NVIDIA and Google Cloud Build Massive AI Infrastructure Scaling to Nearly 1 Million GPUs

NVIDIA and Google Cloud Partnership Reaches New Scale

NVIDIA and Google Cloud unveiled a significant expansion of their collaboration at Google Cloud Next in Las Vegas, introducing infrastructure designed to power the next generation of agentic AI and physical AI applications 1

1

. The partnership, which spans more than a decade of co-engineering efforts, now enables customers to build AI factories with NVIDIA Vera Rubin-powered A5X instances that can scale up to 960,000 NVIDIA Rubin GPUs in a multisite cluster 4

4

. This represents a fundamental shift in how enterprises can approach large-scale AI deployment, moving complex workflows from laboratory environments into production systems.

The announcement positions the full-stack AI platform as a turnkey solution for developers, startups, and enterprises seeking to deploy everything from agents managing complex workflows to robotics and digital twins on factory floors 1

1

. Google has quietly built one of the world's largest accelerated infrastructure deployments, with well over a million NVIDIA GPUs already deployed across its global fleet for internal products and Google Cloud services 2

2

.

Google Cloud AI Hypercomputer Unifies Computing Resources

At the heart of this expansion sits the Google Cloud AI Hypercomputer, a high-performance computing architecture that unifies performance-optimized hardware for compute, storage, networking, open software, and machine learning frameworks 3

3

. The AI Hypercomputer brings together Google's custom TPUv8 series, Axion CPUs, and NVIDIA Rubin GPUs to create what both companies describe as cloud-scale clusters for training, fine-tuning, inference, and simulation 2

2

.

The new NVIDIA Vera Rubin NVL72 rack-scale systems deliver up to 10 times lower inference cost per token and 10 times higher token throughput per megawatt than the prior generation through extreme co-design across chips, systems, and software 1

1

. The A5X instances use NVIDIA ConnectX-9 SuperNICs combined with next-generation Google Virgo networking, scaling to up to 80,000 NVIDIA Rubin GPUs within a single site cluster 1

1

.

NVIDIA Blackwell GPUs Power Diverse Workloads

Google Cloud's comprehensive NVIDIA Blackwell portfolio ranges from A4 VMs with NVIDIA HGX B200 systems to rack-scale A4X VMs with NVIDIA GB200 NVL72 and A4X Max NVIDIA GB300 NVL72 systems, extending to fractional G4 VMs with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs 4

4

. This variety allows customers to right-size their acceleration capabilities, whether using multiple interconnected NVL72 racks that scale out to tens of thousands of NVIDIA Blackwell GPUs or just one-eighth of a GPU 1

1

.

Leading AI organizations are already deploying this AI infrastructure at scale. OpenAI is running large-scale inference on NVIDIA GB300 and GB200 NVL72 systems on Google Cloud for some of its most demanding inference workloads, including for ChatGPT 4

4

. Thinking Machines Lab is scaling its Tinker API on A4X Max VMs with GB300 NVL72 systems to accelerate training clusters 1

1

.

Confidential Computing and Sovereign AI Capabilities

Google Gemini models running on NVIDIA Blackwell and Blackwell Ultra GPUs are now in preview on Google Distributed Cloud, enabling customers to bring frontier models wherever their most sensitive data resides 1

1

. The introduction of Confidential G4 VMs with NVIDIA RTX PRO 6000 Blackwell GPUs marks the first confidential computing offering of NVIDIA Blackwell GPUs in the cloud 4

4

.

Confidential computing with the NVIDIA Blackwell platform enables Gemini models to run in a protected environment where prompts and fine-tuning data stay encrypted and cannot be seen or altered by unauthorized parties, including infrastructure operators 1

1

. This capability addresses critical security concerns for regulated industries seeking to access AI power without compromising on data protection.

Physical AI and Robotics Applications Expand

The partnership extends into physical AI with NVIDIA Omniverse libraries and the NVIDIA Isaac Sim robotics simulation framework now available on Google Cloud Marketplace 4

4

. Solutions from Cadence and Siemens Digital Industries Software are now available on Google Cloud, accelerated on NVIDIA AI infrastructure for complex simulations and digital twins applications 4

4

.

For agentic AI development, NVIDIA Nemotron 3 Super is now available on Gemini Enterprise Agent Platform, and a new managed reinforcement learning API built with NVIDIA NeMo RL accelerates training at scale 4

4

. CrowdStrike uses NVIDIA NeMo open libraries to generate synthetic data and fine-tune Nemotron and other open large language models for cybersecurity applications 4

4

.

Customers including the US Department of Energy, Boston Dynamics, Citadel Securities, and Axia Energy are already leveraging the Google AI Hypercomputer for their workloads 3

3

. The collaboration reduces organizational friction by providing infrastructure teams, data scientists, and application teams with a common, battle-tested platform that shortens time-to-value for AI adoption 2

2

.