Key Moments
- NVIDIA Corporation (NASDAQ:NVDA) plans to supply 1 million GPUs to Amazon.com’s (NASDAQ:AMZN) AWS between 2026 and 2027, with deliveries starting this year and running through 2027.
- The arrangement spans a full stack of Nvidia technology, including GPUs, Spectrum networking chips, Groq chips, and Connect X and Spectrum X networking equipment for AWS data centers.
- The AWS timeline lines up with CEO Jensen Huang’s view of a $1 trillion sales opportunity through 2027 for Nvidia’s Rubin and Blackwell chip families, excluding CPUs, networking, and Groq-related products.
Broad Multi-Year Supply Agreement With AWS
NVIDIA Corporation (NASDAQ:NVDA) will supply 1 million graphics processing units to Amazon.com’s (NASDAQ:AMZN) Amazon Web Services cloud division between 2026 and 2027, according to comments an executive made to Reuters. The disclosure marks the first time Nvidia has provided a clear timeframe for a large-scale cloud partnership that spans multiple chip families and networking systems.
Ian Buck, Nvidia’s vice president of hyperscale and high-performance computing, told Reuters that shipment of the GPUs will begin this year and continue through 2027, covering the entire multi-year commitment.
Beyond GPUs: Groq and Networking Technologies
Buck indicated that the transaction is not limited to GPU hardware. Under the agreement, Amazon Web Services will also buy Nvidia’s Spectrum networking chips along with recently introduced Groq chips. Nvidia obtained rights to Groq technology via a $17 billion licensing agreement with AI chip company Groq in late 2025.
AWS intends to deploy Groq chips together with six additional Nvidia chip types to fine-tune performance for AI inference tasks, where AI models generate outputs and carry out operations in response to input data.
“Inference is hard. It’s wickedly hard,” Buck told Reuters. “To be the best at inference, it is not a one chip pony. We actually use all seven chips.”
Shift in AWS Networking Strategy
The collaboration also covers Nvidia’s Connect X and Spectrum X networking products, which are slated to be installed in AWS data centers. This represents a notable change for AWS, which has long leaned on in-house networking hardware developed and refined internally over many years.
“They’re still going to do that, of course,” Buck said regarding AWS’s proprietary solutions. “But we are collaborating now on deploying Connect X and Spectrum X for those important workloads and biggest customers across AI with AWS.”
The companies have not publicly released any financial details related to the agreement.
Alignment With Nvidia’s Long-Term Revenue Ambitions
The schedule for the AWS partnership lines up with CEO Jensen Huang’s outlook that Nvidia is pursuing a $1 trillion sales opportunity for its Rubin and Blackwell chip platforms through 2027. That projection excludes CPUs, networking hardware, Groq-based offerings, and a Rubin variant labeled Rubin Ultra, implying that the overall potential revenue tied to the broader portfolio could go beyond that figure.
Huang has suggested that incorporating Groq technology could generate $300 billion in annual revenue per gigawatt, with expectations that about 25% of GPU workloads will be paired with Groq chips. The combined Nvidia-Groq configuration, referred to as LPX, is currently positioned as an optional component of Nvidia’s Vera Rubin platform and is not yet broadly deployed at scale.
Deal Structure Overview
| Component | Details |
|---|---|
| GPU supply | 1 million units to AWS between 2026 and 2027, with deliveries starting this year and extending through 2027 |
| Additional chips | Groq chips plus six other Nvidia chip types for AI inference optimization |
| Networking | Nvidia Spectrum networking chips, Connect X, and Spectrum X for AWS data centers |
| Groq deal structure | $17 billion licensing agreement with Groq in late 2025 |
| Financial terms | Not disclosed by Nvidia or AWS |
Key Issues for Market Participants
For investors tracking Nvidia, Amazon, and the broader AI infrastructure landscape, several elements of this relationship could be particularly important:
- GPU delivery pace – How Nvidia executes on the 1 million unit commitment and how those shipments are distributed across different GPU generations.
- Groq deployment timeline – When AWS moves from pilots to broader use of Groq-based, inference-focused chips.
- Networking adoption – The degree to which AWS incorporates Nvidia’s Connect X and Spectrum X gear relative to its existing custom-built networking solutions.
- Fiscal 2027 impact – Fiscal 2027 results, expected in early 2027, will offer a clearer view of how the AWS agreement contributes to Nvidia’s financial performance.
- Competitive backdrop – How AMD and other competitors respond as Nvidia deepens its integration within AWS’s infrastructure stack.





