TensorWave announces a six-year agreement to supply OpenAI with 457 megawatts of dedicated AI computing capacity built on AMD chips. The company says the capacity will be delivered across U.S. data centers using AMD Instinct MI455X GPUs in AMD Helios rack-scale systems, describing the pact as a “landmark” deal.

The announcement is presented as an early, public indication of OpenAI’s broader push to use AMD hardware. Several outlets frame the deal within OpenAI and AMD’s wider partnership, which includes an ambition of up to gigawatt-scale capacity. Reporting also highlights that power is a key metric in AI infrastructure contracting, reflecting data center electricity limits that constrain how much computing can be delivered.

Where details differ, sources vary mainly on how much context they provide about the technology stack and the wider AMD partnership; some focus on the overall compute commitment and timeline, while others add the specific GPU and rack system names. None of the provided summaries include pricing, the exact data center locations, or the timing for when capacity becomes available.