TensorWave Signs Six-Year OpenAI Deal for 457MW AMD Capacity
TensorWave will deliver 457 megawatts of dedicated AMD Instinct MI455X capacity to OpenAI across U.S. sites starting in 2027, under a six-year agreement tied to AMD’s 6-gigawatt partnership.
By Daniel Okafor
3 min read

What's News
- TensorWave announced the six-year, 457-megawatt OpenAI agreement on November 16, 2026.
- First capacity is expected in the first half of 2027, with full deployment by early 2028.
- The build uses AMD Instinct MI455X GPUs and Helios systems across multiple U.S. data center sites.
- The deal extends an AMD-OpenAI partnership announced October 6, 2025 that covers 6 gigawatts of AMD Instinct GPUs.
- TensorWave reports more than 2 gigawatts of secured long-term data-center capacity.
By Daniel Okafor
TensorWave will supply OpenAI with 457 megawatts of dedicated AI infrastructure under a six-year agreement announced on November 16, 2026. The deployment will use AMD Instinct MI455X GPUs and AMD Helios rack-scale systems across multiple U.S. data center sites.
What does the OpenAI agreement cover?
TensorWave stated the deployment schedule and operating terms:
- First capacity online: The first half of 2027.
- Full deployment: Early 2028.
- Capacity model: Dedicated, single-tenant space reserved for OpenAI workloads, rather than shared cloud capacity.
- Intended uses: Large-scale training and inference.
TensorWave called the agreement its largest deployment to date. It also described the contract as a significant expansion of its U.S. infrastructure footprint. It applies TensorWave’s experience operating large-scale AMD GPU clusters to a purpose-built system serving one customer’s next-generation AI workloads throughout the six-year agreement.
How will the infrastructure be built?
TensorWave said AMD Helios rack-scale systems will sit at the center of the deployment. High-performance Ethernet will connect the racks for large-scale training and inference. The build links rack-scale compute while requiring coordinated management of GPU systems, networking, power, cooling and operations across several locations.
Piotr Tomasik, TensorWave’s president and co-founder, said: “TensorWave was built to prove that AMD infrastructure could operate at the scale and reliability that the world’s most demanding AI companies require.” His statement puts operating reliability at the center of the deal.
Where does it fit in the AMD-OpenAI partnership?
TensorWave said the contract builds on the AMD-OpenAI strategic partnership announced on October 6, 2025. The multi-year, multi-generation arrangement covers 6 gigawatts of AMD Instinct GPUs. Its initial 1-gigawatt deployment of MI450-series GPUs was scheduled to begin in the second half of 2026.
TensorWave said AMD and OpenAI also agreed to share technical expertise to optimize their product roadmaps. Their hardware and software collaboration began with MI300X and continued with MI350X. TensorWave said it is the first publicly announced specialist AI-cloud provider supporting that collaboration. The deal makes TensorWave the first named specialist AI-cloud operator in OpenAI’s AMD buildout.
The arrangement adds an infrastructure operator to a partnership previously framed around the chip supplier and model developer.
What operating base does TensorWave bring?
TensorWave said it was the first public cloud to deploy AMD Instinct MI300X GPUs. It later expanded across MI325X and MI355X hardware. The company now operates one of the world’s largest AMD-exclusive AI clouds. TensorWave also operates one of North America’s largest AMD-based AI training clusters. It is an AMD Helios launch partner.
The OpenAI project extends that earlier GPU work into purpose-built Helios systems. TensorWave reports more than 2 gigawatts of secured long-term data-center capacity, which the company says provides a broader base for U.S. expansion. For scale, the OpenAI contract represents 457 megawatts of dedicated capacity. Execution across multiple sites is now the immediate measure of that pipeline.
What must TensorWave deliver?
For OpenAI, the arrangement adds a dedicated delivery channel for AMD compute alongside its direct strategic work with the chip supplier. For TensorWave, the agreement converts secured capacity and operating experience into a multi-site delivery obligation on a fixed timetable. The operator must deliver reliable single-tenant infrastructure across sites supporting both training and inference.
The first checkpoint arrives in the first half of 2027, when TensorWave expects initial capacity online. Full completion by early 2028 will test whether the operator can scale the deployment on schedule. OpenAI’s dedicated capacity depends on TensorWave clearing both deadlines.
Source: unite.ai
More from Daniel Okafor
Show full bio
Correspondent covering money, funding and market shifts around entrepreneurship, startups and venture funding at Business Bearings.
2,189 articles