AI workloads are moving beyond simple model training toward agentic AI, real-time inference, scientific computing, automation, and robotics. The expanded AWS NVIDIA partnership aims to give businesses and AI developers the computing power and infrastructure needed to support that shift.
AWS and NVIDIA announced on August 26, 2026, that they plan to deploy 2 million additional NVIDIA GPUs across AWS’s global infrastructure in 2027–2028 while expanding their collaboration across CPUs, networking, AI models, data processing, and robotics.
Caption: AWS and NVIDIA are expanding their collaboration to support large-scale AI workloads.
What the AWS NVIDIA Partnership Expansion Includes
The new agreement goes beyond adding more GPUs. AWS and NVIDIA are working together across several parts of the AI technology stack.
Key areas include:
- 2 million additional NVIDIA GPUs planned for AWS infrastructure in 2027–2028
- NVIDIA Vera CPUs coming to AWS
- Advanced networking for large-scale AI clusters
- Support for NVIDIA Nemotron open models
- Faster data processing and vector indexing
- AI infrastructure for government workloads
- Greater support for robotics and physical AI
This broader approach is important because modern AI systems require more than GPUs. Data movement, CPU processing, networking, storage, and software can all affect overall performance.
NVIDIA GPUs on AWS Will Scale Further
The biggest headline from the AWS NVIDIA partnership is the planned addition of 2 million GPUs.
AWS says the new capacity will include NVIDIA Blackwell Ultra, Rubin, and Rubin Ultra GPUs across its global infrastructure and AI factories. The goal is to support workloads such as:
- AI model training
- Agentic AI
- Large-scale inference
- Scientific discovery
- Enterprise automation
- Robotics and physical AI
For businesses, increased capacity could make it easier to scale AI workloads without building and operating their own specialized data centers.
AWS is also expanding NVIDIA Blackwell capacity, including RTX PRO 4500 Blackwell Server Edition GPUs for Amazon EC2 G7 instances.
NVIDIA Vera CPUs Are Coming to AWS
GPUs get much of the attention in AI infrastructure, but CPUs remain essential for many workloads.
AWS and NVIDIA are working to bring NVIDIA Vera CPU-based infrastructure to AWS. Vera is designed for workloads associated with agentic AI and reinforcement learning, including data processing, code execution, tool use, orchestration, and analytics.
This could be particularly useful as AI agents become more complex.
An AI agent may need to reason with a model, access databases, execute code, call external tools, and process information before producing an answer. Those tasks place demands on both accelerated computing and CPU infrastructure.
A Broader AWS NVIDIA AI Infrastructure Strategy
The partnership also extends into networking and data processing.
AWS and NVIDIA are collaborating on networking technologies intended to improve communication between GPUs in large AI clusters. Efficient networking becomes increasingly important as AI systems scale to thousands or potentially millions of accelerators.
The companies are also integrating NVIDIA technologies with AWS infrastructure such as the AWS Nitro System and Elastic Fabric Adapter, with a focus on security, reliability, and high-performance workloads.
For data-heavy AI applications, NVIDIA CUDA-X libraries such as cuDF and cuVS will also be used with AWS services including Amazon EMR and Amazon OpenSearch.
Why This Matters for AI Developers and Businesses
The expanded AWS AI infrastructure could help organizations move AI projects from experimentation into production.
For example, an enterprise developing an AI assistant may need:
- GPUs to train or run models.
- CPUs to handle application logic and agent workflows.
- High-speed networking to connect compute resources.
- Data processing infrastructure to prepare information.
- Cloud services to deploy and manage the application.
Bringing more of these components together can simplify the infrastructure required for large AI deployments.
Agentic AI and Physical AI Are Major Focus Areas
The partnership is particularly focused on two emerging areas: agentic AI and physical AI.
Agentic AI systems can perform multi-step tasks, use tools, interact with data, and take actions with less direct human intervention. Physical AI applies AI to real-world environments, including robotics and automated systems.
AWS says Amazon Robotics is also adopting NVIDIA’s physical AI platform, potentially helping accelerate development of warehouse automation and next-generation robots.
The companies are additionally planning AI factories for the U.S. government, including 100,000 GPUs on secure AWS infrastructure for federal and national-security workloads.
What the AWS NVIDIA Partnership Means for the Future
The AWS NVIDIA partnership shows how AI infrastructure is evolving from individual GPU instances into complete computing platforms.
The next generation of AI workloads will require a combination of:
- High-performance GPUs
- Specialized CPUs
- Fast networking
- Efficient data pipelines
- AI software and models
- Secure cloud infrastructure
For cloud customers, the expansion could mean more choices and greater access to NVIDIA-powered computing through AWS as AI demand continues to grow.
Caption: Next-generation AI infrastructure will combine GPUs, CPUs, networking, software, and cloud services.