NVIDIA and Google Cloud have worked together for more than a decade to co-design full-stack AI platforms that span all technology layers, from performance-optimized libraries and frameworks to enterprise-grade cloud services.
This foundation enables developers, startups, and enterprises to push agent and physical AI out of the lab and into production, from agents that manage complex workflows to factory floor robots and digital twins.
At Google Cloud Next in Las Vegas this week, the partnership will reach a new milestone as we advance the expansion of Google Cloud AI hypercomputers for AI factories that power the next frontiers of agent and physical AI.
These include new NVIDIA Vera Rubin– Powered A5X bare metal instance. be preview Google Gemini running on Google Distributed Cloud NVIDIA Blackwell NVIDIA Blackwell Ultra GPU. Confidential VM with NVIDIA Blackwell GPU. Agent AI on Gemini Enterprise Agent Platform and NVIDIA Nemotron open model and NVIDIA NeMo framework.
Next Generation Infrastructure: From NVIDIA Blackwell to Vera Rubin
At Google Cloud Next, Google announced A5X with: NVIDIA Vera Rubin NVL72 Rack-scale systems deliver up to 10x lower inference cost per token and 10x higher token throughput per megawatt compared to previous generations due to thorough co-design across chips, systems, and software.
A5X uses NVIDIA ConnectX-9 SuperNICscalable to maximum with next-generation Google Virgo networking 80,000 NVIDIA Rubin GPU in a single-site cluster and up to 960,000 NVIDIA Rubin GPUs in multi-site clusters enable customers to run their largest AI workloads on NVIDIA-optimized infrastructure.
“At Google Cloud, we believe the next decade of AI will be shaped by our customers’ ability to run their most demanding workloads on a truly integrated, AI-optimized infrastructure stack.” Mark Lohmeyer, vice president and general manager of AI and compute infrastructure at Google Cloud, said: “Google Cloud’s scalable infrastructure and managed AI services, combined with NVIDIA’s industry-leading platforms, systems, and software, give us the flexibility to train, tune, and deliver everything from frontier and open models to agentic and physical AI workloads while optimizing performance, cost, and sustainability.”
Google Cloud’s extensive 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. Fractional G4 VM and NVIDIA RTX PRO 6000 Blackwell Server Edition GPU.
Customers can size their acceleration capabilities accordingly, whether using multiple interconnected NVL72 racks to scale out to tens of thousands of NVIDIA Blackwell GPUs, a single rack and up to 72 Blackwell GPUs on 5th generation NVIDIA NVLink and NVLink 5 switches, or just one-eighth of a GPU.
This comprehensive platform helps teams optimize any workload, from expert mixed inference, multimodal inference, and data processing to complex simulations for the next frontiers of physical AI and robotics.
State-of-the-art AI labs are already putting this infrastructure to work. Thinking Machine Laboratory is extending the Tinker application programming interface (API) on A4X Max VM with GB300 NVL72 system to accelerate training. OpenAI runs large-scale inference on NVIDIA GB300 (A4X Max VM) and GB200 NVL72 systems (A4X VM) on Google Cloud for some of the most demanding inference workloads, such as ChatGPT.
Protecting AI where it needs to run: Sovereignty and confidentiality
Google Gemini models running on NVIDIA Blackwell and Blackwell Ultra GPUs are currently In preview Available on Google Distributed Cloud, customers can deploy Google’s frontier models wherever their most sensitive data resides.
NVIDIA Confidential Computing The NVIDIA Blackwell platform allows Gemini models to run in a protected environment, where prompts and tweak data remain encrypted and cannot be viewed or modified by unauthorized parties, including infrastructure operators.
In the public cloud, preview Confidential G4 VMs with NVIDIA RTX PRO 6000 Blackwell GPUs bring these protections to multi-tenant environments, helping protect prompts, AI models, and data, allowing customers in regulated industries to access the power of AI without sacrificing security or performance.
This is the first classified computing offering for NVIDIA Blackwell GPUs in the cloud, providing Google Cloud customers with a new foundation for secure, high-performance AI.
Agentic AI open model and API
The NVIDIA platform on Google Cloud is optimized to run all types of models, starting with Google’s frontier Gemini. Gemma Integrates the NVIDIA Nemotron open model and family into the broader open weight ecosystem, enabling developers to build agent AI systems that reason, plan, and execute.
NVIDIA Nemotron 3 Super is available on the Gemini Enterprise Agent Platform, giving developers a direct path to discover, customize, and deploy NVIDIA-optimized multimodal models for inference and agent workflows.
Google Cloud and NVIDIA also make it easy to train and customize open models at scale. Managed training clusters on Gemini Enterprise Agent Platform introduces a new managed reinforcement learning (RL) API built on: NVIDIA NeMo RL Accelerate RL training at scale while automating cluster sizing, disaster recovery, and job execution, so your team can focus on agent behavior and model quality instead of infrastructure management.
Cyber security leader cloud strike Purpose NVIDIA NeMo Generate synthetic data using open libraries such as NeMo Data Designer, NeMo Automodel, and NeMo Megatron Bridge to fine-tune Nemotron and other open large-scale language models for domain-specific cybersecurity. These capabilities run on managed training clusters on the Gemini Enterprise Agent Platform, powered by NVIDIA Blackwell GPUs, to accelerate threat detection, investigation, and response.
Building the future of industrial and physical AI
Building industrial and physical AI at scale requires powerful hardware and a combination of open models, libraries, and frameworks to develop complex end-to-end workflows.
NVIDIA AI Infrastructure, open models and physical AI libraries available on Google Cloud are becoming mainstream industrial Physical AI applications allow customers to simulate, optimize, and automate real-world workflows.
Solutions from leading industrial software providers including: cadence and Siemens Digital Industries Softwarenow available on Google Cloud and accelerated by NVIDIA AI infrastructure. These applications are driving the next generation of design, engineering, and manufacturing of everything from chips to self-driving cars, robotics, aerospace platforms, heavy equipment, and large-scale production systems.
and NVIDIA Omniverse Libraries and open source NVIDIA Isaac Sim Robotics Simulation Framework is available at Google Cloud Marketplacedevelopers can build physically accurate digital twins and develop custom robot simulation pipelines to train, simulate, and validate robots before actual deployment.
NVIDIA NIM microservices for models such as: Why NVIDIA Cosmos 2 Can be deployed on Google Vertex AI and Google Kubernetes Engine. This enables robots and vision AI agents to see, reason, and act on the physical world just like humans, powering use cases such as automated data curation and annotation, advanced robot planning and reasoning, and intelligent video analytics agents for real-time insights and decision-making.
Together, these technologies enable developers to seamlessly move from computer-aided design to real-world industrial digital twins and AI-powered robots, accelerating processes from design approval to factory optimization on the NVIDIA platform running on Google Cloud.
Proven effectiveness: from startups to global enterprises
Global enterprises, AI labs, and high-growth startups use NVIDIA and Google Cloud’s co-design platform to move from prototyping to production faster. snap, schrodinger and sales force. snap reduces the cost of A/B testing at scale by moving their data pipeline to GPU-accelerated Spark on Google Cloud. schrodinger uses NVIDIA accelerated computing on Google Cloud to reduce weeks of drug discovery simulations to just hours.
Startups are orchestrating the next wave of AI innovation by building new agents and AI-native applications using NVIDIA-accelerated computing on Google Cloud.
as part of broader ecosystem highlighted through NVIDIA Inception Google for Startups, code rabbit and factory Powering code reviews and autonomous software development agents using NVIDIA Nemotron-based models on Google Cloud. Able, Mantis AI, photo room and base ten builds enterprise data, video intelligence, generated imagery, and managed inference solutions on the full-stack NVIDIA platform on Google Cloud.
Over 90,000 developers join NVIDIA and Google Cloud joint project developer community In just over a year, we were able to leverage this platform to build and scale new AI applications.
Additionally, NVIDIA received the following awards at Next: Google Cloud Partner of the Year Two categories recognized our deep technical expertise and go-to-market alignment: AI Global Technology Partners and Infrastructure Modernization Computing.
Together, NVIDIA and Google Cloud offer customers a cloud-scale platform to turn experimental agents and simulations into production systems that review code, secure fleets, enable new AI applications, and optimize real-world factories.
Join us to learn more about corporate collaboration NVIDIA sessions, demos, and workshops On Google Cloud Next.
