OpenAI Hands TensorWave A 457MW Deal To Run On AMD Chips
Little-known cloud firm TensorWave has landed a six-year deal to supply OpenAI with 457 megawatts of AI computing power built on AMD chips, the first named partner in OpenAI’s giant bet on AMD.
Key Takeaways
- TensorWave will deliver 457 megawatts of dedicated AI infrastructure for OpenAI across U.S. data centers
- The deal runs for six years and uses AMD Instinct MI455X GPUs in AMD Helios rack-scale systems
- TensorWave says it is the first publicly announced neocloud supporting OpenAI and AMD’s strategic partnership
- The company has not shared the price, the data center sites or when the capacity comes online
TensorWave announced the agreement on Monday, calling it “a landmark six-year agreement” and saying 457 megawatts “is just the beginning” for the company.
Why 457 Megawatts Matters
Megawatts are how the AI industry now measures computing deals, because the chips are limited by how much power a data center can feed them. A 457-megawatt commitment means hundreds of thousands of homes’ worth of electricity going into one customer’s AI work.
All of it will run on AMD’s newest Instinct MI455X graphics chips, packed into Helios racks that AMD sells as complete systems. That makes the deal a big test of whether AMD hardware can carry OpenAI’s work at a scale long dominated by Nvidia.
“TensorWave was built to prove that AMD infrastructure could operate at the scale and reliability that the world’s most demanding AI companies require,” said Piotr Tomasik, TensorWave’s president and co-founder.
Also Read: Why Nvidia And AMD Fight Over Inference As Data Center Hits $89B
The Bigger AMD Bet
OpenAI and AMD struck a strategic partnership in October 2025 for up to 6 gigawatts of AMD chips. TensorWave is the first neocloud, a cloud company built just for AI, to say publicly that it is helping deliver that capacity.
“AI infrastructure is entering a new phase, where the ability to deliver hundreds of megawatts of advanced compute reliably and at speed will separate the companies that can support frontier AI from those that cannot,” Tomasik said.
What To Watch Next
TensorWave has not said which data centers will host the chips, when the first racks switch on or what OpenAI is paying. Those details will show how fast AMD’s newest chips can actually reach OpenAI.
The bigger question is who comes next. If more neoclouds sign on to OpenAI’s AMD plan, it would show that the AI boom can run on more than one chipmaker, a shift that would matter for every company paying for AI computing.
Read Next: Forget GPUs, A Memory Chip Squeeze Could Choke AI Even After OpenAI’s $250B Splurge
