Amazon triples Nvidia chip order to 2 million GPUs as AI demand surges

Amazon and Nvidia have expanded their partnership, with Amazon committing to add an additional 2 million Nvidia GPU chips to its data centers by 2027 and 2028, the companies announced Wednesday during Nvidia’s quarterly earnings call. This follows an earlier agreement, made just five months ago, for more than 1 million GPUs to be deployed across Amazon Web Services (AWS) infrastructure starting this year. Nvidia stated that demand has since exceeded expectations, though financial terms were not disclosed. Industry analysts estimate the deal’s value in the tens of billions of dollars based on GPU unit costs.

Expanded partnership beyond chip supply

The collaboration extends beyond mere chip procurement. Nvidia will integrate its networking hardware, open models, CPUs, data processing software, and robotics platform across AWS. This includes Nvidia’s Vera CPUs, which will be deployed in AWS data centers, with some integrated with Rubin GPUs and others standalone, according to Nvidia CFO Colette Kress. The companies cited “surging demand” from startups, enterprises, AI labs, and governments as a key driver.

Amazon’s in-house chip efforts and Nvidia’s dominance

This expansion comes even as Amazon invests in its own AI chip development, including Trainium chips, which compete directly with Nvidia’s offerings for deep learning workloads, and Graviton CPUs, which challenge Intel and AMD in general-purpose processing. Amazon has indicated its custom chip business is growing, surpassing a $25 billion annualized revenue run rate, with $225 billion in total commitments from AI labs like Anthropic and OpenAI. However, Nvidia remains the dominant force in AI chips, as evidenced by this latest order.

Why this matters for the AI industry

The deal underscores the intensifying demand for AI compute infrastructure and Nvidia’s continued market leadership. It also highlights the strategic importance of cloud providers like AWS in meeting the needs of AI companies, even as those providers develop their own silicon. Nvidia’s CEO Jensen Huang emphasized that AI is now generating “profitable tokens,” justifying massive infrastructure investments. The company reported Q2 sales of $96.2 billion, with data center revenue up 117% year-over-year, and expects $108 billion in Q3 revenue, partly driven by its next-gen Rubin GPUs.

Broader implications and future outlook

Nvidia’s commitment to secure supply and manufacturing capacity has risen to $279 billion, up from $119 billion last quarter, reflecting its aggressive push to meet AI demand. The partnership with Amazon also extends to physical AI, with Amazon adopting Nvidia’s robotics stack (Omniverse, Cosmos, Isaac, Jetson) for its warehouse robots, and AWS will offer Nvidia’s Nemotron open models on Bedrock and SageMaker. Investors will be watching whether the massive capital expenditures by AI companies translate into sustained profits, as Huang suggests.

Conclusion

The Amazon-Nvidia expansion marks a significant milestone in the AI infrastructure race, demonstrating the scale of demand for advanced GPUs and the deepening interdependence between cloud providers and chipmakers. While Amazon continues to diversify its own chip portfolio, Nvidia’s technology remains central to AWS’s AI capabilities, and the partnership is poised to shape the future of AI computing.

FAQs

Q1: What is the value of the new Amazon-Nvidia deal?
Financial terms were not disclosed, but based on GPU unit costs, the deal is estimated to be worth tens of billions of dollars.

Q2: When will the additional 2 million GPUs be deployed?
The GPUs, including Nvidia Blackwell Ultra, Rubin, and Rubin Ultra, will be added to AWS data centers in 2027 and 2028.

Q3: How does this affect Amazon’s own chip development?
Amazon continues to invest in its own AI chips like Trainium and Graviton, but the expanded partnership indicates Nvidia remains the primary supplier for AWS’s AI compute needs.