TensorWave commits 457MW of AMD GPU power to OpenAI for six years

What happened
TensorWave, a cloud provider specializing in AI infrastructure, has signed a six-year agreement to supply OpenAI with 457 megawatts of dedicated AMD-powered computing capacity. The deployment will use AMD Instinct MI455X GPUs and Helios rack-scale systems across multiple data centers in the United States. The project is set to begin delivering capacity in early 2027, with full deployment expected by early 2028. This marks TensorWave's largest single deployment and is designed exclusively for OpenAI's use.
Key details
- 457 megawatts of AMD MI455X GPU capacity will be delivered to OpenAI.
- The contract spans six years, starting in early 2027 and completing by early 2028.
- Deployment will occur across several U.S. data center sites.
- TensorWave will provide single-tenant, dedicated infrastructure for OpenAI, not shared cloud resources.
- AMD's Helios rack-scale systems will be used for large-scale training and inference workloads.
Why it matters
OpenAI will have access to a large, dedicated pool of AMD-powered compute resources, allowing it to run training and inference for its AI models without sharing capacity with other customers. For TensorWave, this contract represents a significant expansion of its operations and a test of its ability to deliver large-scale, single-customer infrastructure on a strict timeline. The agreement also adds a new infrastructure provider to OpenAI's AMD-based buildout, giving OpenAI more options for scaling its AI workloads.

