iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
OpenAI’s November 2025 agreement with Amazon Web Services was announced as a $38 billion, multi-year deal for AI computing capacity, including hundreds of thousands of NVIDIA GPUs in AWS EC2 UltraServers. OpenAI named GB200 and GB300 clusters for ChatGPT inference and training future models, with deployment targeted before the end of 2026—not confirmed as complete. In 2026, the companies announced a further $100 billion expansion over eight years, including about two gigawatts of AWS Trainium capacity.
What was the original $38 billion deal?
On November 3, 2025, OpenAI and AWS announced a multi-year agreement valued at $38 billion. Under it, AWS would provide computing capacity for OpenAI workloads, with the arrangement planned to grow over seven years. OpenAI described access to hundreds of thousands of NVIDIA GPUs and said the infrastructure could scale to tens of millions of CPUs for agentic workloads. OpenAI’s announcement said the capacity would support ChatGPT inference and training next-generation models.
The deal is an infrastructure commitment, not a consumer product launch. EC2 UltraServers are AWS data-center systems, and the announcement does not give an itemized price per GPU, a public deployment count, or independent performance results.
Which Blackwell chips and systems are involved?
OpenAI’s announcement specifically names clusters of NVIDIA GB200 and GB300 GPUs. Separately, AWS and NVIDIA have described AWS plans involving Blackwell products including GB200 systems and B100 GPUs. Those broader platform announcements provide context; they do not establish that every AWS Blackwell product is part of OpenAI’s contract.
#1 Best Overall
- PLEASE NOTE: Exporting an NVIDIA RTX Pro 6000 GPU outside the US requires strict adherence to the U.S. Export Administration Regulations (EAR) and issuance of an export license from the Bureau of Industry and Security (BIS). Compliance and Know Your Customer (KYC) screening may be required as a condition of order acceptance. [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations.
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
NVIDIA describes the GB200 NVL72 as a system with 72 Blackwell GPUs and 36 Grace CPUs connected through fifth-generation NVLink, alongside AWS networking and infrastructure components. That is a configuration detail for the NVL72 system, not a reported count of systems deployed for OpenAI. NVIDIA’s AWS collaboration announcement does not establish an independently measured performance result for OpenAI’s workloads.
What was the deployment timeline?
OpenAI said in November 2025 that all capacity under the original agreement was targeted for deployment before the end of 2026, with possible expansion into 2027 and beyond. That was a target, not confirmation that the full capacity had been installed or was in service. The announcement does not provide a later verified deployment total. The Associated Press’s November 3, 2025 report likewise covered the announced deal, not a completed deployment milestone.
Rank #2
- Professional GPU with Blackwell Architecture
- Blackwell Architecture
- 24GB GDDR7 with PCIe 5.0 & Ray Tracing
- AI Workstation
How did the 2026 expansion change the deal?
In 2026, OpenAI and Amazon announced an expansion of the existing agreement by $100 billion over eight years. The expansion includes OpenAI’s commitment to consume approximately two gigawatts of AWS Trainium capacity, spanning Trainium3 and next-generation Trainium4. The companies associated the capacity with Stateful Runtime Environment, Frontier, and other advanced workloads. Trainium4 delivery was expected to begin in 2027, a forward-looking timeline rather than a confirmed delivery. OpenAI’s 2026 announcement and Amazon’s announcement describe the expansion and related plans.
Free tools Windows power users keep installed
One-click scans. No signup required.
The companies also said AWS would be the exclusive third-party cloud distribution provider for OpenAI Frontier, and that a joint Stateful Runtime Environment was planned for availability through Amazon Bedrock. They described customized models for Amazon customer-facing applications. These are service and distribution plans in addition to compute capacity; they should not be confused with the NVIDIA GPU portion of the 2025 announcement.
Rank #3
- Form Factor: Plug-in Card
- Cooler Type: Active Cooler
- Maximum Power Consumption: 70W
- Length: 6.6
- Height: 2.7
How to read the deal’s headline figures
| Figure | What the companies said it represents | Important qualification |
|---|---|---|
| $38 billion | Value of the original AWS agreement, announced by OpenAI in 2025 | Multi-year agreement; not a disclosed per-GPU price |
| Seven years | Period over which the original arrangement was to continue growing | OpenAI’s stated plan in 2025 |
| Hundreds of thousands of NVIDIA GPUs | Compute access planned under the original agreement | Company-stated scale, not an independently verified delivered total |
| Tens of millions of CPUs | Potential scale cited for agentic workloads | Described as capacity to scale, not a confirmed deployed count |
| $100 billion over eight years | Announced 2026 expansion of the existing agreement | Company-announced expansion |
| Approximately two gigawatts | Trainium capacity OpenAI committed to consume under the expansion | Company-announced commitment spanning Trainium3 and Trainium4 |
| 2027 | Expected start of Trainium4 delivery | Forward-looking expectation in the 2026 announcement |
| 72 Blackwell GPUs and 36 Grace CPUs | Configuration NVIDIA gave for a GB200 NVL72 system | System specification, not OpenAI deployment data |
What the deal does—and does not—tell us
The announcements show OpenAI pursuing compute across two AWS chip platforms: NVIDIA Blackwell GPUs in the original deal and AWS Trainium capacity in the expansion. They also point to AWS infrastructure being used for both model training and inference, alongside planned enterprise-facing services. The announcements do not establish realized performance, completed capacity delivery, or that the whole arrangement runs on a single chip type.
The Associated Press placed the 2025 agreement among OpenAI’s large infrastructure commitments and reported investor concerns about their scale and financing. Those concerns are reported reactions, not evidence by themselves of OpenAI’s ability or inability to meet its commitments. Amazon’s 2026 release identifies factors including energy costs, resource and supply volatility, customer demand, and other business conditions as risks to forward-looking plans.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
