Key Takeaways
- Amazon Web Services has committed to purchasing 1 million Nvidia GPUs through the end of 2027.
- Deliveries commence in 2025 and continue for three years.
- The agreement encompasses networking infrastructure, Groq inference processors, and upcoming Blackwell and Rubin architectures.
- AWS plans to deploy seven distinct Nvidia chip types for AI inference operations.
- Shares of NVDA and AMZN both gained ground during extended trading hours after the disclosure.
The partnership between Nvidia and Amazon Web Services represents one of the semiconductor giant’s most substantial single-client hardware commitments to date. Closer examination reveals increasingly compelling strategic implications.
Ian Buck, Vice President at Nvidia, disclosed to Reuters that the company will commence shipping the 1 million GPUs starting in 2025, with deliveries extending through 2027. This delivery schedule aligns precisely with CEO Jensen Huang’s forecast projecting a $1 trillion addressable market for the Blackwell and Rubin processor families during this identical timeframe.
The arrangement extends considerably beyond simple GPU quantities. Amazon Web Services is acquiring Nvidia’s comprehensive hardware ecosystem, which includes Spectrum-X and ConnectX networking solutions. This element carries particular significance given that AWS has traditionally relied on proprietary networking infrastructure. Integrating Nvidia’s networking technology into its data center operations signals a notable strategic evolution.
Amazon Embraces Comprehensive Nvidia Inference Strategy
AI inference — the computational stage where artificial intelligence systems produce outputs and execute tasks — forms the foundation of this arrangement’s technical blueprint. Amazon Web Services intends to utilize seven different Nvidia chips for managing inference operations.
Buck articulated the strategy clearly: “Inference is hard. It’s wickedly hard. To be the best at inference, it is not a one chip pony. We actually use all seven chips.”
The Groq processors, unveiled by Nvidia this week after securing a $17 billion licensing arrangement with an AI semiconductor startup, form part of the inference infrastructure. These chips operate in conjunction with six additional Nvidia processors to provide what the company characterizes as industry-leading inference capabilities.
Amazon Web Services will also integrate Nvidia’s Blackwell processors and anticipates implementing the forthcoming Rubin platform upon its market availability. Neither Nvidia nor Amazon have revealed the monetary terms of this partnership.
Both companies saw their share prices advance moderately during Thursday’s after-hours session following the announcement. NVDA concluded regular trading approximately 1% lower, while AMZN finished down roughly 0.5%.
Amazon Maintains Internal Chip Development Programs
Amazon continues advancing its proprietary AI processors, including the Trainium2 chip. Alongside this internal development, the company maintains its commitment to Nvidia hardware for the most resource-intensive applications. These two strategies function as complementary approaches within AWS’s broader infrastructure plan.
The agreement demonstrates ongoing substantial capital deployment in AI infrastructure among leading cloud service providers. AWS continues operating its custom-designed systems while simultaneously adding Nvidia hardware for particular high-intensity computing scenarios.
The Nvidia-AWS partnership received its initial announcement earlier this week without specific timeline details. Buck’s Thursday statements to Reuters delivered the most comprehensive overview to date: revenue generation beginning in 2025, extending through December 2027, and encompassing a diverse portfolio of Nvidia offerings across computing, networking, and inference categories.
