labor Optimizer — that coordinate real-time decisions across store and distribution center operations.

Learn from us: "Operationally, the platform has become more elastic and efficient. The system automatically handles hundreds of thousands of scaling events per day, ensuring performance remains consistent during peak surges without expensive overprovisioning." — Narayana Reddy Kothapu, Senior Director, Manhattan Associates & Rajkumar Ramani, Technical Director, Manhattan Associates

Cool stuff Google Cloud customers built, Feb. edition: Telco data reinvention; Golden State’s “G.O.A.T.T.”; John Lewis explores DORA

Thu, 26 Feb 2026 17:00:00 +0000

AI and cloud technology are reshaping every corner of every industry around the world. Without our customers, there would be no Google Cloud, as they are the ones building the future on our platform. In this regular round-up, we dive into some of the exciting projects redefining businesses, shaping industries, and creating new categories.

For our latest edition, we explore a new data approach for Vodafone and Fastweb; evaluating John Lewis Partnership’s developer platforms; the Golden State Warrior’s AI playbook; healthy, stable networks at Hackensack Meridian Health; and Ab Initio brings better context to data for AI.

Be sure to check back next year to see how more industry leaders and exciting startups are putting Google Cloud technologies to use. And if you haven’t already, please peruse our list of 1,001 real-world gen AI use cases from our customers.

Fastweb + Vodafone reimagined data workflows

Who: Following the acquisition of Vodafone Italy by Swisscom in 2025, these leading European telecom providers wanted to rethink how they serve customers and deliver timely, personalized experiences across mobile, broadband, and digital channels.

What they did: Both companies had already begun modernizing customer data workflows with BigQuery, but combining ecosystems exposed certain limits of the existing setup. In order to give every channel real-time access to accurate customer data, they implemented Spanner as a service and governance layer, delivering low-latency reads, horizontal scalability, high availability, and a fully managed environment with zero ops overhead. The team is also using Gemini to generate clear documentation directly from the code, which saves hours of manual work.

Why it matters: Using Spanner Graph allowed the organization to map lineage in a way that reflects how its platform actually works: which tables drive specific jobs, how transformations cascade, and where dependencies sit. Call centers now see more complete, up-to-date customer information, digital channels can rely on consistent data without custom integrations, and partners can access what they need with low latency through Apigee.

Learn from us: “Rebuilding our Customer 360 platform with Google Cloud services has already changed how Fastweb + Vodafone works. Workflow monitoring is simpler, pipelines are leaner, and real-time serving is now the norm. ” – Vincenzo Forciniti, IT AI Adoption & Platform Engineering Lead, Fastweb + Vodafone

John Lewis measures the value of its developer platform

Who: The John Lewis Partnership is a major UK retailer operating John Lewis department stores and Waitrose supermarkets. To power their digital transformation, they built the John Lewis Digital Platform (JLDP) to support dozens of product teams building high-quality software for johnlewis.com.

What they did: Moving beyond simple usage metrics, John Lewis developed a sophisticated, multi-stage approach to measuring the real value of their platform. They transitioned from initial speed-based metrics (like "Onboarding Lead Time") to a comprehensive model using DORA metrics and subjective engineer feedback via the DX platform. This included a custom "Technical Health" feature that uses small, automated jobs to monitor more than 35 health measures — such as Kubernetes best practices, security, and operational readiness — providing teams with real-time "traffic light" indicators of their service health.

Why it matters: By focusing on value rather than just activity, John Lewis ensured the platform was actually reducing friction for developers rather than just being a mandatory tool. Their automated Technical Health checks allow product teams to manage technical debt and security vulnerabilities proactively. This approach has decoupled centralized operations teams from individual services, leading to faster incident resolution (MTTR), fewer outages, and significant cost savings.

Learn from us: "Measurement is a journey, not a destination. Start by measuring something meaningful to your stakeholders, but be prepared to adapt as your platform evolves. The things that mattered when you were proving out the platform's viability are unlikely to be what are important several years later when your features are mature." – Alex Moss, Principal Platform Engineer, John Lewis Partnership

Hackensack Meridian Health de-risks network migration using VPC Flow Logs

Who: Hackensack Meridian Health is a leading not-for-profit healthcare organization and the largest hospital system in New Jersey. System reliability is a cornerstone value for HMH as they manage a vast network of hospitals, urgent care centers, and physician practices.

What they did: Preparing for a large-scale migration to a new Google Cloud network design, Hackensack Meridian Health used VPC Flow Logs and Flow Analyzer to eliminate the "black box" of hybrid traffic. By enabling logs on their Cloud Interconnect VLAN attachments, they captured granular telemetry — including source/destination IPs, ports, and protocols. They then exported this data to create a visual "who-is-talking-to-what" map. This allowed them to identify critical traffic patterns between on-premises data centers and specific Google Cloud regions, VPCs, and applications.

Why it matters: In a healthcare environment, even minor network disruptions can have major consequences. By mapping traffic proactively, Hudson Meridian Health pinpointed exactly which moments in the cutover carried the highest risk. This preparation allowed them to detect a migration issue in just three minutes and resolve it within five — a process that previously could have taken hours. Beyond migration, this level of visibility enables the organization to better manage capacity planning, cost attribution, and security compliance across their hybrid infrastructure.

Learn from us: "Getting a clear picture of our interconnect traffic always felt like a black box. Enabling VPC Flow Logs and feeding it into Flow Analyzer finally gave us the map we needed. Identifying those critical traffic flows before we changed any routes was key to de-risking the entire migration." — Randall Brokaw, Cloud Engineering Manager, Hackensack Meridian Health

The Golden State Warriors’ AI-powered back office

Who: The Golden State Warriors are one of the NBA’s most successful modern franchises. Behind their on-court wins are a specialized operations team who run what might be called organization’s "G.O.A.T.T." (Greatest of All-Time Technologies), a data and AI platform that helps drive game-time insights, trading decisions, and fan experience enhancements.

What they did: The Warriors transitioned from a "gut-feeling" culture to an "analytics-first" strategy by building an internal "digital brain" on Google Cloud. Using BigQuery and Gemini, the team now automates complex workflows that previously took hours, such as generating pre-game scouting reports. They use machine learning to run thousands of trade simulations that prioritize "team fit" over raw individual stats and employ computer vision to track the "shot quality" of every attempt in the NBA. On the business side, they built a content recommendation engine using the Discovery API to deliver personalized digital experiences to their global fan base.

Why it matters: This AI-driven approach narrows the decision tree for leadership, allowing them to focus human expertise on the most viable options. By automating the “science” of data processing, coaches and scouts have more time for the "art" of face-to-face training, planning, and player development. This integration has not only influenced on-court strategy — like the three-point revolution — but has also improved business efficiency, with employees now proactively bringing AI-driven ideas to the IT team rather than waiting for top-down mandates.

Learn from us: "You can never reach a point where either humans or machines are making all the decisions. The sweet spot is finding that middle ground where intuition and data converge on the same conclusion. Data helps us narrow our decision tree before we even start evaluating specific options." — Nick Manning, Senior Director of Consumer Products & Emerging Technology, Golden State Warriors

Ab Initio unlocks enterprise data for the agentic AI era

Who: Ab Initio is an enterprise software company specializing in high-volume data integration and governance. Their platform is trusted by large-scale organizations to manage complex data lifecycles across hybrid and multi-cloud environments.

What they did: To solve the challenge of grounding AI agents in accurate data, Ab Initio partnered with Google Cloud to integrate its data fabric with BigQuery, Dataplex Universal Catalog, and Gemini. They launched a suite of more than 500 metadata and data connectors that bridge the gap between legacy systems (like mainframes, COBOL, and SAS) and modern cloud environments. This integration provides field-level, end-to-end lineage, allowing Gemini to access well-documented, "AI-ready" data regardless of where it resides.

Why it matters: AI agents are only as effective as the data they can access. By using Ab Initio as a "neutral hub," enterprises can federate data from on-premises and multi-cloud sources into a single unified layer without moving the data itself. This provides the rich semantic context and lineage needed for Gemini to perform grounded, explainable reasoning. For businesses, this means faster transition from experimental AI to production-ready agentic workflows that are auditable, compliant, and capable of making complex, automated decisions.

Learn from us: "Agentic AI requires trusted, AI-ready data and metadata. Understanding the origin, quality, and meaning of information matters as much as the data itself. Gemini serves as a key component of the agentic layer, using this context to make decisions that are explainable and auditable." — Scott Studer, Head of Development, Ab Initio & Chai Pydimukkala, Data Governance, Sharing & Integration Product Lead, Google Cloud

Accelerate migrations with new incentives from the Rapid Migration and Modernization Program (RaMP)

Wed, 21 Jan 2026 16:00:00 +0000

To lead in 2026, you need to be AI-ready, lean, and optimized across every workload — and for many organizations, that means migrating and modernizing applications like SAP, Oracle, NetApp, and VMware to the cloud. At Google Cloud, we’ve helped thousands of customers with successful migrations through the Rapid Migration and Modernization Program (RaMP), and today, we’re introducing the new RaMP, so that the more you migrate, the more you save. Highlights of the new program include:

Google Cloud Service Credits: Earn credits based on your incremental usage of eligible workloads on Google Cloud.
Partner and Google Cloud Professional Services funds: We will fund partners and Google Cloud Professional Services to help you assess your needs, build your business case and implement your migration and modernization roadmap.
Earn more with advanced workloads: Earn additional credits and incentives for advanced workloads (like SAP, Oracle, VMware, Data Analytics, etc.) to offset higher technical costs.

From technical debt to AI readiness

You deserve a migration and modernization path that leads to less cost and complexity, not more. Moving to Google Cloud provides the infrastructure foundation you need to replace your technical debt with flexibility, so you can reduce cost and complexity.

But migrating and modernizing your infrastructure is about more than that — it’s about data accessibility, optimization, and innovation. For example, when you migrate a legacy SAP environment or a massive Oracle database to Google Cloud, you aren't just changing where the data sits; you are making that data accessible to Vertex AI and our Gemini models.

RaMP can accelerate your migration and replace technical debt with a scalable, secure foundation that can support existing enterprise workloads and the next generation of AI applications — whatever they may be.

Building for the future

Migrating and modernizing with RaMP gives you immediate access to world-class infrastructure, data, and AI solutions, giving you the foundation you need to succeed in 2026 and beyond. To get started, visit our RaMP page to learn more and start your assessment. Get ready to rapidly enter the AI era. Welcome to the fast lane.

Cool stuff Google Cloud customers built, Dec. edition: AI for better toys, reliable mapping tech, Gemini stumps an all-star & more

Wed, 31 Dec 2025 16:00:00 +0000

For our latest edition, we look into how Waze made its network more reliable; NBA superstar Stephen Curry gets quizzed by Gemini; a financial market transformation at CME Group; a multi-agent business forecasting platform from AppOrchid; Mattel crunches customer feedback with AI; VMO2 uses decentralized contracts for reliable data; Mercado Libre’s strategic use of Spanner; and how Ericsson enhances data governance.

Waze keeps traffic flowing with Memorystore

Who: Waze (a division of Google parent company Alphabet) is a community-driven, crowd-sourced navigation app with tens of millions of users who share real-time data to provide optimal driving routes, traffic updates, and alerts for hazards, police, and more.

What they did: Waze depends on vast volumes of dynamic, real-time user session data to power its core navigation features, but scaling that data to support concurrent users worldwide required a new approach. Their team built a centralized Session Server backed by Memorystore for Redis Cluster, a fully managed service with 99.99% availability that supports partial updates and easily scales to Waze’s use case of over 1 million MGET commands per second with ~1ms latency.

Why it matters: Moving from Memcached’s 99.9% SLA to Memorystore for Redis Cluster’s 99.99% means higher availability and resiliency from the service. And because Memorystore for Redis supports partial updates, Waze can change individual fields within a session object rather than rewriting the entire record. That reduces network traffic, speeds up write performance, and makes the system more efficient overall.

Learn from us: “Real-time data drives the Waze app experience. Our turn-by-turn guidance, accident rerouting, and driver alerts depend on up-to-the-millisecond accuracy. But keeping that experience seamless for millions of concurrent sessions requires robust and battle hardened infrastructure that is built to manage a massive stream of user session data.” – Eden Levin, Waze BE infrastructure developer & Yuval Kamran Waze site reliability engineer

What Stephen Curry learned from a custom Gemini agent

Who: Stephen Curry is arguably of the greatest three-point shooter of all-time in the NBA — as well as Google’s performance advisor and an all-around stats-obsessive.

What they did: For a special engagement with Curry, the Google Cloud team wanted to showcase the power of Gemini for creative thinking, analysis, and data mining. They took every regular season, play-in, and playoff game from Curry’s career (through the end of the 2024-2025 season) and input the data into a custom-built agent using Google Cloud’s Agent Development Kit and Gemini APIs.The system could then be queried for obscure stats, to see if the team could stump Curry and teach him more about his game.

Why it matters: For example, it found that his three-point shooting percentage after more than seven dribbles, with a minimum 105 attempts was 40.2%, and how many points Curry generated for his teammates off of screens since 2013: 1,105. Instead of countless hours of manual research, the team got query results in less than a minute. Some queries were so obscure, the team wouldn’t have reached a valid answer without the ability of the agent to analyze the rich data.

Learn from us: “Gemini is going to be in my head this year, cause I'm going to be looking at all these details.” – Stephen Curry, Golden State Warriors point guard and 4x NBA champ

How CME Group builds a faster, smarter exchange

Who: CME Group has evolved from a nineteenth-century commodities exchange into one of the most advanced financial market infrastructures in the world. To support real-time trading and risk management at a global scale, the company launched a strategic partnership with Google Cloud.

What they did: By migrating to Cloud SQL and adopting AI-powered insights, CME Group empowered developers, paid down technical debt, and unlocked new opportunities for data-driven innovation across financial markets.

Why it matters: Cloud SQL has given CME a foundation for increased developer and team agility. Fewer performance issues mean more time focused on innovation: expanding CME’s analytics capabilities, accelerating AI initiatives, and exploring new ways to commercialize data responsibly. When teams stopped chasing outages, they unlocked more time to take bigger bets and build the future.

Learn from us: “With Cloud SQL, we’ve found a way to keep our data layer as fast and dependable as the markets we serve. Cloud SQL gives our teams real-time visibility into what’s happening inside the database. When an application slows, we can identify the root cause in minutes instead of hours. Those insights are built into the platform, which means we don’t need custom tooling or manual analysis to keep operations steady.” – Kristofer Shane Sikora, Executive Director, Cloud Data Engineering, CME Group

AppOrchid’s multi-agent system for superior business forecasting

Who: App Orchid is an enterprise AI builder and a leader in making data actionable with AI, with a mission to make AI a force for good. Their goal is to empower every employee with trusted, understandable, and accessible data.

What they did: The business forecasting agent is actually built on the foundation of two powerful, specialized AI agents: a prediction agent built by Google Cloud and App Orchid’s Data Agent offering. These agents work in concert to solve complex business problems, acting as complementary specialists. App Orchid’s agent possesses unparalleled understanding of an enterprise's past and present, while Google’s agent brings world-class capabilities in predicting the future.

Why it matters: Adopting a multi-agent approach provides clear, tangible advantages that directly address the forecasting problems that often plague businesses, including improved accuracy; increased operational efficiency; faster insights; and reduced costs and increased revenue; and greater agility and adaptability. Neither of the underlying agents could achieve these results on their own, but working together, this agent is more than the sum of its subagents.

Learn from us: “As the agentic era gets underway, it is evolving quickly. Our multi-agent approach demonstrates both how true agentic systems are most successful when multiple agents are at play, and the importance of finding strong partners with distinct capabilities to help build and assemble these agentic systems.” – Brian Mills, Director, Enterprise AI, Google Cloud & Taka Shinagawa, Gen AI Field Solution Architect, Google Cloud

How Mattel uses AI for real-time product updates

Who: Since its humble beginnings in a garage in 1945, Mattel has consistently been reshaping play for children and families across the globe with iconic franchises like Barbie, Hot Wheels, and Fisher-Price.

What they did: To improve its understanding of consumer sentiment, Mattel developed an AI-powered feedback classification system, which can analyze millions of customer interactions from a diverse range of sources in a matter of seconds. At its core, the system relies on BigQuery for storing and efficiently processing its massive customer datasets and then utilizes Vertex AI and Google’s multimodal Gemini models to refine and train the sophisticated consumer feedback model.

Why it matters: Already, the new AI-powered system has delivered significant wins, delivering a staggering 100x boost in data processing capacity and reducing analysis times from a month to a single minute. By automating the analysis of many processes, analysts are now freed from the noise of everyday tasks, enabling them to focus on deeper research across the company’s iconic portfolio brand.

Learn from us: “Our big motto is ‘From months to minutes,’ but it’s real. We were literally spending months-worth of analysis and just getting data into the place that an analyst could tally up all the sentiment — and now it’s just at our fingertips.” – Shaun Applegate, Director of Product Quality Analytics, Mattel

Virgin Media O2 uses data contracts for scalable AI products

Who: Virgin Media O2 is one of Europe’s largest telecommunications and media providers, with 45.8 million broadband, mobile, phone, and home subscribers across the UK. To build AI products that are adaptable and data-driven, they needed a decentralized system that internal customers could count on for clean, reliable data.

What they did: New decentralized data contracts, built with Dataplex, serve as the data quality and assurance layer for VMO2’s data products; these ensure every dataset they publish is reliable, documented, and ready for consumption. Defined at the asset level, such as individual BigQuery tables or Google Cloud Storage buckets, data contracts are redefining how VMO2 manage and share data, enabling the creation of trusted and scalable AI products across their data mesh.

Why it matters: The power of this approach lies in moving beyond static documentation. Because they are machine-readable, data contracts become living guarantees with continuous enforcement and real-time validation directly within data pipelines. This proactive monitoring allows teams to detect schema changes or SLA breaches early, transforming data quality from a reactive fix into a scalable, automated mechanism.

Learn from us: “By operationalizing trust through data contracts, we are fostering a culture of shared responsibility and data-first thinking. This federated model does more than simply fix pipelines; it builds the trusted foundation needed to scale next-generation AI. It ensures that the resilient AI tools empowering our teams are built on data that is reliable, consistent, and well-defined.” – Chandu Bhuman, Head of Data Strategy, Cloud & Engineering, Virgin Media O2 & Dženan Softić, Data & AI Architect, Google Cloud

Inside Mercado Libre's multi-faceted Spanner architecture

Who: Mercado Libre, an e-commerce and fintech pioneer across Latin America, operates at a staggering scale, demanding an infrastructure that's not just resilient and scalable, but also a catalyst for rapid innovation.

What they did: At the heart of Mercado Libre's strategy is Fury, an in-house middleware platform designed to abstract away the complexities of various backend technologies, providing developers with standardized, simplified interfaces to build applications. Spanner provides Fury with an always-on, globally consistent, multi-model database with virtually unlimited scale. By designating Spanner as a choice within Fury, Mercado Libre ensures that applications built on the platform using Spanner stay consistent globally, scale without breaking, and rarely go down.

Why it matters: The strategic adoption of Spanner, amplified by internal platforms like Fury and sophisticated data workflows, has yielded significant benefits, including: significant cost savings and low total cost of ownership; business impact and agility for developers; and low operational overhead thanks to automation.

Learn from us: “Mercado Libre's adoption of Spanner demonstrates how to use a powerful, globally consistent database not just for its core capabilities, but as a strategic enabler for developer productivity, operational efficiency, advanced analytics, and future AI ambitions.” – Pablo Leopoldo Arrojo, Software Technical Leader, Mercado Libre

Ericsson achieves data integrity and superior governance with Dataplex

Who: Ericsson is one of the world's leading providers of telecommunications and networking technology and solutions. Its Managed Services unit provides network operations and optimization, including field operations, for various telecom and enterprise customers, including outsourcing network performance management, future provisioning, network vulnerability management, and network energy infrastructure management.

What they did: To power the future of its autonomous network operations and deliver on its strategic priorities across a global network of more than 710,000 sites, Ericsson's Managed Services has been on a transformative data journey with governance at the center of its strategy. Ericsson moved from foundational practices to a sophisticated, business-enabling data governance framework using the Dataplex Universal Catalog — turning data from a simple resource into a strategic asset.

Why it matters: With Dataplex as the governance foundation, Ericsson began implementing the core pillars of its governance program, moving from manual processes to an automated, intelligent data fabric. More specifically, Ericsson established a unified business vocabulary within Dataplex, which helped eliminate ambiguity and ensure their teams — from data scientists to data analysts — were speaking the same language.

Learn from us: “Governance is a value enabler, not a blocker. A modern data governance program should focus on business enablement first, driving value and innovation in order to complement policies, rules and risk management. Also remember this work is a journey, not a destination. Be prepared to fail fast, learn, and adapt. The landscape is constantly changing at breakneck speed.” – William McCann Murphy, Head of Data Authority, Ericsson & Akanksha Bhagwanani, EMEA Data Analytics Solution Lead, Google Cloud

How CME Group builds a faster, smarter exchange on Cloud SQL

Wed, 03 Dec 2025 17:00:00 +0000

Editor’s note: The Chicago Mercantile Exchange (CME Group) has evolved from a nineteenth-century commodities exchange into one of the most advanced financial market infrastructures in the world. To support real-time trading and risk management at a global scale, the company launched a strategic partnership with Google Cloud. By migrating to Cloud SQL and adopting AI-powered insights, CME Group empowered developers, paid down technical debt, and unlocked new opportunities for data-driven innovation across financial markets.

From butter and eggs to bandwidth

CME Group is where risk meets opportunity. Every transaction that happens in our exchange — every order placed, trade executed, or risk calculated — relies on data moving flawlessly and instantly. The integrity of our markets depends on it.

Behind each of those trades is a database storing valuations, ownership, and so much more information, all of which can shift from millisecond to millisecond throughout the day. At our scale, those databases have to store and retrieve that information under relentless demand. We’re processing millions of messages a day with no margin for latency or error. That level of precision doesn’t come easily, especially in a highly regulated industry where performance has to coexist with security and reporting. Every change we make must align with strict compliance standards and global regulatory frameworks.

Speed has always been our currency, but scale became a challenge. CME Group's legacy database estate required significant engineering effort to maintain performance and meet regulatory demands. We needed to reduce operational overhead while improving our security posture. This required a managed database solution that offered transparent observability and clear compliance controls.

When Cloud SQL meets the trading floor

Our 10-year strategic partnership with Google Cloud aims to address this by migrating all our technology to the cloud, enabling us to innovate and collaborate on pushing the boundaries of what cloud infrastructure can support. Together, we’re experimenting with new ways to achieve ultra-low-latency performance in the cloud. As data volumes surge and AI becomes increasingly central to risk management, the ability to move and interpret information in milliseconds is a technical requirement. We’re building systems with Google Cloud that let us keep the market running, even as we lead it into the future.

With Cloud SQL, we’ve found a way to keep our data layer as fast and dependable as the markets we serve. Cloud SQL gives our teams real-time visibility into what’s happening inside the database. When an application slows, we can identify the root cause in minutes instead of hours. Those insights are built into the platform, which means we don’t need custom tooling or manual analysis to keep operations steady.

But for us, the value of Cloud SQL goes beyond performance tuning. It’s about confidence. Our database administrators can focus on strategic improvements, and our developers can validate and optimize queries without waiting for escalation. Taken together, we have faster troubleshooting and a data foundation ready for the always-on demands of global trading.

aside_block: <ListValue: [StructValue([('title', 'Build smarter with Google Cloud databases!'), ('body', <wagtail.rich_text.RichText object at 0x7f2e2707d8e0>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Cloud SQL is our new favorite teammate

The more we use Cloud SQL, the more it feels like we’ve added a new member to the team. AI-assisted insights from Cloud SQL have changed how the CME Group team works. When an application slows, Cloud SQL tells us why. It surfaces anomalies, walks us through guided analysis, and even suggests query optimizations that restore performance in minutes. Developers can see those recommendations right in their workflows, test fixes, then move on. No waiting, no hand-offs, no firefights.

In other words, AI-assisted troubleshooting has made performance management into a shared responsibility. And because Cloud SQL delivers a consistent experience, our teams can move seamlessly between environments. There’s less training – and a lot more collaboration. The end result is a smarter, more unified data culture at CME Group.

Performance is our competitive advantage

The work we’re doing with Google Cloud is about more than modernization. Every improvement in speed, reliability, and visibility translates directly into business confidence. CME Group can now deploy new features faster while maintaining the continuity our clients depend on.

Cloud SQL has given us a foundation for that agility. Fewer performance issues mean more time focused on innovation: expanding our analytics capabilities, accelerating AI initiatives, and exploring new ways to commercialize data responsibly. When you stop chasing outages, it turns out you have more time to take bigger bets and build the future.

For us at CME Group, performance has always been the product. Now, it’s also the platform. We’re building the infrastructure with Google Cloud that keeps global markets moving and the intelligence that will shape what comes next.

Learn more:

Sign up for the new Cloud SQL free trial, a dedicated 30-day program designed to give both new and existing Google Cloud users hands-on access to premium, enterprise-grade features of Cloud SQL (PostgreSQL and MySQL).
Download this IDC report to learn how migrating to Cloud SQL can lower costs, boost agility, and speed up deployments.
Learn how Ford and Yahoo gained high performance and cut costs by modernizing with Cloud SQL.

AWS and Google Cloud collaborate to simplify multicloud networking

Sun, 30 Nov 2025 19:00:00 +0000

As organizations increasingly adopt multicloud architectures, the need for interoperability between cloud service providers has never been greater. Historically, however, connecting these environments has been a challenge, forcing customers to take a complex "do-it-yourself" approach to managing global multi-layered networks at scale.

To address these challenges and advance a more open cloud environment, Amazon Web Services (AWS) and Google Cloud collaborated to transform how cloud service providers could connect with one another in a simplified manner.

Today, AWS and Google Cloud are excited to announce a jointly engineered multicloud networking solution that uses both AWS Interconnect - multicloud and Google Cloud’s Cross-Cloud Interconnect. This collaboration also introduces a new open specification for network interoperability, enabling customers to establish private, high-speed connectivity between Google Cloud and AWS with high levels of automation and speed.

“Integrating Salesforce Data 360 with the broader IT landscape requires robust, private connectivity. AWS Interconnect - multicloud allows us to establish these critical bridges to Google Cloud with the same ease as deploying internal AWS resources, utilizing pre-built capacity pools and the tools our teams already know and love. This native, streamlined experience — from provisioning through ongoing support — accelerates our customers' ability to ground their AI and analytics in trusted data, regardless of where it resides.” - Jim Ostrognai, SVP Software Engineering, Salesforce

Previously, to connect cloud service providers, customers had to manually set up complex networking components including physical connections and equipment; this approach required lengthy lead times and coordinating with multiple internal and external teams. This could take weeks or even months. AWS had a vision for developing this capability as a unified specification that could be adopted by any cloud service provider, and collaborated with Google Cloud to bring it to market.

Now, this new solution reimagines multicloud connectivity by moving away from physical infrastructure management toward a managed, cloud-native experience. By integrating AWS with Google Cloud’s Cross-Cloud Network architecture, we are abstracting the complexity of physical connectivity, network addressing, and routing policies. Customers no longer need to wait weeks for circuit provisioning: they can now provision dedicated bandwidth on demand and establish connectivity in minutes through their preferred cloud console or API.

Reliability and security are the cornerstone of this collaboration. We have collaborated on this solution to deliver high resiliency by leveraging quad-redundancy across physically redundant interconnect facilities and routers. Both providers engage in continuous monitoring to proactively detect and resolve issues. And this solution is built on a foundation of trust, utilizing MACsec encryption between the Google Cloud and AWS edge routers.

“This collaboration between AWS and Google Cloud represents a fundamental shift in multicloud connectivity. By defining and publishing a standard that removes the complexity of any physical components for customers, with high availability and security fused into that standard, customers no longer need to worry about any heavy lifting to create their desired connectivity. When they need multicloud connectivity, it's ready to activate in minutes with a simple point and click.” - Robert Kennedy, VP of Network Services, AWS

“We are excited about this collaboration which enables our customers to move their data and applications between clouds with simplified global connectivity and enhanced operational effectiveness. Today's announcement further delivers on Google Cloud’s Cross-Cloud Network solution focused on delivering an open and unified multicloud experience for customers.” - Rob Enns, VP/GM of Cloud Networking, Google Cloud

This collaboration between AWS and Google Cloud is more than a multicloud solution: it’s a step toward a more open cloud environment. The API specifications developed for this product are open for other providers and partners to adopt, as we aim to simplify global connectivity for everyone. We invite you to explore this new capability today. To learn more about how to streamline your multicloud operations please visit the in-depth Google Cloud Cross-Cloud Interconnect blog and the AWS Interconnect - multicloud website to get started.

How Lightricks trains video diffusion models at scale with JAX on TPU

Tue, 11 Nov 2025 17:00:00 +0000

Training large video diffusion models at scale isn't just computationally expensive — it can become impossible when your framework can't keep pace with your ambitions.

JAX has become a popular computational framework across AI applications, now recognized for its capabilities in training large-scale AI models, such as LLMs and life sciences models. Its strength lies not just in performance but in an expressive, scalable design that gives innovators the tools to push the boundaries of what's possible. We're consistently inspired by how researchers and engineers leverage JAX's ecosystem to solve unique, domain-specific challenges — including applications for generative media.

Today, we're excited to share the story of Lightricks, a company at the forefront of the creator economy. Their LTX-Video team is building high-performance video generation models, and their journey is a masterclass in overcoming technical hurdles. I recently spoke with Yoav HaCohen and Yaki Bitterman, who lead the video and scaling teams, respectively. They shared their experience of hitting a hard scaling wall with their previous framework and how a strategic migration to JAX became the key to unlocking the performance they needed.

Here, Yoav and Yaki tell their story in their own words. – Srikanth Kilaru, Senior Product Manager, Google ML Frameworks

The creator's challenge

At Lightricks, our goal has always been to bring advanced creative technology to consumers. With apps like Facetune, we saw the power of putting sophisticated editing tools directly into people's hands. When generative AI emerged, we knew it would fundamentally change content creation.

We launched LTX Studio to build generative video tools that truly serve the creative process. Many existing models felt like a "prompt and pray" experience, offering little control and long rendering times that stifled creativity. We needed to build our own models—ones that were not only efficient but also gave creators the controllability they deserve.

Our initial success came from training our first real-time video generation model on Google Cloud TPUs with PyTorch/XLA. But as our ambitions grew, so did the complexity. When we started developing our 13-billion-parameter model, we hit a wall.

Hitting the wall and making the switch

Our existing stack wasn’t delivering the training step times and scalability we needed. After exploring optimization options, we decided to shift our approach. We paused development to rewrite our entire training codebase in JAX, and the results were immediate. Switching to JAX felt like a magic trick, instantly providing the necessary runtimes.

This transition enabled us to effectively scale our tokens per sample (the amount of data processed in each training step), model parameters, and chip count. With JAX, sharding strategies (sharding divides large models across multiple chips) that previously failed now work out of the box on both small and large pods (clusters of TPU chips).

These changes delivered linear scaling that translates to 40% more training steps per day — directly accelerating model development and time to market. Critical issues with FlashAttention and data loading also worked reliably. As a result, our team's productivity skyrocketed, doubling the number of pull requests we could merge in a week.

Why JAX worked: A complete ecosystem for scale

The success wasn't just about raw speed; it was about the entire JAX stack, which provided the building blocks for scalable and efficient research.

A clear performance target with MaxText: We used the open-source MaxText framework as a baseline to understand what acceptable performance looked like for a large model on TPUs. This gave us a clear destination and the confidence that our performance goals were achievable on the platform.
A robust toolset: We built our new stack on the core components of the JAX ecosystem based on the MaxText blueprint. We used Flax for defining our models, Optax for implementing optimizers, and Orbax for robust checkpointing — all core components that work together natively.
Productive development and testing: The transition was remarkably smooth. We implemented unit tests to compare our new JAX implementation with the old one, ensuring correctness every step of the way. A huge productivity win was discovering that we could test our sharding logic on a single, cheap CPU before deploying to a large TPU slice. This allowed for rapid, cost-effective iteration.
Checkpointing reliability: For sharded models, JAX’s checkpointing is much more reliable than before, making training safer and more cost-effective.
Compile speed & memory: JAX compilation with lax.fori_loop is fast and uses less memory, freeing capacity for tokens and gradients.
Smooth scaling on a supercomputer: With our new JAX codebase, we were able to effectively train on a reservation of thousands of TPU cores. We chose TPUs because Google provides access to what we see as a "supercomputer" — a fully integrated system where the interconnects and networking were designed first, not as an afterthought. We manage these large-scale training jobs with our own custom Python scripts on Google Compute Engine (GCE), giving us direct control over our infrastructure. We also use Google Cloud Storage and stream the training data to the TPU virtual machines.

Architectural diagram showing the Lightricks stack

Build your models with the JAX ecosystem

Lightricks' story is a great example of how JAX's powerful, modular, and scalable design can help teams overcome critical engineering hurdles. Their ability to quickly pivot, rebuild their stack, and achieve massive performance gains is a testament to both their talented team and the tools at their disposal.

The JAX team at Google is committed to supporting innovators like Lightricks and the entire scientific computing community.

Share your story: Are you using JAX to tackle a challenging scientific problem? We would love to learn how JAX is accelerating your research.
Help guide our roadmap: Are there new features or capabilities that would unlock your next breakthrough? Your feature requests are essential for guiding the evolution of JAX.

Please reach out to the team via GitHub to share your work or discuss what you need from JAX. Check out documentation, examples, news, events and more at jaxstack.ai and jax.dev.

Sincere thanks to Yoav, Yaki, and the entire Lightricks team for sharing their insightful journey with us. We're excited to see what they create next.

11 ways to reduce your Google Cloud compute costs today

Mon, 06 Oct 2025 16:00:00 +0000

As the saying goes, "a penny saved is a penny earned," and this couldn't be more true when it comes to cloud infrastructure. In today's competitive business landscape, you need to maintain the performance to meet your business needs. Luckily, Google Cloud’s Compute Engine and block storage services offer numerous opportunities to reduce costs without sacrificing performance, especially in the context of your migration and modernization initiatives.

In this article, we'll explore 11 key ways to optimize your infrastructure spending on Google Cloud, from simple adjustments to strategic decisions that can result in significant long-term savings.

1. Choose the right VM instances

One of the most effective ways to reduce Compute Engine costs is to ensure that you’ve properly selected and right-sized your virtual machines (VMs) for their workloads to support your migration and modernization efforts. Whether you're new to Google Cloud or already using Compute Engine, adopting the latest-generation VMs — such as N4, C4, C4D, and C4A — can deliver substantial savings and improved price-performance.

Powered by Google Cloud’s Titanium architecture, our latest-generation VMs offer faster CPUs, higher memory bandwidth, and more efficient virtualization than their predecessors, so you can handle the same workloads with fewer resources. For existing customers, migrating from older VM generations to the newest VMs can significantly lower total costs while helping you exceed current performance levels. Organizations that have made the switch often report 20–40% better performance along with meaningful reductions in cloud compute spend. For example, Elastic leveraged the general-purpose C4A machine series based on Google Cloud's Arm-based Axion CPUs, to achieve a compelling efficiency and performance uplift for their workloads.

Beyond general-purpose VMs, we also offer specialized machine types to address unique customer requirements. Compute-optimized HPC VMs like H4D are designed for high-performance computing and data analytics, offering extreme performance for demanding workloads. M4 and X4 instances cater to memory-intensive applications, while Z3 instances are ideal for storage-intensive workloads. Furthermore, if you need complete control over your hardware environment and maximum performance isolation, we offer bare metal instances.

These options help ensure that even the most specialized and performance-sensitive workloads can find an optimal and cost-effective home within the Compute Engine portfolio.

2. Optimize your block storage selections

The best way to lower your block storage TCO, while ensuring your workloads remain successful, is to drive high resource efficiency. Hyperdisk makes it simple to drive high performance and high efficiency by enabling you to optimize your block storage to your workload and through Storage Pools. We’ll discuss each of these capabilities, and how you can use them to lower your block storage TCO below.

Workload Optimized: With Hyperdisk, you can independently tune capacity and performance to match your block storage resources to your workload. Hyperdisk enables you to independently provision performance and capacity at the volume level. You can leverage this capability to purchase just the capacity and performance you need, no more and no less. You can also take advantage of Hyperdisk Balanced’s “baseline” performance (i.e. included free with every volume), you can serve the vast majority of your VMs without purchasing any extra performance.

Storage Pools: Hyperdisk is the only hyperscale cloud block storage to offer thin-provisioned performance and capacity. With Hyperdisk Storage Pools, you can provision the aggregate performance and capacity your workload requires, while still provisioning the volume level capacity performance your workloads request (also known as thin-provisioning). This allows you to pay for the resources you need, not the sum of the volumes you’ve provisioned. As a result, you can lower your overall block storage TCO by as much as 50%.

For more information on how to select the right block storage for your workload and to see how customers have benefitted from Hyperdisk, read this blog.

3. Consider custom compute classes

To get the most out of our latest-generation VMs, Google Kubernetes Engine (GKE) custom compute classes (CCC) offer an advanced way to optimize compute choices and provide high availability. Instead of being limited to a single machine type for your workloads, you can define a prioritized list of VM instance types. This allows you to set the newest, most price-performant VMs — including our latest-generation VMs — as your top priority. GKE custom compute classes provide the capability to automatically and seamlessly spin up instances based on your specified priority list. This feature helps you maximize the availability of your compute capacity while still aiming for the most cost-effective options, so your workloads can scale reliably without manual intervention.

Here are some specific use cases for how custom compute classes can help you optimize costs:

Autoscaling cost-performant fallbacks: When demand peaks, you might be tempted to autoscale using a highly available but less cost-efficient VM type. CCC allows you to take a tiered approach. You can set up several cost-efficient fallback alternatives, so that as demand increases, GKE first attempts to use the most cost-effective options, and progressively moves to the other choices in your list when necessary to meet demand.
AI/ML inference: Running AI/ML inference workloads often involves significant compute resources. Instead of maintaining a large, static reservation that might sit idle during off-peak times, CCC lets you provision a minimal base reservation and leverage more cost-effective capacity types, such as Spot VMs, to handle peak inference demand — all orchestrated through your CCC configuration.
Adopting new VM generations: Combine the power of GKE custom compute classes with Compute Flexible committed use discounts (Flex CUDs) to de-risk the adoption of new, cost-efficient VM series like N4 and C4. With CCC, you can define fallback options, providing workload resilience, while Flex CUDs offer financial adaptability, as the discounts apply across your total eligible compute spend, regardless of the specific VM series you use. This dual approach is a safe, cost-effective strategy for leveraging the latest hardware without disruption. For more information, read this blog.
Using flexible Spot VMs: Spot VMs offer significant savings but can be preempted. Being constrained to a single Spot VM shape increases the risk that capacity will not be available. With CCC, you can define multiple fallback Spot VM types. This "spot surfing" capability allows the application to remain on cost-efficient Spot capacity by automatically pivoting to alternative Spot instance types if the primary choice is unavailable.

In short, by leveraging GKE CCC, you can artfully mix and match various VM types and consumption models, including On-Demand, Spot, DWS FlexStart, and instances covered by CUDs, to build a resilient and highly cost-optimized infrastructure that adapts to the unique needs and patterns of your workloads.

4. Leverage custom machine types (CMT)

Custom machine types, available on N4 VMs, allow you to precisely configure virtual machines to your exact specifications. Rather than selecting from predefined machine types that might include excess capacity, you can tailor the CPU-to-memory ratio specifically for your workloads, so you only pay for resources you actually use. This targeted approach minimizes waste and can significantly reduce your cloud spend, especially when migrating from on-premises to Google Cloud or from other cloud providers.

This flexibility becomes particularly valuable if your applications have unique resource profiles that don't align well with our standard offerings. Custom machine types let you create the perfect environment for your needs. By avoiding the compromise of over-provisioning certain resources while potentially constraining others, you can achieve both better performance and more efficient spending across your Compute Engine deployment.

As an example, take a memory-intensive workload that runs best with 16 vCPU, and 70 GB memory. Normally, you would need to pick a VM with 128 GB memory with our standard shapes, or in other cloud contexts, resulting in higher costs to run your workload due to the extra provisioned resources. Instead, with custom machine types, you can easily launch a VM with 16 vCPU and 70 GB memory, resulting in an 18% cost savings vs standard N4-highmem-16 VMs.

5. Make the most of committed use discounts

CUDs are a strategic cost-saving opportunity for organizations with steady, predictable computing needs. By committing to resource usage over one- or three-year periods, you can reduce cloud costs by up to 70% compared to on-demand pricing. This approach not only helps ensure budget predictability but also converts fixed infrastructure spending into a financial advantage, making it ideal for stable workloads that support core business functions.

Google Cloud offers flexible CUD structures to align with various operational models. Resource-based commitments target specific machine types and regions, flexible commitments apply discounts across projects, regions, and machine series — great for dynamic environments. By analyzing historical usage and forecasting future needs, you can identify workloads suited for these discounts, reinvesting the savings into innovation and scaling initiatives.

6. Manage unused disk space

You pay for the total provisioned disk space, regardless of how much you actually use. Many organizations tend to over-provision storage "just in case," which often leads to unnecessary and costly waste. For instance, if you provision a 100GB disk but only use 20GB, you're still paying for the entire 100GB. Being intentional and precise with your storage allocations — rather than rounding up to common sizes — can lead to significant cost savings.

To optimize spending, it's important to adopt a few best practices. Using Ops Agent, regularly audit disk usage across your infrastructure to identify and eliminate inefficiencies. Resize disks to align with actual consumption, allowing a reasonable buffer for growth. Implement automated alerts in Cloud Monitoring to detect underutilized disks and take corrective action. For stateless applications, consider using smaller boot disk images to minimize overhead and reduce costs even further.

In addition, consider the following optimization strategies to further reduce costs and improve efficiency:

Use Google Cloud’s monitoring tools to track CPU, memory, and disk usage over time.
Establish a regular review cycle to identify and right-size over-provisioned resources.
Test workloads across different VM configurations to find the optimal balance between cost and performance.

7. Use Spot VMs

Spot VMs provide the same machine types and configuration options as standard virtual machines but at a significantly reduced cost — typically offering a 60% to 91% discount. This cost efficiency comes with the tradeoff of potential preemption at short notice, making them most suitable for workloads that are fault-tolerant and can recover quickly from unexpected interruptions. Spot VMs are designed to take advantage of unused compute capacity, allowing you to optimize your cloud spending without compromising access to high-performance resources.

Strong use cases for Spot VMs include batch processing jobs, big data and analytics workloads, continuous integration and deployment (CI/CD) pipelines, stateless web servers running in autoscaling groups, and compute-heavy tasks. When properly architected to handle interruptions — for example, by using job checkpointing, load balancing, task queues, or via GKE custom compute classes (see more above) — Spot VMs can play a critical role in minimizing infrastructure costs while maintaining high availability and system resilience. Leveraging Spot VMs in these scenarios lets you scale cost-effectively, especially when compute demand is variable or time-flexible.

8. Use optimization recommendations

Google Cloud's Recommenders are a powerful tool designed to help you optimize your cloud resources efficiently. When browsing the Google Cloud console, you may see lightbulb icons next to specific resources — these indicate potential improvements identified by Google's recommendation engine. By analyzing real-time usage patterns and current resource configurations, the Recommender delivers actionable insights tailored to each user's unique environment. This intelligent system highlights opportunities not only to reduce costs but also to enhance security, performance, reliability, management efficiency, and environmental sustainability.

For example, there are idle VM recommendations to help you identify VM instances that have not been used over the last 1 to 14 days. Common recommendations include switching to more suitable machine types, rightsizing underutilized compute instances, or adopting more cost-effective storage solutions. The tool allows you to apply many of these changes directly, streamlining the optimization process. By continuously evaluating workloads and offering these automated, data-driven suggestions, the Recommendation Hub helps organizations maintain cloud performance while managing costs more effectively.

9. Take advantage of auto-scaling and scheduling

Matching your compute resources to actual demand patterns is one of the most effective ways to reduce cloud waste and improve overall cost efficiency. Many organizations over-provision their resources to handle peak workloads, leaving machines underutilized during off-peak periods. By aligning compute capacity more closely with real-time or predictable usage patterns, such as business hours or seasonal trends, you can significantly cut unnecessary spending without sacrificing performance.

Autoscaling is the key to achieving this efficiency. In fact, customers who leverage Google Compute Engine autoscaling for their virtual machines have seen average infrastructure cost savings of more than 40%.

You can implement autoscaling strategies to dynamically adjust resources based on CPU utilization, load balancing capacity, or custom application metrics, so that workloads receive the necessary compute power when needed, while scaling down automatically during low-demand periods.

For workloads with predictable patterns, such as those that fluctuate with business hours or planned seasonal events, schedule-based scaling is a particularly powerful tool. This approach allows you to proactively increase resources in anticipation of high demand and scale them down during lulls, for the performance you need without constant over-provisioning.

In addition to autoscaling, several practical implementation techniques can further optimize your resource usage. Setting up instance scheduling lets you automatically start and stop development and test environments according to business hours — a simple yet highly effective approach that can lead to cost savings of up to 70%. You can also leverage maintenance windows to reduce disruptions and resource consumption, by concentrating updates and system changes into low-usage periods. Together, these tactics help maintain high availability and performance while keeping infrastructure costs under control.

10. Understand your spend with detailed billing analysis

Before implementing any cost-saving strategies in Google Cloud, it’s essential to understand your current spending in detail. Google Cloud’s billing panel offers granular visibility into your expenses, including costs broken down by individual SKUs. This level of transparency lets you track where your money is going and identify potential inefficiencies. Begin by regularly reviewing your billing dashboard to monitor usage trends and spot anomalies. Applying labels and tags to your resources can further help categorize and attribute costs accurately, especially in complex environments with multiple projects or departments.

In addition, setting up budget alerts is a practical way to stay ahead of overspending by notifying you when costs approach or exceed predefined thresholds. It’s also important to identify and eliminate unused or idle resources, such as virtual machines or persistent disks that are no longer in active use — these can often be shut down or deleted to immediately reduce costs. By thoroughly analyzing your cost structure, you can uncover “low-hanging fruit” — resources that provide little or no value — and make data-driven decisions to optimize your cloud usage efficiently.

11. Consider serverless alternatives

Last but not least, Google Cloud's serverless computing offerings provide a compelling alternative to traditional virtual machines, can deliver better cost efficiency, simplified operations, and greater scalability. By abstracting away infrastructure management, serverless platforms allow teams to focus on writing and deploying code without worrying about provisioning, scaling, or maintaining servers. This shift can not only reduce operational overhead but also cut costs by aligning compute spending directly with application usage.

There are multiple serverless options available, each tailored to different workloads. Cloud Run is designed for running containerized applications that need rapid scaling and flexible deployment. Cloud Run Functions supports lightweight, event-driven code execution for microservices or automation tasks. GKE (Autopilot Mode) simplifies Kubernetes operations by automatically managing nodes and scaling, allowing you to run Kubernetes workloads without handling the underlying infrastructure. All these options charge based on usage not allocation, significantly reducing costs associated with idle resources and over-provisioning. This makes them especially beneficial for variable or unpredictable workloads. Cloud Run and GKE both support GPU’s and flexibility to move between the two. You can start with Cloud Run then move to GKE or vice-versa. Some customers also leverage both offerings for workloads. The rule of thumb is to start with GKE if you need access to the Kubernetes API. Otherwise, start with Cloud Run.

Start reducing your costs today

Migrate to Google Cloud and optimize your infrastructure costs without compromising on what your workloads need. If you are new to Google Cloud, start with a migration assessment. Google Cloud’s Migration Center can help you with a clear understanding of your potential savings by migrating to Google Cloud, with detailed recommended paths for your workloads, along with TCO reports. Apply the strategies in this article and unlock substantial cost savings.

How Baseten achieves 225% better cost-performance for AI inference (and you can too)

Thu, 04 Sep 2025 17:00:00 +0000

Baseten is one of a growing number of AI infrastructure providers, helping other startups run their models and experiments at speed and scale. Given the importance of those two factors to its customers, Baseten has just passed a significant milestone.

By leveraging the latest Google Cloud A4 virtual machines (VMs) based on NVIDIA Blackwell, and Google Cloud’s Dynamic Workload Scheduler (‘DWS’) Baseten has achieved 225% better cost-performance for high-throughput inference and 25% better cost-performance for latency-sensitive inference.

Why it matters: This breakthrough in performance and efficiency enables companies to move powerful agentic AI and reasoning models out of the lab and into production affordably. For technical leaders, this provides a blueprint for building next-generation AI products — such as real-time voice AI, search, and agentic workflows — at a scale and cost-efficiency that has been previously unattainable.

The big picture: Inference is the cornerstone of enterprise AI. As models for multi-step reasoning and decision-making demand exponentially greater compute, the challenge of serving them efficiently has become the primary bottleneck. Enter Baseten, a six-year-old Series C company that partners with Google Cloud and NVIDIA to provide enterprise companies a scalable inference platform for their proprietary models as well as open models like Gemma, DeepSeek ,and Llama, with an emphasis on performance and cost efficiency. Their success hinges on a dual strategy: maximizing the potential of cutting-edge hardware and orchestrating it with a highly optimized, open software stack.

We wanted to share more about how Baseten architected its stack — and what this new level of cost-efficiency can unlock for your inference applications.

aside_block: <ListValue: [StructValue([('title', '$300 in free credit to try Google Cloud AI and ML'), ('body', <wagtail.rich_text.RichText object at 0x7f2e26dcc3a0>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Hardware optimization with the latest NVIDIA GPUs

Baseten delivers production-grade inference by leveraging a wide range of NVIDIA GPUs on Google Cloud, from NVIDIA T4s through the recent A4 VMs (NVIDIA HGX B200). This access to the latest hardware is critical for achieving new levels of performance.

With A4 VMs, Baseten now serves three of the most popular open-source models — DeepSeek V3, DeepSeek R1, and Llama 4 Maverick — directly on their Model APIs with over 225% better cost-performance for high throughput inference, and 25% better cost-performance for latency- sensitive inference.
In addition to its production-ready model APIs, Baseten provides additional flexibility with NVIDIA B200-powered dedicated deployments for customers seeking to run their own custom AI models with the same reliability and efficiency.

Advanced software for peak performance

Baseten’s approach is rooted in coupling the latest accelerated hardware with leading and open-source software to extract the most value possible from every chip. This integration is made possible with Google Cloud’s AI Hypercomputer, which includes a broad suite of advanced inference frameworks, including NVIDIA’s open-source software stack — NVIDIA Dynamo and TensorRT-LLM — as well as SGLang and vLLM.

Using TensorRT-LLM, Baseten optimizes and compiles custom LLMs for one of its largest AI customers, Writer. This has boosted their throughput by more than 60% for Writer’s Palmyra LLMs. The flexibility of TensorRT-LLM also enabled Baseten to develop a custom model builder that speeds up model compilation.
To serve reasoning models like DeepSeek R1 and Llama 4 on NVIDIA Blackwell GPUs, Baseten uses NVIDIA Dynamo. The combination of NVIDIA’s HGX B200 and Dynamo dramatically lowered latency and increased throughput, propelling Baseten to the top GPU performance spot on OpenRouter’s LLM ranking leaderboard.
The team leverages techniques such as kernel fusion, memory hierarchy optimization, and custom attention kernels to increase tokens per second, reduce time to first token, and support longer context windows and larger batch sizes — all while maintaining low latency and high throughput.

Building a backbone for high availability and redundancy

For mission-critical AI services, resilience is non-negotiable. Baseten runs globally across multiple clouds and regions, requiring an infrastructure that can handle ad hoc demand and outages. Flexible consumption models, such as the Dynamic Workload Scheduler within the AI Hypercomputer, help Baseten manage capacity similar to on-demand with additional price benefits. This allows them to scale up on Google Cloud if there are outages across other clouds.

"Baseten runs globally across multi-clouds and Dynamic Workload Scheduler has saved us more than once when we encounter a failure,” said Colin McGrath, head of infrastructure at Baseten. “Our automated system moves affected workloads to other resources including Google Cloud Dynamic Workload scheduler and within minutes, everyone is up and running again. It is impressive — by the time we’re paged and check-in, everything is back and healthy. This is amazing and would not be possible without DWS. It has been the backbone for us to run our business.”

Baseten’s scalable inference platform architecture

Unlocking new AI applications for end-users

Baseten's collaboration with Google Cloud and NVIDIA demonstrates how a powerful combination of cutting-edge hardware and flexible, scalable cloud infrastructure can solve the most pressing challenges in AI inference through Google Cloud’s AI Hypercomputer.

This unique combination enables end-users across industries to bring new applications to market, such as powering agentic workflows in financial services, generating real-time audio and video content in media, and accelerating document processing in healthcare. And it’s all happening at a scale and cost that was previously unattainable.

You can easily get started with Baseten's platform through the Google Cloud Marketplace, or read more about their technical architecture in their own post.

From clicks to clusters: Expanding Confidential Computing with Intel TDX

Fri, 29 Aug 2025 16:00:00 +0000

Privacy-protecting Confidential Computing has come a long way since we introduced Confidential Virtual Machines (VMs) five years ago. The technology, which can protect data while in use, strengthens a security gap beyond data encryption at rest and in transit.

Since then, customers have used Confidential Computing to protect patient medical data, comply with privacy guidance of GDPR and Schrems II for U.S.-Europe data transfers, and run high-performance computing (HPC) workloads securely.

By isolating workloads in hardware-based Trusted Execution Environments (TEEs), Confidential Computing empowers customers to process their most sensitive information in the public cloud with assurance.

As part of the advancements we’ve made with Confidential Computing, we added even more security capabilities with the introduction of Confidential VMs with Intel Trust Domain Extensions (TDX) last year. Intel TDX creates an isolated trust domain (TD) in a VM, uses hardware extensions for managing and encrypting memory to protect cloud workloads, and offers hardware-based remote attestation for verification.

Today, we are excited to highlight our greatly expanded, and generally available, Intel TDX-based offerings, which includes Confidential GKE Nodes, Confidential Space, Confidential GPU, and more regions and zones where customers can use Confidential Computing.

Click to create a Confidential VM

Google Cloud Console now offers Google Compute Engine (GCE) customers a new interface for Intel TDX — no code changes required. To get started, follow these steps:

Start at the GCE Create an instance page
Go to the Security tab and under Confidential VM service, click Enable
Then select Intel TDX from the dropdown menu and click Confirm.

It’s that simple to create a Confidential VM.

Create a new Confidential VM with Intel TDX in the Google Cloud console.

Get Confidential Computing in more regions and zones

Confidential VMs with Intel TDX were first available with support for three regions (and nine zones.) To accommodate growing demand, we’ve expanded support for Intel TDX on the C3 machine series to 10 regions (and 21 zones,) and we are planning more for the future. The full list is available here. As regional availability and scalability are critical, your account team is available to help you plan early to ensure your capacity needs are met.

Confidential GKE Nodes with Intel TDX, now generally available

Confidential GKE Nodes are built on top of Confidential VM and deliver hardware-based protections to your Google Kubernetes Engine (GKE) clusters and node pools to ensure that your containerized workloads remain encrypted in memory. Today, Confidential GKE Nodes are generally available with Intel TDX on GKE Standard and GKE Autopilot.

Confidential GKE Nodes with Intel TDX on the C3 machine series can be created on GKE Standard via CLI, API, UI, and Terraform. The confidential setting can be set at the cluster level or the node pool level with no code changes. You can learn more here.

Confidential GKE Nodes with Intel TDX on the C3 machine series can also be created on GKE Autopilot. It can be enabled through the use of custom compute classes. In GKE, a compute class is a profile that consists of a set of node attributes that GKE uses to provision the nodes that run your workloads during autoscaling events. Check out our documentation to get started.

aside_block: <ListValue: [StructValue([('title', '$300 in free credit to try Google Cloud security products'), ('body', <wagtail.rich_text.RichText object at 0x7f2dfccc4d60>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Confidential Space with Intel TDX, now generally available

Also built on Confidential VM, our Confidential Space offering is a robust solution for many common issues including addressing insider threats, enabling joint machine-learning training and private gen AI inference, and fostering multi-party collaboration on sensitive data. Here are just a few examples of what our customers have built with Confidential Space:

Confidential matching enabled customers to securely connect their first-party data for Google Ads measurement and audience solutions.
Symphony demonstrated with its Confidential Cloud how SaaS companies can guarantee isolation of customer data from privileged insiders in the highly regulated financial industry.
Duality delivered privacy-preserving federated learning solutions for a broad range of use cases in healthcare, financial services, and the public sector.
Flare spearheaded innovation in verifiable AI on blockchain.

Previously, Confidential Space was only available with AMD-based technology and hardware (on the N2D, C2D, C3D, and C4D machine series), but now it is also available with Intel-based technology and hardware. This is ideal for those wanting attestation guarantees with a hardware root of trust and for those focused on Intel’s C3 machine series.

Additionally, Confidential Space with Intel TDX is measured into runtime measurement registers (RTMR) and the measurements are verified by Google Cloud Attestation. Note that for Confidential VMs with Intel TDX, RTMRs are now populated as well. Confidential Space benefits are highlighted in the NCC Group’s latest independent security evaluation.

Confidential VM and Confidential GKE Nodes with NVIDIA H100 GPUs, now generally available

If you’re looking for performance and security while protecting data in use, Confidential VM and Confidential GKE Nodes with NVIDIA H100 GPUs on the accelerator-optimized A3 machine series are now generally available. These offerings deliver Google Cloud’s first Confidential GPUs, focus on ease of use to meet the demand for secure computing, and extend security to data-intensive, AI and ML workloads by having Intel TDX enabled on the CPU and NVIDIA Confidential Computing enabled on the GPU. You now have the ability to secure your data performantly during inference and training across models.

Confidential VM with NVIDIA H100 GPUs is available with the a3-highgpu-1g machine type and in three zones: europe-west4-c, us-central1-a, and us-east5-a. No code changes are needed for most AI and ML workloads. For pricing details, see here. Confidential GKE Nodes with NVIDIA H100 GPUs are generally available on both GKE Standard and GKE Autopilot (through custom compute class). To get started, click here.

And, we also have Confidential Space with NVIDIA H100 GPUs in preview.

Intel has a free tier for independent attestation

Intel’s attestation verifier service, Intel Tiber Trust Authority, now has a free tier. Google Cloud Confidential VMs and Confidential Space are both integrated with Intel Tiber Trust Authority as a third party attestation service, but now Intel Tiber Trust Authority is making secure attestation more accessible for all by offering a free tier (with optional paid support).

When Confidential VM and Confidential Space customers use Intel Tiber Trust Authority, they can gain stronger separation of duties security guarantees. Click here to learn more.

What our customers say

"Thanks to the joint efforts of Super Protocol, Google Cloud, and NVIDIA, the world now gains a new layer of possibility — unlocking Confidential AI without cloud borders. With A3 Confidential VMs built on NVIDIA H100 GPUs now integrated into Super’s decentralized infrastructure and marketplace, companies can securely run, monetize, and collaborate on sensitive AI and data — across any environment. This enables seamless collaboration between Google Cloud customers and partners in other clouds — with no need for shared trust, manual agreements, or compromise. For the broader market, A3 instances at scale accelerate global access, while Super ensures confidentiality, verifiability, and self-sovereignty — fully automated and requiring no expertise in confidential computing. We are excited to open this next chapter of Confidential AI, built to work wherever you and your partners are," said Nukri Basharuli, founder and CEO, Super Protocol.

“We’re proud to have partnered with Google Cloud to validate their Confidential Computing-enabled GPU solution — a major step forward in securing sensitive data for AI and machine learning workloads, without compromising on performance or scalability. Confidential Computing allows organizations to process sensitive workloads in the cloud while protecting sensitive data and models from both the cloud provider and the organization's insiders and internal threats. However, for gen AI and agentic AI use cases, protecting the CPU alone isn’t enough — both CPU and GPU must also run in confidential mode with mutual trust. With Google Cloud’s new offering, Anjuna can now launch Confidential Containers that leverage Intel TDX and NVIDIA H100 GPUs in confidential mode. This ensures that data, configurations, secrets, and code remain protected end-to-end from any untrusted entity, bringing state-of-the-art security for sensitive data.” said Steve Van Lare, CTO, Anjuna Security.

“With data processing worldwide growing up to three times faster than ever before and doubling every six months, the future of cloud computing must be built on trust. In collaboration with Google, Modelyo leverages Confidential VMs on the A3 machine series with NVIDIA H100 GPUs, transforming Confidential Computing into a seamless, intuitive, and fully integrated cloud experience. This enables us to deliver end-to-end managed solutions across interconnected environments, empowering organizations to innovate confidently knowing their data remains effortlessly protected at every stage.” said Benny Meir, CEO, Modelyo.

How to get started with Confidential Computing

To add that extra layer of protection and privacy to your sensitive workloads, check out our documentation for Confidential VMs and Confidential GKE Nodes today.

Run Gemini anywhere, including on-premises, with Google Distributed Cloud

Thu, 28 Aug 2025 05:00:00 +0000

Earlier this year, we announced our commitment to bring Gemini to on-premises environments with Google Distributed Cloud (GDC). Today, we are excited to announce that Gemini on GDC is now available to customers.

For years, enterprises and governments with the strictest data security and sovereignty requirements have faced a difficult choice: adopt modern AI or protect their data. Today, that compromise ends. We are announcing the general availability of Gemini on GDC air-gapped and preview of Gemini on GDC connected, bringing Google's most advanced models directly into your data center.

We are inspired by initial feedback from customers, including Singapore’s Centre for Strategic Infocomm Technologies (CSIT), Government Technology Agency of Singapore (GovTech Singapore), Home Team Science and Technology Agency (HTX), KDDI, and Liquid C2, who are excited to gain the advantages of generative AI with Gemini on GDC.

Transformative AI capabilities, on-premises

Gemini models offer groundbreaking capabilities, from processing extensive context to native multimodal understanding of text, images, audio, and video. This unlocks a wide array of high-impact use cases on secure infrastructure:

Unlock new markets and global collaboration: Instantly break down language barriers across your international operations, creating a more connected and efficient global workforce.
Accelerate decision-making: Make faster, data-driven decisions by using AI to automatically summarize documents, analyze sentiment, and extract insights from your proprietary datasets.
Improve employee efficiency and customer satisfaction: Deliver instant, 24/7 support and enhance user satisfaction by developing intelligent chatbots and virtual assistants for customers and employees.
Increase development velocity: Ship higher-quality software faster by using Gemini for automated code generation, intelligent code completion, and proactive bug detection.
Strengthen safety & compliance: Protect your users with AI-powered safety tools that automatically filter harmful content and ensure adherence to industry policies.

aside_block: <ListValue: [StructValue([('title', 'Try Google Cloud for free'), ('body', <wagtail.rich_text.RichText object at 0x7f2e29576d30>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Secure AI infrastructure where you need it

It takes more than just a model to drive business value with generative AI; you need a complete platform that includes scalable AI infrastructure, a library with the latest foundational models, high-performance inferencing services, and pre-built AI agents like Agentspace search. GDC provides all that and more with an end-to-end AI stack combining our latest-generation AI infrastructure with the power of Gemini models to accelerate and enhance all your AI workloads.

Delivering these transformative capabilities securely requires a complete, end-to-end platform that only Google is providing today :

Performance at scale: GDC utilizes the latest NVIDIA GPU accelerators, including the NVIDIA Hopper and Blackwell GPUs. A fully managed Gemini endpoint is available within a customer or partner data center, featuring a seamless, zero-touch update experience. High performance and availability are maintained through automatic load balancing and auto-scaling of the Gemini endpoint, which is handled by our L7 load balancer and advanced fleet management capabilities.
Foundation of security and control: Security is a core component of our solution, with audit logging and access control capabilities that provide full transparency for customers. This allows them to monitor all data traffic in and out of their on-premises AI environment and meet strict compliance requirements. The platform also features Confidential Computing support for both CPUs (with Intel TDX) and GPUs (with NVIDIA's confidential computing) to secure sensitive data and prevent tampering or exfiltration.
Flexibility and speed for your AI strategy: the platform supports a variety of industry-leading models including Gemini 2.5 Flash and Pro, Vertex AI task-specific models (translation, optical character recognition, speech-to-text, and embeddings generation), and Google’s open-source Gemma models. GDC also provides managed VM shapes (A3 & A4 VMs) and Kubernetes clusters giving customers the ability to deploy any open-source or custom AI model, and custom AI workloads of their choice. This is complemented by Vertex AI services that provide an end-to-end AI platform including a managed serving engine, data connectors, and pre-built agents like Agentspace search (in preview) for a unified search experience across on-premises data.

What our customers are saying

“As a key GDC collaboration partner in shaping the GDC air-gapped product roadmap and validating the deployment solutions, we’re delighted that this pioneering role has helped us grow our cutting-edge capabilities and establish a proven deployment blueprint that will benefit other agencies with similar requirements. This is only possible with the deep, strategic collaboration between CSIT and Google Cloud. We’re also excited about the availability of Gemini on GDC, and we look forward to building on our partnership to develop and deploy agentic AI applications for our national security mission.” - Loh Chee Kin, Deputy Chief Executive, Centre for Strategic Infocomm Technologies (CSIT)

“One of our priorities is to harness the potential of AI while ensuring that our systems and the services citizens and businesses rely on remain secure. Google Cloud has demonstrated a strong commitment to supporting the public sector with initiatives that enable the agile and responsible adoption of AI. We look forward to working more closely with Google Cloud to deliver technology for the public good.” - Goh Wei Boon, Chief Executive, Government Technology Agency of Singapore

“The ability to deploy Gemini on Google Distributed Cloud will allow us to bridge the gap between our on-premises data and the latest advancements in AI. Google Distributed Cloud gives us a secure, managed platform to innovate with AI, without compromising our strict data residency and compliance requirements.” - Ang Chee Wee, Chief AI Officer, Home Team Science & Technology Agency (HTX)

“The partnership with Google Cloud and the integration of Google's leading Gemini models will bring cutting-edge AI capabilities, meet specific performance requirements, address data locality and regulatory needs of Japanese businesses and consumers.” - Toru Maruta, Executive Officer, Head of Advancing Business Platform Division, KDDI

"Data security and sovereignty are paramount for our customers. With Gemini on Google Distributed Cloud, our Liquid Cloud and Cyber Security solution would deliver strategic value to ensure our customers in highly regulated industries can harness the power of AI while keeping their most valuable data under their control." - Oswald Jumira, CEO Liquid C2

Gemini everywhere is here

The era of on-premises AI without compromise is here. To bring the power of Gemini to your on-premises environment, request a strategy session with our experts.

An efficient path to production AI: Kakao’s journey with JAX and Cloud TPUs

Tue, 19 Aug 2025 16:00:00 +0000

When your messaging platform serves 49 million people – 93% of South Korea’s population – every technical decision carries enormous weight. The engineering team at Kakao faced exactly this challenge when their existing infrastructure hit critical limitations. Their solution? A strategic shift to Google Cloud TPUs using the JAX framework that not only solved their immediate scalability needs but opened new possibilities for advanced AI model development.

Kakao’s approach provides a compelling example of leveraging the high-performance array computing framework JAX for AI model development at scale. While their primary training environment was GPU-based, the team made a strategic decision to adopt the JAX stack on Google Cloud TPUs to optimize for cost and efficiency.

This work laid the groundwork for the development of their proprietary Kanana model family, and several Kanana models — including Kanana-MoE — have recently been released as open source on Hugging Face Hub.

In this post, Minho Ryu and Nayeon Kim detail Kakao’s technical journey. They cover their specific implementation details, from adapting the JAX large language model framework and MaxText for custom data pipelines to their work on mixture-of-experts (MoE) model training.

aside_block: <ListValue: [StructValue([('title', 'Try Google Cloud for free'), ('body', <wagtail.rich_text.RichText object at 0x7f2e2431f250>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Kakao’s journey by Minho and Nayeon:

As engineers at Kakao, we develop models that serve KakaoTalk, a platform supporting services that extend far beyond text. Our rich ecosystem includes chat with over 700,000 images and stickers (emojis), voice and video calls, finance, and navigation.

KakaoTalk’s massive scale and complexity demand that our language models are not only highly efficient but also excel at understanding the Korean language and are flexible enough for diverse applications. These real-world product requirements directly influenced our technical decisions and our need for a customizable training framework.

Our journey with JAX began at an important inflection point. Our existing GPU-based infrastructure was reaching power and budget capacity constraints. We had two options: expand our GPU infrastructure and maintain our existing codebase, or adopt Cloud TPUs, which offered cost-performance advantages while requiring adoption of a new toolchain. We chose Cloud TPUs, viewing the short-term investment as worthwhile for long-term cost-performance benefits, and built our stack on JAX.

We use XPK for Kubernetes cluster management, which simplifies job creation and management on GKE without requiring Kubernetes expertise. For the data pipeline, we adopted Grain due to its deterministic behavior, which is essential for the stability of long-running AI model training jobs.

We focused on adapting the MaxText framework to fit our specific research and compatibility needs. We made two key customizations to the pipeline:

1. Multi-source data blending: When we began exploring training with MaxText, it assumed a single, pre-mixed corpus. Our research requires blending different data sources — such as web text, code, and math — with specific, dynamically-adjusted weights during different training phases. To achieve this flexibility without reprocessing terabytes of data for each experiment, we implemented a solution using Grain's mix function. This approach allows us to define blending ratios in our configuration, providing the adaptability essential for our iterative research process. We filed a PR for this feature to be supported in MaxText natively, and it has been incorporated here since.

2. Token Processing for Efficiency and Compatibility: To maintain compatibility with our existing Megatron-LM pipeline and improve efficiency, we modified MaxText's token processing logic. Our data preparation method constructs each training sequence by appending the first token of the subsequent sequence. This creates overlapping, continuous sequences, ensuring that no information is lost at the boundaries and maximizing data utilization.

To validate our new TPU-based workflow, we trained two models. First, we trained the Kanana 2.1 billion parameter model from scratch, and the results demonstrated that our MaxText implementation achieved performance comparable to our existing GPU-based Megatron-LM pipeline at each stage. Second, we performed depth upscaling with continued pre-training from our existing 8B model to a 9.8B architecture. Both approaches succeeded and showed consistent improvements across various benchmarks, confirming that the results on GPU were effectively reproduced on TPU.

Advancing our approach: Training Mixture-of-Experts (MoE) models with MaxText

With the core pipeline validated, we began experimenting with more advanced architectures, specifically MoE models, to build inference-efficient models that maintain strong performance. Our objectives were to explore upcycling an existing dense model into an MoE structure and to evaluate the suitability of the TPU and MaxText stack for this task.

For the experiment, we upcycled our 2.1B dense model into a 13.4B parameter (2.3B active) MoE architecture with 64 experts and 8 active experts per token. We trained this model on the exact same dataset as the original dense model to isolate the impact of the architectural change. The training was performed on v5e TPUs using MaxText with Fully Sharded Data Parallelism (FSDP).

The implementation process was straightforward. We found that MaxText's flexible design, built on Flax, Optax, and Orbax, was well-suited for the wide range of ablations required for MoE research. Specifically:

Integrated Kernels: Megablocks MoE kernels which support optimized MoE features like Group GEMM were already integrated into JAX.
Combining Schedules: We used the optax.join_schedules function to combine multiple learning rate schedules (e.g. warmup, constant, and annealing) into a single, custom schedule for our training run. This ability to combine different schedules is very useful to experiment with different training strategies.

Code Customization: We needed to enable the load balancing loss for our sparse matmul implementation. This required inserting a single line of code in the permute function within the MoE block of MaxText to calculate the loss directly from the router logits.

The results showed performance improvements, particularly in code and math benchmarks, suggesting domain specialization among the experts.

Performance Evaluation

This met our objectives and further demonstrated the JAX stack's utility for advanced model development. We are now extending this work by experimenting with shared experts and replacing initial MoE layers with dense layers, modifications which are simple to implement within the MaxText framework.

Performance improvements and key takeaways

During our work, we gained early access to Trillium TPUs. We managed the transition from v5e by changing a few parameters in our XPK cluster and workload configurations. We observed an immediate and substantial throughput increase of 2.7x across our models, along with improved cost-performance efficiency.

Based on our experience, the JAX stack on TPUs provides a comprehensive and efficient environment for AI model development. The key advantages for our team include:

Performance and scalability: The JAX and XLA combination provides just-in-time compilation, and MaxText is optimized for large-scale parallel computing with support for paradigms like SPMD and FSDP.
Customizability and control: The codebase, being pure Python and built on libraries like Flax, Optax, and Orbax is intuitive and easy to modify. This allows us to implement custom data pipelines, training strategies, and novel architectures with minimal overhead.
Rapid feature adoption: The MaxText framework is updated quickly with features from new state-of-the-art models, allowing us to stay current with our research.

These strengths have made the JAX stack a powerful and flexible foundation for our work in training large language models at Kakao.

Build your Language Models with the JAX Ecosystem:

Kakao's journey demonstrates how the JAX ecosystem’s modular design — including MaxText, Flax, Optax, and Orbax — enables the customization required for both production pipelines and advanced research, from tailored data blending to rapid experimentation with MoE architectures.

Our sincere thanks to Minho, Nayeon and their team for sharing their insightful engineering work. We look forward to seeing how they and other leading enterprises worldwide continue to use the JAX ecosystem to build the next generation of powerful and efficient language models.

How Yahoo Calendar broke free from hardware queues and DBA bottlenecks

Mon, 11 Aug 2025 16:00:00 +0000

Editor's note: Yahoo Mail is in the midst of one of its largest infrastructure transformations to date: a multi-year effort to modernize hundreds of petabytes of services by moving to Google Cloud.The Yahoo Mail migration - a high-scale always-on workload - began with Yahoo Calendar, a product that is an essential part of the experience for hundreds of millions of Yahoo Mail users. It was a massive undertaking with no room for error, and the result was a smooth cutover with no customer impact that proved Cloud SQL could handle the complexity and pace of Yahoo’s operations. It also marked a shift in how Yahoo works by reducing manual overhead, unlocking developer agility, and laying the foundation for what comes next.

At Yahoo, we knew migrating our cornerstone platform — Yahoo Mail — to the cloud would be one of the most significant infrastructure efforts we’d ever taken on. With over 500 petabytes of interconnected systems, we knew we needed to start with a smaller, high-impact workload to build early confidence. That’s how Yahoo Calendar, a product that is an essential part of the experience for hundreds of millions of Yahoo Mail users, , became the first production service to make the move.

We needed to migrate a high-scale, always-on service without disrupting the experience users rely on every day — or risk millions of people missing standups, birthday dinners, or that dentist appointment they actually remembered to schedule.

We chose Google Cloud to help us modernize our operations with managed infrastructure, reduce manual effort, and tap into a trusted ecosystem for large-scale transformation. Migrating Yahoo Calendar became our proving ground for running mission-critical services on Cloud SQL and would set the pace for the rest of our multi-year migration plan for Yahoo Mail.

aside_block: <ListValue: [StructValue([('title', 'Build smarter with Google Cloud databases!'), ('body', <wagtail.rich_text.RichText object at 0x7f2e2706eb80>), ('btn_text', ''), ('href', ''), ('image', None)])]>

Modernizing infrastructure without skipping a single invite

The infrastructure we were replacing included tens of on-premises MySQL (Percona) instances. It was solid but not built for operational speed. Scaling meant filing hardware requests and often waiting weeks or even months. Routine tasks like backups or upgrades had to go through a separate database administration (DBA) team. And as demand grew, the need for agility grew with it. To meet that growing need with more flexibility and speed, we took on a massive lift:

Migrating tens of database shards across multiple regions
Moving over 20+ TB of storage (excluding replicas)
Supporting peak traffic of 1 million QPS reads and 2,500 QPS writes
Replatforming our application stack to run on Google Kubernetes Engine (GKE) to support the Calendar experience for the hundreds of millions of Yahoo Mail users

Cloud SQL's support for our existing MySQL workloads with minimal changes lets us replicate our on-prem shards without a full re-architecture. That compatibility provided the foundation to restructure our full stack. To make it all work, we migrated the UI, API, and backend to GKE and connected everything to Cloud SQL deployments in multiple Google Cloud regions. All of this had to be migrated incrementally, with no downtime for public users. Traffic continued flowing through existing endpoints, and our proxy layer routed requests based on each user’s location and migration state. As database shards became ready, we carefully flipped them into read-write mode on Cloud SQL to keep Calendar users running on schedule while shifting the backend in stages.

Fig. 1 - MySQL On-Prem to Cloud SQL Initial Load + CDC

A migration this big needed backup

Google Cloud’s Professional Services Organization (PSO) played a critical role in getting us there. From the earliest stages, they were embedded with our team. They helped us evaluate Cloud SQL and Database Migration Service (DMS), guide proof-of-concept work, and stress-test our migration architecture.

When we hit a roadblock replicating data with DMS, PSO worked closely with Cloud SQL engineering and our internal security and DBA teams to design a custom workaround. During cutover, they were right there with us to help with hiccups like debugging capacity constraints or troubleshooting connection spikes during shadow traffic. They also helped us resolve reverse replication failures caused by permission changes — an edge case we wouldn’t have anticipated without their guidance.

Fig. 2 - Yahoo Calendar migration diagram

Cloud SQL helped us block time for what matters

With managed infrastructure, we’ve significantly reduced manual operations, reduced database admin overhead, and gained the agility to scale up without the wait. Our application teams now deploy and manage database shards ourselves using infrastructure as code (IaC), without relying on manual processes. Backups, patching, and failovers are automated to reduce risk and manual effort. Usage and cost monitoring are built-in, helping us optimize across the board. And thanks to tight integration with our security protocols, we’re able to maintain high confidence in operating a large-scale public-facing service.

Today, Yahoo Calendar processes hundreds of thousands of queries per second, operates 26 Cloud SQL instances with disaster recovery (DR), and runs on infrastructure that includes 2,500 virtual CPUs and 17 TB of memory for databases alone. Our application tier spans 850 pods and 2,200 vCPUs, with 10 TB of memory to match. We now run at scale, with confidence — and without waiting on hardware or handoffs.

Fig .3 - Architecture diagram of Yahoo Calendar’s services

Up next on our calendar

We’re seeing the benefits of infrastructure that works with us, not against us. And we’re doing it all without compromising on scale, performance, or security. Now that we've pressure-tested our migration strategy and refined how we operate in the cloud, we're ready to take on Yahoo Mail’s full environment — 500 petabytes and counting.

The next couple of years will be about scaling smart, staying nimble, and proving that modernization doesn’t have to mean disruption. But with the hardest part of any journey behind us (starting), and a calendar that runs on Cloud SQL, we’re in sync and right on schedule.

Learn more:

Discover how Cloud SQL can transform your business! Start a free trial today!
Download this IDC report to learn how migrating to Cloud SQL can lower costs, boost agility, and speed up deployments.
Learn how Ford and Lightricks gained high performance and cut costs by modernizing with Cloud SQL.

Google is a Leader in the Gartner® Magic Quadrant for Strategic Cloud Platform Services

Fri, 08 Aug 2025 16:00:00 +0000

For the eighth consecutive year, Gartner® has named Google a Leader in the Gartner Magic Quadrant™ for Strategic Cloud Platform Services. This year, however, marks a major milestone: Google is now positioned furthest for completeness of vision.

Download the complimentary 2025 Magic Quadrant for Strategic Cloud Platform Services.

How do we keep getting recognized year over year? In our opinion, and what our customers consistently relay, are three major advantages of working with Google Cloud infrastructure:

A purpose-built modern cloud, specifically designed for AI-first apps and services.
Workload-optimized compute, storage, and the most versatile cloud networking that make it easier to migrate and modernize key enterprise workloads — SAP, VMware, Microsoft, Oracle, OpenShift, and even mainframes.
Leading reliability and security among hyperscalers — customers see Google as an extension of their security team.

But that’s not the full story. What does Google Cloud do behind the scenes to provide these advantages to our customers? Let’s dive into the five principles that Google Cloud lives by as we continue to build and enhance our infrastructure:

1. Run workloads on precisely optimized infrastructure

At Google Cloud, we believe that simply adding more hardware is an unsustainable and ineffective approach to scaling application performance. That’s why we develop strategic, workload-optimized infrastructure technologies. For example, Titanium is a system of hardware and software offloads, boosting performance, reliability, and security and maximizing workload efficiency.

In the past 12-months we also enhanced our Compute Engine platform to drive enterprise transformation with more massive developments:

Fourth-generation compute instances and Hyperdisk block storage, focusing on optimizing performance and costs across workloads, while delivering enterprise-grade scalability, reliability, security, and workload consistency, ultimately enabling businesses to grow efficiently and invest more in innovation.
We also developed our own custom Arm®-based CPUs, Google Axion processors, aimed at maximizing performance, reducing infrastructure costs, and supporting sustainability goals for general-purpose workloads.
Continued enhancements to our fully managed database services, such as AlloyDB for PostgreSQL and Cloud SQL, and easy migration from traditional databases to Google Cloud infrastructure.

2. Build AI-centric infrastructure

Google, like in 2024, is again ranked #1 for AI/ML in the 2025 Gartner Critical Capabilities for Strategic Cloud Platform Services. And we’ve only added to our depth and breadth since then. For evidence of our commitment to AI, there’s AI Hypercomputer, built on Google AI technology developed over the past decade, underpinning nearly every AI workload run on Google Cloud.

This integrated supercomputing system delivers more intelligence at a consistently low price for both training and serving AI workloads. It powers Vertex AI, our managed machine learning (ML) platform, that unifies the entire ML lifecycle—from data preparation to model deployment and monitoring. From an infrastructure perspective, it allows data scientists and developers to focus on building models rather than provisioning and managing the infrastructure required to run them.

Beyond core infrastructure, we’re also dedicated to fostering an open and innovative generative AI partner ecosystem, helping companies rapidly transform their software development, business processes, and information retrieval.

We’ve also released a plethora of new AI products and services, including:

Ironwood, announced at Next ‘25, is our 7th generation TPU, built specifically for inference, offering 5x more peak compute capacity and 6x more high-bandwidth memory than its predecessor and achieving a staggering 42.5 exaFLOPS of compute in its larger configuration while being 2x more power efficient.
Google Cloud Managed Lustre, a fully managed parallel file system built on the DDN EXAScaler Lustre file system, which provides PB scale at under 1ms latency, millions of IOPS, and TB/s of throughput for AI workloads; Rapid Storage, a new Cloud Storage zonal bucket that provides industry-leading <1ms random read and write latency, 20x faster data access, 6 TB/s of throughput, and 5x lower latency for random reads and writes compared to other leading hyperscalers; Anywhere Cache, which provides an SSD-backed zonal read cache for Cloud Storage buckets; Hyperdisk Exapools, entering preview this quarter, providing exabytes of block storage capacity and TB/s of throughput for AI clusters.
Gemini CLI (Command Line Interface), an open-source AI agent that brings Gemini 2.5 Pro directly into the terminal with unmatched free usage limits (60 model requests per minute, 1,000 requests per day) and integrates with Gemini Code Assist for AI-first coding and problem-solving.

3. Meet the AI moment with containers

With Google Kubernetes Engine (GKE), Google is once again a Leader in the Gartner Magic Quadrant for Container Management this year. And like our advancements elsewhere, we feel it's only gotten better. Built on years of experience and a deep commitment to Kubernetes, GKE has emerged as the foundation for the next generation of AI workloads. In fact, we use GKE to power our own leading AI services at scale, including Vertex AI, leveraging the same cutting-edge technologies and best practices we share with you.

To continue empowering your AI innovation, we've rolled out several impressive new releases:

GKE Autopilot, with an impressive 30% of active GKE clusters created in 2024 operating in Autopilot mode, showcases its effectiveness in simplifying operations and enhancing resource efficiency for critical workloads.
Cluster Director, now GA, allows for the deployment and management of massive clusters of accelerators (GPUs and TPUs) as a single, high-performance unit, crucial for demanding AI models and distributed workloads.
For optimizing AI inference, the new GKE Inference Quickstart and GKE Inference Gateway (both in public preview) simplify model deployment with benchmarked profiles and provide intelligent routing, leading to up to a 30% reduction in serving costs and up to a 60% decrease in tail latency.

4. Empower customers with true sovereignty

At Google Cloud, we believe that digital sovereignty is about more than just compliance; it's about providing organizations with flexibility, control, choice, and robust security without compromising functionality to enable innovation. We do that by delivering technology capabilities that align with our customers' diverse needs, backed by extensive engagement with local partners and policymakers.

Our strength in this area comes from our massively scaled global infrastructure, encompassing over 42 cloud regions, 127 zones, and 202 network edge locations, alongside significant subsea cable investments. Two of our most important recent developments here include:

Forging key partnerships with independent local and regional partners across Asia, Europe, the Middle East, and the United States, such as Schwarz Group, T-Systems, S3NS (with Thales), Minsait, Telecom Italia, and World Wide Technology
Designing a portfolio of sovereign solutions designed to fit your unique business needs, regulatory requirements, and risk profiles. With three precisely designed solutions, Google Cloud Data Boundary, Google Cloud Dedicated, and Google Cloud Air-Gapped, organizations can choose what you need based on their unique requirements.

5. Deliver a “planet-scale” network

Our strength in networking is built upon over 25 years of Google's foundational innovations, connecting billions of people globally to essential services like Gmail, YouTube, and Search. This deep expertise has allowed us to build planet-scale network infrastructure that powers Google Cloud and our Cross-Cloud Network solutions. Our vast backbone network features 202 points of presence (PoPs), over 2 million miles of fiber, and 33 subsea cables, all backed by a 99.99% reliability SLA, providing a robust and resilient global platform.

With the rapid emergence of AI, today’s industries are demanding unprecedented network capabilities for training, inference, and serving AI models, including massive capacity, seamless connectivity, and robust security. We're addressing this with continuous innovation in our cloud networking products and Cross-Cloud Network solutions, enabling customers to easily build and deliver distributed AI applications.

To meet these demanding requirements, we've launched a suite of impressive new networking solutions, including:

Cloud WAN, our new, fully managed, reliable, and secure enterprise backbone designed to transform wide area network (WAN) architectures by leveraging Google's planet-scale network. It offers up to 40% faster performance compared to the public internet and up to a 40% savings in total cost of ownership (TCO) over customer-managed solutions.
Cloud Interconnect and the new Cross-Site Interconnect (options in Cloud WAN, currently in preview), make Google Cloud the first major cloud provider to offer transparent Layer 2 connectivity over its network.
For AI-optimized networking, 400G Cloud Interconnect (4x more bandwidth for massive data ingestion), networking support for up to tens of thousands of GPUs per cluster, and Zero-Trust RDMA security for high-performance GPU/TPU traffic.

Take the next steps on your journey with Google Cloud

From optimizing legacy apps to building next-gen AI and everything in between, Google Cloud lets you tackle your biggest challenges with a platform that works across your data centers, other clouds, and the edge. With decades of experience and a proven track record of reliability, Google Cloud infrastructure is equipped to handle your most visionary workloads and is the ideal partner to drive your business transformation.

You can download a complimentary copy of the 2025 Magic Quadrant for Strategic Cloud Platform Services on our website.

Selecting the right Hyperdisk block storage for your workloads

Wed, 11 Jun 2025 16:00:00 +0000

As you adopt Google Cloud or migrate to the latest Compute Engine VMs or to Google Kubernetes Engine (GKE), selecting the right block storage for your workload is crucial. Hyperdisk, Google Cloud's workload-optimized block storage that’s designed for our latest VM families (C4, N4, M4, and more), delivers high-performance storage volumes that are cost-efficient, easily managed at scale, and enterprise-ready. In this post, we guide you through the basics and help you choose the optimal Hyperdisk for your environment.

Introduction to Hyperdisk block storage

With Hyperdisk, you can independently tune capacity and performance to match your block storage resources to your workload. Hyperdisk is available in a few flavors:

Hyperdisk Balanced: Designed to fit most workloads and offers the best combination and balance of price and performance. This is also the boot disk for your compute instances. With Hyperdisk Balanced, you can independently configure the capacity, throughput, and IOPS of each volume. Hyperdisk Balanced is available in High Availability and Multi-writer mode.
Hyperdisk Extreme: Delivers the highest IOPS of all Hyperdisk offerings and is suited for high-end, performance-critical databases. With Hyperdisk Extreme, you can drive up to 350K IOPS from a single volume.
Hyperdisk Throughput: Delivers capacity at the cost of cold object storage with the semantics of a disk. Hyperdisk Throughput offers high throughput for bandwidth and capacity-intensive workloads that do not require low latency. It also can be used to deliver cost-effective disks for cost-sensitive workloads (e.g., cold disks).
Hyperdisk ML: Purpose-built for loading static data into your compute clusters. With Hyperdisk ML, you hydrate the disk with a fixed data set (such as model weights or binaries), then connect up to 2,500 compute instances to the same volume, so a single volume can serve over 150x more compute instances than competitive block storage volumes¹ in read-only mode. You get exceptionally high aggregate throughput across all of those nodes, enabling you to accelerate inference startup, train models faster, and ensure your valuable compute resources are highly utilized.

You can also leverage Hyperdisk Storage Pools, which lowers TCO and simplifies operations by pre-provisioning an aggregate amount of capacity and performance, which is then dynamically consumed by volumes in that pool. You create a storage pool with the aggregate capacity and performance that your workloads will need, and then create disks in the storage pool. You can then attach the disks to your VMs. When you create the disks, you can create them with a much larger size or provisioned performance limit than is needed. This simplifies planning and provides room for growth later, without needing to change the disk's provisioned size or performance.

You can also use a set of comprehensive data protection capabilities such as high availability, cross-region replication and recovery, backup, and snapshots to protect your business critical workloads.

For specifics around capabilities, capacity, machine support, and performance, please visit the documentation.

aside_block: <ListValue: [StructValue([('title', 'Try Google Cloud for free'), ('body', <wagtail.rich_text.RichText object at 0x7f2e294cdd00>), ('btn_text', 'Get started for free'), ('href', 'https://console.cloud.google.com/freetrial?redirectPath=/welcome'), ('image', None)])]>

Recommendations for the most common workloads

To make choosing the right Hyperdisk architecture simpler, here are high-level recommendations for some of the most common workloads we see. For an enterprise, the Hyperdisk portfolio lets you optimize an entire three-tier application matching the needs of each component of your application to the different flavors of Hyperdisk.

Enterprise applications including general-purpose databases:

Hyperdisk Balanced combined with Storage Pools offers an excellent solution for a wide variety of general-purpose workloads, including common database workloads. Hyperdisk Balanced can meet the IOPS and throughput needs for most databases including Clickhouse, MySQL, and PostgreSQL, at general-purpose pricing. Hyperdisk Balanced offers 160K IOPS per volume — better than AWS EBS gp3 volumes². With Storage Pools you can enhance efficiency and radically simplify planning. Storage Pools allows customers to save approximately 20-40% on storage costs for typical database workloads when compared to Hyperdisk Balanced Volumes or AWS EBS gp3 volumes³.

“At Sentry.io, a platform used by over 4 million developers and 130,000 teams worldwide to quickly debug and resolve issues, adopting Google Cloud's Hyperdisk has enabled us to create a flexible architecture for the next-generation of our Event Analytics Platform, a product at the core of our business. Hyperdisk Storage Pools with advanced capacity and performance enabled us to reduce our planning cycles from weeks to minutes, while saving 37% in storage costs, compared to persistent disks.” - Dave Rosenthal, CTO, Sentry

“High Availability is essential for Blackline — we run database failover clustering, at massive scale, for our global and mission-critical deployment of Financial Close Management. We are excited to bring this workload to Google Cloud leveraging Hyperdisk Balanced High Availability to meet the performance, capacity, cost efficiency, and resilience requirements that our customers demand, and helps us address our customer’s financial regulatory needs globally.” - Justin Brodley, SVP of Cloud Engineering and Operations, Blackline

Tier-0 databases

For high-end, performance-critical databases like SAP HANA, SQL Server, and Oracle Database, Hyperdisk Extreme delivers uncompromising performance. With Hyperdisk Extreme, you can obtain up to 350K IOPS and 10 GiB/s of throughput from a single volume.

AI, analytics, and scale-out workloads

Hyperdisk offers excellent solutions for the most demanding next-generation machine learning and high performance computing workloads.

Dynamically scaling AI and analytics workloads and high-performance file systems

Workloads with fluctuating demand, and high peak throughput and IOPS, benefit from Hyperdisk Balanced and Storage Pools. These workloads can include customer-managed parallel file systems and scratch disks for accelerator clusters. Storage Pools’ dynamic resource allocation helps ensure that these workloads get the performance they need during peak times without requiring constant manual adjustments or inefficient over-provisioning. Further, once your Storage Pool is set up, planning at the per-disk level is significantly simpler. Note: If you want a fully managed file system, Managed Lustre is an excellent option for you to consider.

“Combining our use of cutting-edge machine learning in quantitative trading at Hudson River Trading (HRT) with Google Cloud's accelerator-optimized machines, Dynamic Workload Scheduler (DWS) and Hyperdisk has been transformative in enabling us to develop [state-of-the-art] models. Hyperdisk storage pools have delivered substantial cost savings, lowering our storage expenses by approximately 50% compared to standard Hyperdisk while minimizing the amount of planning needed.” - Ragnar Kjørstad, Systems Engineer, Hudson River Trading

AI/ML and HPC data-load acceleration

Hyperdisk ML is specifically optimized for accelerating data load times for inference, training and HPC workloads — Hyperdisk ML accelerates model load time by 3-5x compared to common alternatives⁴. Hyperdisk ML is particularly well-suited for serving tasks compared to other storage services on Google Cloud because it can concurrently provide to many VMs exceptionally high aggregate throughput (up to 1.2 TiB/s of aggregate throughput per volume, offering greater than 100x higher performance than competitive offerings)⁵. You write once (up to 64 TiB per disk) and attach multiple VM instances to the same volume in a read-only mode. With Hyperdisk ML you can accelerate data load times for your most expensive compute resources, like GPUs and TPUs. For more, check out g.co/cloud/storage-design-ai.

“At Resemble AI, we leverage our proprietary deep-learning models to generate high-quality AI audio through text-to-speech and speech-to-speech synthesis. By combining Google Cloud’s A3 VMs with NVIDIA H100 GPUs and Hyperdisk ML, we’ve achieved significant improvements in our training workflows. Hyperdisk ML has drastically improved our data loader performance, enabling 2x faster epoch cycles compared to similar solutions. This acceleration has empowered our engineering team to experiment more freely, train at scale, and accelerate the path from prototype to production." - Zohaib Ahmed, CEO, Resemble AI

High-capacity analytics workloads:

For large-scale data analytics workloads like Hadoop and Kafka, which are less sensitive to disk latency fluctuations, Hyperdisk Throughput provides a cost-effective solution with high throughput. Its low cost per GiB and configurable throughput are ideal for processing large volumes of data with low TCO.

How to size and set up your Hyperdisk

To select and size the right Hyperdisk volume types for your workload, answer a few key questions:

Storage management. Decide if you want to manage the block storage for your workloads in a pool or individually. If your workload will have more than 10 TiB of capacity in a single project and zone, you should consider using Hyperdisk Storage Pools to lower your TCO and simplify planning. Note that Storage Pools do not affect disk performance; some data protection features such as Replication and High Availability are not supported in Storage Pools.
Latency. If your workload requires SSD-like latency (i.e., sub-millisecond), it likely should be served by Hyperdisk Balanced or Hyperdisk Extreme.
IOPS or throughput. If your application requires less than 160K IOPS or 2.4 GiB/s of throughput from a single volume, Hyperdisk Balanced is a great fit. If it needs more than that, consider Hyperdisk Extreme.
Sizing performance and capacity. Hyperdisk offers independently configurable capacity and performance, allowing you to pay for just the resources you need. You can leverage this capability to lower your TCO by understanding how much capacity your workload needs (i.e., how much data, in GiB or TiB, is stored on the disks which serve this workload) and the peak IOPS and throughput of the disks. If the workload is already running on Google Cloud, you can see many of these metrics in your console under “Metrics Explorer.”

Another important consideration is the level of business continuity and data protection required for your workloads. Different workloads have different Recovery Point Objective (RPO) and Recovery Time Objective (RTO) requirements, each with different costs. Think about your workload tiers when making data-protection decisions. The more critical an application or workload, the lower the tolerance for data loss and downtime. Applications critical to business operations likely require zero RPO and RTO in the order of seconds. Hyperdisk business continuity and data protection helps customers meet the performance, capacity, cost efficiency, and resilience requirements they demand, and helps them address their financial regulatory needs globally.

Here are a few questions to consider when selecting which variety of Hyperdisk to use for a workload:

How do I protect my workloads from attack and malicious insiders? Use Google Cloud Backup vault for cyber resilience, backup immutability, and indelibility for managed backup reporting and compliance. If you want to self-manage your own backups, Hyperdisk standard snapshots are an option for your workloads.
How do I protect data from user errors and bad upgrades cost efficiently with low RPO / RTO? You can use our point-in-time recovery with Instant Snapshots. This feature minimizes the risk of data loss from user error and bad upgrades with ultra-low RPO and RTO — creating a checkpoint is nearly instantaneous.
How do I easily deploy my critical workload (e.g., MySQL) with resilience across multiple locations? You can utilize Hyperdisk HA. This is a great fit for scenarios that require high availability and fast failover, such as SQL Server that leverages failover clustering. For such workloads, you can also choose our new capability with Hyperdisk Balanced High Availability with Multi-Writer support. This allows you to run clustered compute with workload-optimized storage in two zones with RPO=0 synchronous replication.
When a disaster occurs, how do I recover my workload elsewhere quickly and reliably, and run drills to confirm my recovery process? Utilize our disaster recovery capabilities with Hyperdisk Async Replication which enables cross-region continuous replication and recovery from a regional failure, with fast validation support for disaster recovery drills via cloning. Further, consistency group policies help ensure that workload data that’s distributed across multiple disks is recoverable when a workload needs to fail over between regions.

In short, Hyperdisk provides a wealth of options to help you optimize your block storage to the needs of your workloads. Further, selecting the right Hyperdisk and leveraging features such as Storage Pools can help you lower your TCO and simplify management. To learn more, please visit our website. For tailored recommendations, always consult your Google Cloud account team.

_{1. As of March 2025 based on published information for Amazon EBS, Azure managed disks.
2. As of May 2025, compared to Amazon EBS gp3 volumes max iops/volume
3. As of March 2025, at list price, 50 to 150 TiB, peak IOPS of 25K to 75K and 25% compressibility, compared to Amazon EBS gp3 volumes.
4. As of March 2025, based on internal Google benchmarking, compared to Rapid Storage, GCSFuse with Anywhere Cache, Parallelstore and Lustre for larger node sizes.
5. As of March 2025 based on published performance for Microsoft Azure Ultra SSD and Amazon EBS io2 BlockExpress}

^{The authors would like to thank David Seidman and Ruwen Hess for their contributions on this blog.}

Advancing sovereignty, choice, and security in the cloud for our customers

Wed, 21 May 2025 08:00:00 +0000

Like most organizations, Google Cloud is continually engaging with customers, partners, and policymakers to deliver technology capabilities that reflect their needs. When it comes to digital sovereignty solutions, Google Cloud has worked with customers for nearly a decade.

Today, we’re pleased to announce significant technical and commercial updates on our sovereign cloud solutions for customers, and details on how we’re helping them achieve greater control, choice, and security in the cloud — without compromising functionality.

Building on the first sovereign solutions we introduced years ago, we’ve massively scaled our infrastructure footprint globally, now consisting of more than 42 cloud regions, 127 zones, 202 network edge locations, and 33 subsea cable investments.

We have also forged key partnerships in Asia, Europe, the Middle East, and the United States to help deliver these sovereign solutions, including Schwarz Group and T-Systems (Germany), S3NS (France), Minsait (Spain), Telecom Italia (Italy), Clarence (Belgium and Luxembourg), CNTXT (Saudi Arabia), KDDI (Japan), and World Wide Technology (United States).

A commitment to customer choice

Digital sovereignty is about more than just controlling encryption keys. At its core, it’s about giving customers the flexibility their global businesses require. It’s about enabling them to operate on multiple clouds. And it’s about securing data with the most advanced technologies.

We’ve long been committed to enabling customers to choose the cloud provider and solution that best fit their needs, and not locking them into a single option. Sovereignty in the cloud is not one-size-fits-all. We offer customers a portfolio of solutions that align with their business needs, regulatory requirements, and risk profiles.

Our strong contractual commitments to our customers are backed by robust sovereign controls and solutions that are all available today. Our updated sovereign cloud solution portfolio includes:

Google Cloud Data Boundary gives customers the ability to deploy a sovereign data boundary and control where their content is stored and processed. This boundary also allows customers to store and manage their encryption keys outside Google’s infrastructure, which can help customers meet their specific data access and control requirements no matter what market.

Google Cloud Data Boundary customers have access to a large set of Google Cloud products, including AI services, and can enable capabilities, including Confidential Computing and External Key Management with Key Access Justifications to control access to their data and deny access for any reason.

In addition, Google Workspace customers can take advantage of Google Cloud Data Boundary’s sovereign controls to limit the processing of their content to the United States or EU, choose a country to locally store data, and use client-side encryption to prevent unauthorized access (even by Google) to their most critical content.

Today, we are also announcing User Data Shield, a solution that adds Mandiant services to validate the security of customer applications built on top of Google Cloud Data Boundary. User Data Shield provides recurring security testing of customer applications to validate sovereignty postures.

Google Cloud Dedicated delivers a solution designed to meet local sovereignty requirements, enabled by independent local and regional partners. As an example, Google Cloud has partnered with Thales since 2021 to build a first-of-its-kind Trusted Cloud by S3NS for Europe.

This offering with Thales is designed to offer a rich set of Google Cloud services with GPUs to support AI workloads and is operated by S3NS, a standalone French entity. Currently in preview, S3NS’ solution is designed to meet the rigorous security and operational resilience requirements of France’s SecNumCloud standards. We are expanding our Google Cloud Dedicated footprint globally, launching next in Germany.

“For France to truly embrace digital sovereignty, it is essential to have a cloud solution that marries the immense power of hyperscale technology with the strictest local security and operational controls. S3NS is committed to providing French organizations with access to advanced cloud services, including critical AI capabilities, all operated within France by a European operator to meet and exceed the rigorous SecNumCloud standards,” said Christophe Salomon, EVP, Information Systems and Secured Communication, at Thales.

Google Cloud Air-Gapped offers a fully standalone and air-gapped solution that does not require connectivity to an external network. This solution is tailored for customers in the intelligence, defense, and other sectors with strict data security and residency requirements. The air-gapped solution can be deployed and operated by Google, the customer, or a Google partner.

It is built with open-source components and comes with a targeted set of AI, database, and infrastructure services. Because air-gapped solutions run on open-source components, they are designed to provide business continuity and survivability in the event of service disruptions. Google Cloud Air-Gapped received authorization in 2024 to host U.S. government Top Secret and Secret-level data.

“Working with Google Cloud to introduce sovereign offerings can give our joint clients greater control, choice, and security in the cloud, without compromising the functionality of their underlying cloud architectures,” said Scott Alfieri, Senior Managing Director and Google Business Group Lead at Accenture. “Google Cloud's extensive global infrastructure, coupled with Accenture’s transformation and industry expertise, helps organizations build an agile and scalable foundation, unlocking opportunities for growth and continuous innovation.”

Local control, global security

Security and sovereignty are two sides of the same coin. Local control of data and operations can provide customers a greater level of confidence in their security, but it’s also true that no organization can be considered sovereign if dependencies on legacy infrastructure leave its data vulnerable to loss or theft.

Analysis from the Google Threat Intelligence Group and Google Cloud’s Office of the CISO suggests that the global cyber threat landscape will only become more complex as malicious actors tap into AI-powered tools and techniques to prey on older software products, platforms, and outdated infrastructures.

With Google Cloud, customers not only get sovereign solutions, but also gain access to our leading security capabilities. This includes our rigorous focus on secure by design technology and deep expertise from Google Threat Intelligence Group and Mandiant Consulting, who operate on the frontlines of cyber conflicts worldwide and maintain trusted partnerships with more than 80 governments around the world.

In addition, Google Cloud CyberShield provides AI and intelligence-driven cyber defense to help governments defend against threats at national scale. And Mandiant Managed Defense services make it easy for customers worldwide to extend their security teams with our security team.

Google Sovereign Cloud solutions ultimately enable customers to leverage the secure foundation of Google Cloud, while gaining access to advanced security features — such as Confidential Computing, Zero Trust, post-quantum cryptography, and AI-powered platform defenses — faster and more cost-effectively than they could achieve on their own.

Sovereign solutions for any organization

We remain dedicated to fostering an environment of trust and control for our customers, empowering organizations globally to navigate the complex landscape of digital sovereignty with confidence. We continue to work with customers, partners, and policymakers around the world to refine our sovereign cloud offerings and deliver technologies that address their needs.

To learn more about how we are enabling our customers’ digital sovereignty capabilities, visit our web page or contact your account manager.

SandboxAQ: Accelerating drug discovery through cloud integration

Tue, 29 Apr 2025 15:00:00 +0000

The traditional drug discovery process involves massive capital investments, prolonged timelines, and is plagued with daunting failure rates. From initial research to obtaining regulatory approval, bringing a new drug to market can take decades. During this time, many drug candidates that had seemed very promising fail to deliver, either due to inefficacy or safety concerns. Only a small fraction of candidates successfully make it through clinical trials and regulatory hurdles.

Enter SandboxAQ, which is helping researchers explore vast chemical spaces, gain deep insights into molecular interactions, and predict biological outcomes with precision. It does so with cutting-edge computational approaches such as active learning, absolute free energy perturbation solution (AQFEP), generative AI, structural analysis, and predictive data analytics, ultimately reducing drug discovery and development timelines. And it does all this on a cloud-native foundation.

Drug design involves an iterative cycle of designing, synthesizing, and testing molecules referred to as the Design-Make-Test cycle. Many customers approach SandboxAQ during the design phase, often when their computational methods are falling short. By improving and accelerating this part of the cycle, SandboxAQ helps medicinal chemists bring innovative and effective molecules to market. For example, in a project related to neurodegenerative disease, SandboxAQ’s approach expanded chemical space from 250,000 to 5.6 million molecules, achieving a 30-fold increase in hit rate and dramatically accelerating the discovery of candidate molecules.

Cloud-native development for scientific insight

SandboxAQ’s software relies on large-scale computation and to maximize flexibility and scale, they use a cloud strategy, which includes Google Cloud infrastructure and tools.

The technologies in large-scale virtual screening campaigns need to be agile and scale cost-effectively. Specifically, SandboxAQ engineers need to be able to quickly iterate on scientific code, immediately run that code at scale cost-effectively, and store and organize all of the data it produces.

SandboxAQ achieved a significant boost in efficiency and scalability with Google Cloud infrastructure. They scaled their computational throughput by 100X to leverage tens of thousands of virtual machines (VMs) in parallel. They also improved utilization by reducing idle time by 90%. By consolidating development and deployment on Google Cloud, SandboxAQ streamlined its workflows, from code development and testing to large-scale batch processing and machine-learning model training.

aside_block: <ListValue: [StructValue([('title', 'Try Google Cloud for free'), ('body', <wagtail.rich_text.RichText object at 0x7f2e271a8b20>), ('btn_text', 'Get started for free'), ('href', 'https://console.cloud.google.com/freetrial?redirectPath=/welcome'), ('image', None)])]>

All of SandboxAQ’s development and deployment takes place in the cloud. Code and data live in cloud-based services, and development is done on a cloud-based platform that provides scientists and engineers with self-service VMs with standardized and centrally maintained environments and tools. This is important, because scientific code often requires heavy-duty computing hardware. Scientists have access to hefty 96-core machines, or instances with large GPUs. They can also create new machines with alternate configurations or CPU types as depicted below, enabling low-friction testing and development processes across heterogeneous resources.

SandboxAQ scientists and developers manage and access their Bench machines (see above) using the company’s `bench` client. They can connect to machines via SSH or use any number of managed tools, for example a browser-based VNC service for instant remote desktop, or JupyterLab for a familiar notebook development flow.

As code is ready to be run at a larger scale, researchers can dispatch SandboxAQ parameterized sets of computations as jobs on an internal tool powered by Batch, a fully managed service to schedule, queue, and execute batch jobs on Google infrastructure. With development and batch runtime environments closely synced, changes can be quickly run at scale. Code developed on bench machines is pushed to GitHub and immediately available for batch execution. Then, as tools are reviewed and merged into `main` of the company’s monorepo, the new tools become automatically available on SandboxAQ scientists’ bench machines, who can launch parallel jobs processing millions of molecules on any kind of Google Cloud VM resource in any global zone, utilizing either on-demand or Spot VMs.

SandboxAQ's implementation of a globally resolved transitive dependency tree, enables simple package and dependency management. With this practice, Google Batch can seamlessly integrate with individual tools developed by engineers to train many instances of a model in parallel.

Machine learning is a core component of SandoxAQ’s strategy, making easy data access especially important. At the same time, SandboxAQ’s Drug Discovery team also works with clients who have sensitive data. To secure customers’ data, bench and batch workloads read and write data from a unified interface that’s managed via IAM, allowing granular control of different data sources within the organization.

Meanwhile, Google Cloud services like Cloud Logging, Cloud Monitoring, Compute Engine and Cloud Run make it simple to develop tools to monitor these workloads, easily surface logs to SandboxAQ scientists, and comb through huge amounts of output data. As new features are tested or bugs show up, changes are made immediately available to the scientific team, without having to wrangle infrastructure. Then, as code becomes stable, they can incorporate it into downstream production applications, all in a centrally secured, unified way on Google Cloud.

In short, having a unified development, batch compute, and production environment on Google Cloud reduces the friction SandboxAQ faces to develop new workloads and run them at scale. With shared environments for scientific workload development and engineering, SandboxAQ makes it quick and easy for customers to move from experimentation to production, delivering the results customers want, fast.

SandboxAQ solution in the real world

SandboxAQ is already having a profound impact on drug discovery programs targeting a range of hard-to-treat diseases. For example, there are advanced collaborations with Professor Stanley Pruisner's lab at University of California San Francisco (UCSF), Riboscience, Sanofi, and with the Michael J Fox Foundation, to name a few. With this approach built on Google CloudSandboxAQ has achieved a superior hit rate compared to other methods like high throughput screening, demonstrating the transformative potential of SandboxAQ on drug discovery and bringing cures to patients faster.

Visit the Google Cloud AI Hypercomputer web page to learn about Google Cloud AI infrastructure.

Google Cloud and Oracle accelerate enterprise modernization with new offerings, regions, and capabilities

Thu, 10 Apr 2025 16:00:00 +0000

Supporting customers where they want to be is a core value at Google Cloud, and a big part of the reason that we have partnered with Oracle — so that you can innovate faster with the best of Google and the best of Oracle.

This week at Google Cloud Next, we announced significant expansions to our Oracle Database offerings, including the preview of Oracle Base Database Service for a flexible and controllable way to run Oracle databases in the cloud; general availability of Oracle Exadata X11M, bringing the latest generation of the Oracle Exadata platform to Google Cloud; and additional enterprise-ready capabilities including customer managed encryption keys (CMEK).

We are continuing to invest in global infrastructure for Oracle, with a total of 20 locations available in the coming months, adding Oracle Database@Google Cloud presence in Australia, Brazil, Canada, India, Italy, and Japan.

These announcements follow our developments with Oracle since last July, when we launched Oracle Database@Google Cloud. This partnership enables customers to migrate and modernize their Oracle workloads and start taking advantage of Google’s industry-leading data and AI capabilities such as BigQuery, Vertex AI platform, and Gemini foundation models.

Additional features provide customers with even more options in their modernization journey, such as the fully managed Oracle Autonomous Database Serverless. They can also benefit from increased reliability and resiliency features, such as cross-region disaster recovery and Oracle Maximum Availability Gold certification.

"Banco Actinver is committed to providing innovative financial solutions to our clients. By combining the security and performance of Oracle Database with Google Cloud's data analytics and AI tools, we're gaining deeper insights into market trends, enhancing our services, and delivering personalized experiences to our customers," said Jorge Fernandez, CIO, Banco Actinver.

aside_block: <ListValue: [StructValue([('title', '$300 in free credit to try Google Cloud databases'), ('body', <wagtail.rich_text.RichText object at 0x7f2e26525ca0>), ('btn_text', 'Start building for free'), ('href', 'http://console.cloud.google.com/freetrial?redirectPath=/products?#databases'), ('image', None)])]>

Innovative new capabilities

We're expanding our offerings to empower customers with the flexibility to manage a diverse set of database workloads cost effectively.

Oracle Base Database Service: The new Base Database Service delivers a highly controllable and customizable foundational database platform, built on Oracle Cloud Infrastructure (OCI) virtual machines and general-purpose infrastructure. It can empower businesses with the flexibility to manage a diverse range of database workloads directly.
Enhanced Oracle Database Services: In addition to the availability of Exadata Cloud Service, Autonomous Database Service, Oracle Linux, and Oracle on Google Compute Engine (GCE) and Google Kubernetes Engine (GKE), we are pleased to share general availability of Oracle Exadata X11M. Oracle Database@Google Cloud now offers the latest generation of Oracle Exadata machines, the X11M, with enhanced performance and scalability for demanding database workloads. These new machines provide significant performance gains and increased capacity, enabling customers to run even the most intensive Oracle applications with ease. X11M will be available in all new regions.

Customers are embracing Oracle Database@Google Cloud, and to support their global needs, we're expanding our footprint while maintaining the highest standards of application performance and reliability.

Expanding to 20 Oracle Database@Google Cloud Locations in the coming months: To further support the growing demand for Oracle workloads on Google Cloud, we are launching in more locations, including U.S. Central 1 (Iowa), North America-Northeast 1 (Montreal), North America-Northeast 2 (Toronto), Asia-Northeast 1 (Tokyo), Asia-Northeast 2 (Osaka), Asia-South 1 (Mumbai), Asia-South 2 (Delhi), South America-East 1 (Sao Paulo), Europe-West (Italy), Australia-Southeast2 (Melbourne), and Australia-Southeast1 (Sydney) — and additional zones in Ashburn, Frankfurt, London, Melbourne, and Italy. The new regions and expanded capacity are in addition to Google Cloud regions across U.S. East (Ashburn), U.S. West (Salt Lake City), U.K. South (London), and Germany Central (Frankfurt) that are available today.
New Partner Cross-Cloud Interconnect availability: Partner Cross-Cloud Interconnect for OCI is pleased to expand our global network offerings with new multicloud connectivity between Google and Oracle Cloud Infrastructure in Toronto and Zurich. This further complements our existing 11 regions already served, ensuring the lowest possible latency between both clouds while keeping traffic private and secure.
Cross Region Disaster Recovery: Cross Region Disaster Recovery support for Oracle workloads on Oracle Autonomous Database ensures high availability and resilience, protecting against potential outages and providing continuous operation for critical applications.
Enterprise-grade networking upgrades: Advanced networking upgrades enable enterprises to efficiently deploy their Oracle resources along with Google Cloud and share resources.

Industry-leading certifications and user experience

Google Cloud is committed to providing a seamless and efficient experience for Oracle customers, ensuring that managing and utilizing Oracle databases is straightforward and effective. We offer a combination of native Google Cloud tools and Oracle Cloud Infrastructure (OCI) interfaces, along with robust support for various applications and systems.

Enhanced user experience: Google Cloud is committed to providing an easy-to-use experience for Oracle customers, offering a Google Cloud integrated user experience for application developers and routine database operations, alongside an OCI-native experience for advanced database management. This includes support for Shared VPC, APIs, SDKs, and Terraform.
Application support: Google Cloud is pleased to announce the support for Oracle applications running on Google Cloud, ensuring compatibility and optimal performance, including Oracle E-Business Suite, Peoplesoft Enterprise, JD Edwards Enterprise One, Hyperion Financial Management, and Retail Merchandising.
SAP and Oracle Capability: Oracle workloads on Google Compute Engine are now supported by SAP and Oracle, further validating Google Cloud as a trusted platform for running enterprise applications.
Integration with Google Cloud Monitoring: Provides enterprises a unified monitoring and alerting mechanism across all their Google Cloud database services, now including Oracle Database.
New support in Google Cloud Backup and DR: Our backup service now provides central, policy-based management for backup of Oracle workloads along with other Google Cloud services using secure backup vaults for data protection — isolating and protecting data from threats like ransomware and accidental deletion.

Google Cloud's strengths make it the preferred hyperscaler for running mission-critical Oracle workloads.

Get started right away from your Google Cloud Console or learn more here.

Bringing Gemini and Google Agentspace to you on-premises

Wed, 09 Apr 2025 12:00:00 +0000

Today we are announcing that Gemini will be available on Google Distributed Cloud (GDC), bringing Google’s most capable models to on-premises environments, with public preview starting in Q3 2025. To do so, we’ve partnered with NVIDIA to bring our Gemini models to NVIDIA Blackwell systems that you can purchase through Google or your preferred channels.

GDC is a fully managed on-prem and edge cloud solution that is offered in both connected and air-gapped options, scaling from a single server to hundreds of racks. It offers infrastructure-as-a-service, security, data, and AI services, and is extensible with a rich ISV ecosystem. GDC takes care of infrastructure management, making it easy for your developers to focus on leveraging the best that AI has to offer and build applications, assistants, and agents.

“NVIDIA and Google Distributed Cloud provide a secure AI platform, bringing Gemini models to enterprise datacenters and regulated industries. With NVIDIA Blackwell infrastructure and confidential computing, Google Distributed Cloud enhances privacy and security, and delivers industry-leading performance on DGX B200 and HGX B200 systems, available from Dell.” – Justin Boitano, VP, Enterprise AI Software, NVIDIA.

Historically, organizations that face strict regulatory, sovereignty, latency, or data volume issues have been unable to access the latest AI technology since they must keep their data on-premises. Their only options have been open-source models and tools. And, in most cases, they have to put together the software and hardware themselves, which increases operational burden and complexity. With Gemini on GDC, you don’t have to compromise between the best of AI and the need to keep your data on-premises.

Our GDC air-gapped product, which is now authorized for US Government Secret and Top Secret missions, and on which Gemini is available, provides the highest levels of security and compliance.

aside_block: <ListValue: [StructValue([('title', '$300 in free credit to try Google Cloud AI and ML'), ('body', <wagtail.rich_text.RichText object at 0x7f2e257d42e0>), ('btn_text', 'Start building for free'), ('href', 'http://console.cloud.google.com/freetrial?redirectPath=/vertex-ai/'), ('image', None)])]>

Gemini on GDC: unlocking generative AI anywhere

Gemini models deliver breakthrough AI performance: they can analyze million-token contexts; are multimodal, i.e., can process diverse data formats such as text, image, audio and video; and operate globally across 100+ languages.

Further, the Gemini API offers AI inferencing without having to worry about infrastructure, OS management, or model lifecycle management. This enables you to:

Add your own business context: Use Retrieval Augment Generation (RAG) to personalize and augment the AI model’s output, eliminating the need for fine tuning or retraining the models.
Automate information processing and knowledge extraction: Improve employee efficiency by using gen AI to quickly summarize long documents, analyze sentiment in reports or feedback, or add captions to image, audio, and video content.
Create interactive conversational experiences: Build deeper customer relationships by enabling Gemini-powered customer support agents, chatbots via natural language, and employee assistants.
Tailor agents for your industry’s use case: Unlock highly specialized capabilities and workflows by developing tailored agents for everyone from financial advisors, to security assistants, to robotics.

“Gemini on Google Distributed Cloud will empower ServiceNow to augment powerful agentic AI capabilities such as reasoning in our existing systems via robust APIs. This strategic deployment allows us to explore and implement cutting-edge advancements while upholding our commitment to customer trust and data protection.” - Pat Casey, Chief Technology Officer & EVP of DevOps, ServiceNow

Vertex AI: one platform for cloud and on-prem

In addition to bringing Gemini to Google Distributed Cloud, customers today already benefit from the Vertex AI platform on GDC, which lets them accelerate the development, deployment, and management of agentic applications.

This complete AI platform offers:

Pre-trained APIs: Ready-to-use, task-optimized, pre-trained APIs based on advanced Google models for translation, speech-to-text, and optical character recognition (OCR). These APIs offer advanced features such as customizable glossaries and in-place document translation
Gen AI building tools: Open-source and third-party models with optimized inferencing on GKE, delivering fast startup and auto-scaling
Retrieval Augmented Generation (RAG): Grounding using Google Agentspace search and LLM API management and governance using Apigee on-prem
Built-in embeddings API and AlloyDB vector database: Powerful applications for personalization and recommendations, enabling improved user experiences

“With Google Distributed Cloud, Vertex AI, and Agentspace search, we will empower our Home Team innovators with a secure AI/ML platform and unified search, enabling the use of AI to enhance productivity and transform public safety for a safer and more secure future.” - Chee Wee Ang, Chief AI Officer, HTX

Google Agentspace: out-of-box access to on-prem data

Enterprises are eager to deploy gen AI, but they also struggle to connect large volumes of siloed information across various repositories and formats such as images, PDFs, and text. This hinders productivity and innovation. At the same time, building an in-house search solution is costly and requires access to scarce AI expertise.

We are excited to announce Google Agentspace search will be available on GDC, with public preview starting in Q3 2025. Google Agentspace search provides all enterprise knowledge workers with out-of-the-box capabilities that unify access to all your data in a secure, permissions-aware way.

Agentspace gives you access to:

Company-branded, multimodal search agent: A conversational search interface that can answer complex questions based on your company’s unique information, acting as a central source of enterprise truth for your entire organization
Pre-built enterprise data connectors: Connectors to index data from the most common on-prem enterprise systems (such as Confluence, Jira, ServiceNow, and Sharepoint)
Permissions-aware search results: Robust access control list (ACL) enforcement that help ensure that search results are permission-aware, maintaining security and compliance for all your on-prem data
Agentspace agents: Vertex AI is integrated out-of-the-box with Agentspace, starting with search agents, with more pre-built agents coming soon, and the ability to build your own

Get started with gen AI on GDC

We're constantly innovating on GDC to be the leading gen AI and modern application development that you can deploy anywhere. To bring Gemini and gen AI to your premises, please contact Google Cloud sales or reach out to any of our accredited global partners.