Sirf Such

Truth that Matters. Stories that Impact

Sirf Such

Truth that Matters. Stories that Impact

Technology

Amazon Expands Nvidia Partnership to Add Two Million AI GPUs Across AWS Data Centers

Amazon and Nvidia have deepened their strategic partnership through an agreement to supply an additional 2 million graphics processing units (GPUs) to Amazon Web Services data centers. The deal expands upon an earlier agreement made five months prior to deploy over 1 million processors, driven by what both firms describe as surging demand across startups, enterprises, AI research labs, and public sector organizations.

What Happened

During its quarterly earnings report, Nvidia announced that Amazon will incorporate 2 million high-performance GPUs into its cloud computing infrastructure. The scheduled hardware includes Nvidia Blackwell Ultra, Rubin, and Rubin Ultra processors, slated to arrive in AWS data centers across 2027 and 2028. Shipments of certain units are planned to begin rolling out to AWS starting in the third quarter.

Alongside GPUs, Nvidia will deliver an unspecified volume of its Vera central processing units (CPUs). As confirmed by Nvidia Chief Financial Officer Colette Kress, some Vera chips will be paired directly with Rubin GPUs, while others will operate as standalone units. The broader integration encompasses Nvidia networking gear, data processing software, open models, and dedicated robotics systems across Amazon operations.

Key Highlights

  • Significant Hardware Expansion: Amazon is adding 2 million advanced Nvidia GPUs, including Blackwell Ultra, Rubin, and Rubin Ultra architectures, to meet compute requirements for training and operating artificial intelligence models.
  • Comprehensive Integration: Beyond chips, AWS is incorporating Nvidia networking systems, CPU hardware, and data processing software, while hosting Nvidia Nemotron open foundation models on Amazon Bedrock and SageMaker.
  • Robotics Deployment: Amazon intends to deploy Nvidia’s physical AI software and hardware stack—comprising Omniverse, Cosmos, Isaac, and Jetson computing modules—to support its automated warehouse robot fleet.
  • Parallel In-House Development: The deal proceeds even as AWS develops its proprietary Trainium and Graviton chips, with Amazon reporting an annualized custom silicon revenue run rate exceeding $25 billion.
  • Strong Financial Performance: Nvidia recorded total revenue of $96.2 billion for its second quarter, with data center sales generating $89 billion, and projected third-quarter revenue of $108 billion.

Why This Matters

The scale of the procurement underscores the intense infrastructure requirements underpinning modern artificial intelligence. Despite Amazon investing heavily in internal silicon alternatives such as Trainium to curb reliance on third-party suppliers, market appetite for compute capacity has required substantial external procurement from Nvidia.

The collaboration also moves past raw processing hardware. By incorporating Nvidia’s full physical AI software and Jetson hardware within its fulfillment facilities, Amazon is tying its logistics automation directly into Nvidia’s specialized robotics ecosystem.

What to Watch Next

Market observers and stakeholders will monitor the rollout of Rubin production shipments, which Nvidia noted began this quarter, and examine third-quarter revenue targets projected at $108 billion. Additionally, attention remains focused on Nvidia’s capital commitments—totaling $279 billion to secure memory and manufacturing supply—to determine whether rapid hardware delivery can keep pace with widespread customer demand through 2028.

Frequently Asked Questions

Which specific Nvidia chips will Amazon deploy?

Amazon will add Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs, along with Vera CPUs deployed both as standalone processors and integrated alongside Rubin hardware.

When are the new processors scheduled to arrive in AWS facilities?

Deliveries are scheduled to roll out to Amazon Web Services data centers in 2027 and 2028, with initial chip additions beginning in the third quarter.

How does this agreement affect Amazon’s own chip projects?

Amazon continues to develop its custom silicon, including Trainium AI chips and Graviton CPUs. AWS AI chief Peter DeSantis confirmed the firm is discussing sales of Trainium hardware to external data centers, despite the concurrent Nvidia partnership expansion.

Source: TechCrunch and Nvidia earnings disclosures.