Business

OpenAI advances AMD partnership with reported TensorWave capacity deal

Fortune reports TensorWave will supply 457 megawatts for OpenAI, an early part of its planned 6-gigawatt AMD buildout.

Daniel Okafor

By Daniel Okafor · Business Editor

2 min read

OpenAI advances AMD partnership with reported TensorWave capacity deal
Photo: Fortune

OpenAI has identified TensorWave as the first cloud provider tied to its AMD agreement, with 457 megawatts of planned capacity across U.S. data-center sites, according to Fortune. The reported OpenAI TensorWave AMD partnership adds capacity from another chip supplier, but it does not show that OpenAI is replacing Nvidia.

TensorWave’s planned 457 megawatts equal 0.457 gigawatts, or about 7.6% of the 6-gigawatt AMD program, Fortune reported. That makes the deployment an initial segment of the broader announced AMD plan rather than its full scope.

How large is the TensorWave deployment in OpenAI’s AMD plan?

Fortune said TensorWave expects the capacity to come online in the first half of 2027. The company told Fortune that the work would be its largest deployment to date and would require it to roughly triple its staff.

AMD Chief Executive Lisa Su described the wider arrangement in an October 2025 Bloomberg Tech interview as a planned 6-gigawatt AI-computing buildout for OpenAI. Su said the first gigawatt was planned for the second half of 2026 using AMD’s next-generation MI450 chip.

OpenAI President Greg Brockman said in that interview that AMD systems would be installed in OpenAI data centers and through cloud providers. He said the scale of the effort would require multiple locations and providers.

Those plans sit alongside a separate Nvidia agreement. Nvidia and OpenAI said in September 2025 that they had signed a letter of intent for OpenAI to deploy at least 10 gigawatts of Nvidia systems. Nvidia said the first phase was targeted for the second half of 2026 on its Vera Rubin platform, and described OpenAI as a preferred strategic compute and networking partner.

Nvidia also said its proposed investment, deployment timetable and other expected results were forward-looking statements subject to risks and uncertainties. The AMD schedule and TensorWave deployment are likewise planned capacity, not evidence of hardware already in service.

The announced programs point to OpenAI seeking additional computing supply from AMD while maintaining a substantial planned relationship with Nvidia. The available reports do not establish how much AMD capacity OpenAI has deployed compared with Nvidia capacity, or quantify a current reduction in OpenAI’s use of Nvidia hardware.

This story draws on original reporting from Fortune.