OpenAI has signed a six-year deal with TensorWave for 457 megawatts of dedicated AI capacity built on AMD chips. TensorWave, an AMD-backed cloud provider, announced the agreement on 16 November 2026. The capacity will run on AMD Instinct MI455X GPUs in AMD Helios rack-scale systems across multiple US data center sites, starting in the first half of 2027.
Quick Take
- The deal: six years, 457 MW of IT capacity, dedicated and single-tenant for OpenAI. Financial terms were not disclosed.
- The hardware: AMD Instinct MI455X GPUs in AMD Helios racks, with Ethernet scale-out networking between racks.
- Timeline: first capacity in H1 2027, full deployment expected by early 2028.
- Why it’s notable: TensorWave is the first neocloud publicly named as a provider under the existing OpenAI and AMD strategic partnership.
- TensorWave’s footprint: the company says it has secured more than 2 gigawatts of long-term data center capacity and calls this its largest deployment to date.
In plain English
OpenAI needs far more computing power than it can build itself. Instead of buying all the hardware directly, it is renting a huge, dedicated slice of a specialist cloud company that only uses AMD chips. The deal is measured in power, 457 megawatts, because electricity is now the main limit on how many AI chips a data center can run.
What happened
TensorWave said the agreement is part of “OpenAI and AMD’s previously announced strategic partnership.” The sites are being designed specifically for AMD Helios, and the deployment takes TensorWave from its existing large AMD clusters to “the hundreds-of-megawatts level required by frontier AI developers.”
“As demand for AI continues to grow, expanding access to reliable, large-scale compute is critical to our ability to build and deliver increasingly capable models,” said Uday Ruddarraju, OpenAI’s chief technology officer for compute. Vamsi Bopanna, senior vice president for AI at AMD, called it “this landmark deployment for OpenAI.”
TensorWave president and co-founder Piotr Tomasik said the company “was built to prove that AMD infrastructure could operate at the scale and reliability that the world’s most demanding AI companies require.”
Background
OpenAI and AMD announced a multi-year partnership in 2025 covering 6 gigawatts of AMD GPUs; we covered the terms in our analysis of the OpenAI and AMD 6GW deal. Until now, it was not public which operators would host that capacity. TensorWave, backed by AMD Ventures, Magnetar and Nexus Venture Partners, was the first public cloud to deploy AMD Instinct MI300X GPUs and raised a Series B in June 2026.
The announcement comes the same day AMD said it had completed its World Labs acquisition, another sign of how hard AMD is pushing to be a full-stack alternative to NVIDIA.
Why it matters
For OpenAI, a second chip supplier at scale reduces dependence on NVIDIA and gives it leverage on price and supply. Dedicated, single-tenant capacity also means OpenAI does not share these clusters with other customers.
For AMD, a named neocloud building for Helios is a proof point that its rack-scale systems are moving from roadmap to real deployments.
For the neocloud market, the deal shows a model that CoreWeave pioneered with NVIDIA: specialist clouds raise capital, sign long contracts with a single frontier lab, and build around one vendor’s hardware. The risk is concentration, since a large part of TensorWave’s growth now depends on one customer.
What’s still open
- The contract value and how it is financed.
- Which US sites will host the capacity and how they will be powered.
- How many more neoclouds will be named under the OpenAI and AMD partnership.
Sources
- TensorWave press release, “TensorWave Announces Agreement with OpenAI to Deliver More Than 450MW of AMD Powered AI Infrastructure,” 16 November 2026
- TensorWave blog, “TensorWave to Provide OpenAI with 457 Megawatts of AMD-Powered AI Infrastructure,” 16 November 2026
i10x Perspective
The AMD-OpenAI deal was always going to need landlords, and TensorWave is the first to step forward. For OpenAI, this is diversification with a long lease. For AMD, it is the clearest sign yet that Helios will run frontier workloads, not just benchmarks. We will add the deal to our tracking in the AI Capex Tracker once terms are clearer.






