AWS and NVIDIA Deepen Strategic Cloud Alliance: 2 Million Additional GPUs and Vera CPUs Headed to AWS
In a major milestone for global cloud computing and enterprise artificial intelligence, Amazon Web Services (AWS) and NVIDIA have announced a comprehensive expansion of their multi-year strategic collaboration. To keep up with unprecedented enterprise demand for generative, agentic, and physical AI, the cloud giant will deploy an additional 2 million NVIDIA GPUs across its global footprint through 2027 and 2028.
Scaling Next-Generation AI Infrastructure
This major expansion triples the commitment previously outlined earlier this year, scaling AWS's deployment pipeline to over 3 million NVIDIA GPUs. As enterprises transition AI initiatives from early prototypes into mission-critical production systems, access to scalable, high-throughput compute infrastructure has emerged as the central bottleneck for AI developers worldwide.
Bringing NVIDIA Vera CPUs to AWS for Agentic AI
Beyond raw GPU acceleration, the expanded partnership brings NVIDIA's specialized Vera CPUs to AWS infrastructure. Agentic AI workloads—characterized by multi-step reasoning, tool execution, sandboxing, and autonomous code synthesis—require substantial general-purpose compute alongside accelerated matrices. Introducing Vera CPU-based instances provides developers with fine-grained architectural flexibility to balance orchestration, data preprocessing, and GPU acceleration.
Key Highlights of the AWS & NVIDIA Collaboration
- 2 Million Additional GPUs: Expanding AWS global availability for frontier model training and real-time inference across 2027–2028.
- NVIDIA Vera CPU Integration: Optimized compute instances specifically engineered for complex agentic workflows and reinforcement learning loops.
- Dedicated Government AI Factories: Construction of secure sovereign AI infrastructure, including 100,000 GPUs on dedicated AWS environments for federal and national security workloads.
- Advanced Networking: Enhancing interconnect architectures through next-gen NVLink Fusion and high-bandwidth memory pipelines to maximize distributed cluster efficiency.
The Multi-Silicon Strategy
AWS emphasized that the expanded NVIDIA collaboration directly complements Amazon's custom silicon roadmap, including its in-house Trainium and Inferentia processors. By offering premier access to both NVIDIA's latest platforms and Amazon's purpose-built ASICs, AWS aims to provide enterprises with complete architectural sovereignty when tailoring compute economics to their specific workload profiles.