In a significant move at the AWS re:Invent event, NVIDIA and Amazon Web Services (AWS) have announced a deepened strategic alliance aimed at advancing the frontiers of artificial intelligence (AI) technology. This collaboration leverages new technology integrations across various domains, including interconnect technology, cloud infrastructure, open models, and physical AI, promising to enhance AI capabilities and infrastructure significantly.
A pivotal aspect of this collaboration is AWS’s support for NVIDIA NVLink Fusion, a robust platform designed for custom AI infrastructure. This platform is set to play a critical role in deploying AWS’s custom-designed silicon, including the next-generation Trainium4 chips. These chips are tailored for inference and agentic AI model training. Additionally, AWS plans to integrate Graviton CPUs for diverse workloads, coupled with the Nitro System virtualization infrastructure, ensuring a seamless deployment process.
NVIDIA NVLink Fusion facilitates AWS’s ability to combine NVIDIA’s NVLink scale-up interconnect and MGX rack architecture with AWS’s custom silicon. This amalgamation is expected to enhance performance considerably and accelerate the rollout of AWS’s next-generation cloud-scale AI capabilities. The integration of these technologies signifies the beginning of a multigenerational collaboration between NVIDIA and AWS, particularly regarding NVLink Fusion.
AWS has already incorporated MGX racks at scale, utilizing NVIDIA GPUs. The integration of NVLink Fusion is anticipated to simplify deployment further and streamline systems management across AWS platforms. Leveraging the NVLink Fusion supplier ecosystem, AWS can access all necessary components for full rack-scale deployment, including rack and chassis, power-delivery, and cooling systems. This comprehensive support will bolster AWS’s Elastic Fabric Adapter and Nitro System, enhancing the NVIDIA Vera Rubin architecture on AWS and offering customers robust networking choices. This ensures full compatibility with AWS’s cloud infrastructure and facilitates the rapid rollout of new AI services.
As highlighted by Jensen Huang, founder and CEO of NVIDIA, the demand for GPU computing is skyrocketing. He posits that this demand is fueling a virtuous cycle of AI development, where increased compute power leads to smarter AI, which in turn broadens AI’s application and drives further demand for compute resources. He emphasized that the integration of NVIDIA NVLink Fusion with AWS Trainium4 represents a unification of scale-up architecture with AWS’s custom silicon, setting the stage for a new generation of accelerated platforms. This collaboration between NVIDIA and AWS is seen as a cornerstone in building the compute fabric for the AI industrial revolution, aiming to make advanced AI accessible to every company globally.
Matt Garman, CEO of AWS, echoed these sentiments, noting that AWS and NVIDIA have been synergistically working for over 15 years. This renewed partnership marks a significant milestone in their journey, focusing on advancing large-scale AI infrastructure to deliver unparalleled performance, efficiency, and scalability. The integration of NVIDIA NVLink Fusion within AWS Trainium4, Graviton, and the Nitro System is expected to unlock new capabilities, enabling customers to innovate more rapidly than ever before.
### Convergence of Scale and Sovereignty
AWS has expanded its accelerated computing portfolio with the NVIDIA Blackwell architecture, featuring NVIDIA HGX B300 and NVIDIA GB300 NVL72 GPUs. This expansion provides customers with immediate access to some of the industry’s most advanced GPUs for training and inference. Furthermore, the NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, designed for visual applications, are anticipated to be available on AWS soon.
These cutting-edge GPUs form a crucial part of the AWS infrastructure, powering AWS AI Factories. These AI Factories represent a new AI cloud offering that equips customers worldwide with the dedicated infrastructure necessary to leverage advanced AI services and capabilities within their data centers, all while maintaining control over their data and adhering to local regulations.
NVIDIA and AWS are committed to deploying sovereign AI clouds globally, aiming to bring the pinnacle of AI innovation to the world. With the launch of AWS AI Factories, the companies are delivering secure, sovereign AI infrastructure, enabling organizations to access unprecedented computing capabilities while meeting rigorous sovereign AI requirements.
For public sector organizations, AWS AI Factories are set to revolutionize the federal supercomputing and AI landscape. These AI Factories allow customers to seamlessly integrate AWS’s industry-leading cloud infrastructure and services, known for reliability, security, and scalability, with NVIDIA Blackwell GPUs and the comprehensive NVIDIA accelerated computing platform, including NVIDIA Spectrum-X Ethernet switches. This unified architecture ensures that customers can access advanced AI services and capabilities, train and deploy massive models, while maintaining absolute control over proprietary data and full compliance with local regulatory frameworks.
### NVIDIA Nemotron Integration With Amazon Bedrock Expands Software Optimizations
Beyond hardware, this partnership extends to the integration of NVIDIA’s software stack with the AWS AI ecosystem. NVIDIA Nemotron open models are now integrated with Amazon Bedrock, enabling customers to build generative AI applications and agents at a production scale. Developers can utilize Nemotron Nano 2 and Nemotron Nano 2 VL to develop specialized agentic AI applications that efficiently process text, code, images, and video with high accuracy.
This integration ensures that high-performance, open NVIDIA models are readily accessible via Amazon Bedrock’s serverless platform, offering proven scalability and zero infrastructure management. Prominent industry players like CrowdStrike and BridgeWise are among the first to adopt this service to deploy specialized AI agents.
### NVIDIA Software on AWS Simplifies Developer Experience
NVIDIA and AWS are co-engineering at the software layer to accelerate the data backbone of enterprises. The Amazon OpenSearch Service now offers serverless GPU acceleration for vector index building, powered by NVIDIA cuVS, an open-source library for GPU-accelerated vector search and data clustering. This marks a fundamental shift toward utilizing GPUs for unstructured data processing, with early adopters witnessing up to 10x faster vector indexing at a fraction of the cost.
These advancements significantly reduce search latency, accelerate data writes, and enhance productivity for dynamic AI techniques like retrieval-augmented generation by delivering the precise amount of GPU power needed at any given moment. AWS is the first major cloud provider to offer serverless vector indexing with NVIDIA GPUs.
To support production-ready AI agents, the collaboration includes Strands Agents for agent development and orchestration, the NVIDIA NeMo Agent Toolkit for deep profiling and performance tuning, and Amazon Bedrock AgentCore for secure, scalable agent infrastructure. This comprehensive suite empowers developers with a predictable path from prototype to production.
Building on AWS’s existing integrations with NVIDIA technologies—such as NVIDIA NIM microservices and frameworks like NVIDIA Riva and NVIDIA BioNeMo, alongside model development tools integrated with Amazon SageMaker and Amazon Bedrock—organizations can deploy agentic AI, speech AI, and scientific applications more swiftly than ever before.
### Accelerating Physical AI With AWS
Developing physical AI requires high-quality and diverse datasets for training robot models, as well as frameworks for testing and validation in simulations before real-world deployment. NVIDIA Cosmos world foundation models (WFMs) are now available as NVIDIA NIM microservices on Amazon EKS, enabling real-time robotics control and simulation workloads with seamless reliability and cloud-native efficiency. For batch-based tasks and offline workloads, such as large-scale synthetic data generation, Cosmos WFMs are also available on AWS Batch as containers.
Cosmos-generated world states can be utilized to train and validate robots using open-source simulation and learning frameworks like NVIDIA Isaac Sim and Isaac Lab. Leading robotics companies, including Agility Robotics, Agile Robots, ANYbotics, Diligent Robotics, Dyna Robotics, Field AI, Haply Robotics, Lightwheel, RIVR, and Skild AI, are utilizing the NVIDIA Isaac platform with AWS for various use cases, from collecting and processing robot-generated data to training and simulation for scaling robotics development.
### Sustained Collaboration
Highlighting years of sustained collaboration, NVIDIA has been recognized with the AWS Global GenAI Infrastructure and Data Partner of the Year award. This accolade acknowledges top technology partners with the Generative AI Competency that support vector embeddings, data storage and management, or synthetic data generation across multiple types and formats.
To explore more about NVIDIA and AWS’s collaboration, sessions are available at AWS re:Invent, which runs through December 5 in Las Vegas.
For more Information, Refer to this article.
