Amazon Web Services (AWS) and NVIDIA announced a significant enhancement to their strategic partnership aimed at meeting the exponentially growing global demand for advanced artificial intelligence infrastructure. The collaboration plans to deploy an additional 2 million NVIDIA GPUs across AWS’s worldwide infrastructure. Furthermore, the partners are deepening their joint efforts across multiple domains—including CPUs, networking, open models, data processing, and robotics—to provide customers with integrated, co-engineered AI solutions designed to accelerate development and deployment at an unprecedented scale.
Addressing Accelerating AI Workloads
The pace of AI development requires massive computational resources, spanning everything from training large language models to processing data and powering intelligent applications. As businesses transition AI capabilities from experimental pilots to full-scale production, they are increasingly focusing on agentic AI, scientific discovery, enterprise automation, and physical robotics. To support these diverse needs, customers require broader model selection, faster data pipelines, and reliable underlying infrastructure that can keep pace with innovation while maintaining stringent security standards for mission-critical operations.
Expanded Technical Commitments and Capabilities
To meet the escalating requirements from global enterprises, government bodies, and research institutions, AWS and NVIDIA are building upon 16 years of joint innovation. The expanded partnership includes several key technical deployments and enhancements:
- GPU Capacity Boost: A deployment of 2 million supplementary NVIDIA GPUs across AWS’s global network is slated for 2027–2028.
- CPU and Interconnects: The partnership will introduce NVIDIA Vera CPU-based infrastructure to AWS and enhance the system through the extension of NVIDIA NVLink Fusion™ with custom NVIDIA high-bandwidth memory (NVHBM).
- Government AI Factories: AWS and NVIDIA are collaborating to establish AI factories for the U.S. government, which will include 100,000 GPUs housed on secure AWS infrastructure to run federal and national-security workloads.
- Security and Reliability: The NVIDIA platform will be integrated with the existing AWS Nitro System and Elastic Fabric Adapter (EFA) to ensure maximum security and reliability for AI operations.
- Model Choice: Support for NVIDIA Nemotron™ open models will continue on Amazon Bedrock and Amazon SageMaker, giving clients expanded choices among open models.
- Data Processing: Faster, more efficient analytics and AI applications will be enabled by accelerating data processing and vector indexing on Amazon EMR and Amazon OpenSearch using NVIDIA cuDF and cuVS CUDA-X™ libraries.
- Robotics: Amazon Robotics will advance next-generation automation by integrating NVIDIA’s comprehensive physical AI platform, speeding up innovations in warehouse and advanced robotics.
Leadership Commentary
Commenting on the expansion, Matt Garman, CEO of AWS, stated, “Customers require the flexibility to select the optimal tools for their AI workloads, and they need assurance that all components function seamlessly together. Therefore, we have made substantial investments with NVIDIA to establish AWS as the premier location for running NVIDIA AI technologies, optimizing performance across our infrastructure, from networking to security. This expanded collaboration provides frontier labs, enterprises, and governments with enhanced methods for building and deploying AI on AWS.”
Jensen Huang, founder and CEO of NVIDIA, noted the strength of the partnership, stating, “NVIDIA and AWS have co-created a major growth engine for the AI era, and the demand is surpassing all forecasts. Over 16 years, we have scaled NVIDIA computing in the cloud together. Now, we are expanding our partnership across the entire stack—including GPUs, CPUs, networking, open models, and software—to actualize agentic and physical AI at a scale and pace that only AWS and NVIDIA can achieve. This growth reflects the strong demand from customers for NVIDIA’s platform on AWS.”
Details on Computational Capacity
AWS continues to offer the most extensive selection of GPU-based computing instances among all cloud providers. Building on previous announcements of adding over 1 million NVIDIA GPUs starting in 2026, AWS plans to deploy an additional 2 million NVIDIA Blackwell Ultra, Rubin, and Rubin Ultra GPUs between 2027 and 2028. This capacity boost will support various applications, including physical AI, scientific research, and enterprise automation. Furthermore, AWS will expand its Blackwell offerings, including NVIDIA RTX PRO™ 4500 Blackwell Server Edition GPUs for Amazon EC2 G7 instances. These G7 instances offer a 4.6x improvement in AI inference performance and a 2.1x improvement in graphics performance compared to the preceding G6 instances. AWS is the first major cloud provider to provide compute instances accelerated by the RTX PRO 4500. Additionally, AWS and NVIDIA are collaborating on NVIDIA Spectrum™ networking to further optimize performance for large-scale AI training across GPU clusters.
Additional Technological Support
AWS and NVIDIA are also advancing infrastructure options by bringing Vera CPU-based computing to AWS. This provides an additional, high-performance CPU compute option for agentic AI workloads, complementing AWS’s strategy to offer the widest variety of compute choices, from custom silicon to advanced accelerators and CPU partners. In terms of hardware interconnectivity, AWS and NVIDIA are expanding support for NVIDIA NVLink Fusion high-speed chip interconnect technology, coupling it with NVIDIA’s new custom high-bandwidth memory (NVHBM) technology. This combination allows for enhanced performance and efficiency for AI workloads, while seamlessly integrating Trainium and GPUs within a common, rack-scale architecture.
