OpenAI Contracts TensorWave for 457 MW of AMD Compute in Nvidia Diversification Push
The six-year deal, the first publicly named neocloud partnership under OpenAI's AMD buildout, puts AMD's new Helios rack-scale systems to their biggest commercial test yet.
OpenAI has signed a six-year contract with TensorWave, an AMD-focused cloud provider, to supply 457 megawatts of dedicated AI compute across multiple U.S. data center sites. TensorWave announced the deal on November 16, and according to reporting reviewed by Data Center Dynamics, it's the company's largest single deployment contract to date.
The hardware underneath it is specific: AMD Instinct MI455X GPUs inside AMD's Helios rack-scale systems, with Ethernet as the scale-out network fabric. That last detail matters. Nvidia's dominant rack architecture leans on InfiniBand and NVLink for inter-GPU connectivity. AMD is betting that Ethernet-based scale-out can close the gap, and the Helios system is the production vehicle for that argument. According to Data Center Dynamics, AMD claims the Helios rack delivers 15 percent more peak FP4 performance than Nvidia's comparable Vera Rubin NVL72, plus 50 percent more high-bandwidth memory capacity. Those are AMD's own numbers against published Nvidia specs, so take them with appropriate skepticism, but they signal where AMD thinks the competitive fight is.
The TensorWave agreement slots into the broader AMD-OpenAI strategic partnership announced in October 2025, which outlined plans for up to 6 gigawatts of AMD GPU deployments over multiple years. According to reporting by The Next Web, TensorWave is the first neocloud publicly named as a delivery partner under that framework. The 457 MW here is roughly 7.6 percent of that 6 GW ceiling, so this is one piece of a much larger infrastructure commitment, not the whole thing. Whether TensorWave's capacity counts toward the 6 GW figure hasn't been spelled out publicly, as FourWeekMBA noted in its coverage of the announcement.
TensorWave itself is a Las Vegas-based neocloud backed by AMD Ventures, which closed a Series B round in June. According to reporting by Futunn, the company has secured 2 gigawatts of long-term U.S. data center capacity, and the OpenAI contract is designed to run on a single-tenant, exclusive basis. First capacity is expected in the first half of 2027.
The structural story here isn't the megawatt number. It's the supply chain architecture OpenAI is building. For years, Nvidia GPUs have been the default substrate for frontier model training and inference, and OpenAI is one of Nvidia's largest customers. The AMD partnership, now with a named operator and a timetable, is an operational hedge. If TensorWave can bring Helios capacity online reliably across several sites, OpenAI gets negotiating leverage it currently lacks.
For AMD, the stakes are just as direct. Nvidia still dominates rented AI compute. According to technologies.org's coverage of the deal, every dedicated MI455X hall that goes live for a customer of OpenAI's scale becomes a reference other buyers can point to. That's how you shift a market, not by winning a benchmark but by proving the ops story at scale.
What to watch: TensorWave's execution between now and that first-half 2027 delivery window. The technical specs are plausible. The network architecture is coherent. But a 457 MW multi-site buildout for the world's most high-profile AI lab is a different operational category than anything TensorWave has run before.
Sources cited:
- Data Center Dynamics (https://www.datacenterdynamics.com/en/news/openai-signs-on-for-457mw-ai-capacity-with-tensorwave/)
- The Next Web (https://thenextweb.com/news/tensorwave-openai-457mw-amd-helios)
- FourWeekMBA (https://fourweekmba.com/ai-openai-signs-457-mw-amd-gpu-deal-with-neocloud-tensorwave/)
- Futunn News (https://news.futunn.com/en/post/1004636255/openai-expands-amd-compute-footprint-signs-six-year-agreement-with)
- technologies.org (https://technologies.org/tensorwave-signs-six-year-deal-to-supply-openai-with-457mw-of-amd-helios-capacity/)
- TensorWave (primary announcement) (https://tensorwave.com/blog/tensorwave-to-provide-openai-with-457-megawatts-of-amd-powered-ai-infrastructure)
This release was originally distributed via ETL Newswire. Visit Data Center Dynamics for the full story, related releases, and contact information.
Visit Data Center Dynamics →