OpenAI Signs 457MW Deal With AMD-Powered Cloud Provider TensorWave

CIO Africa

Source published: Site published:

Original article · فارسی

ByStaff Writer
3 min read
TensorWave will build 457 megawatts of AMD-powered computing capacity for OpenAI across the United States by 2028. This is a major step for a company using chips other than Nvidia's, announced just before Nvidia releases its financial results.
Photo courtesy: Storagereview

TensorWave, a cloud company based in Las Vegas, has signed a six-year deal to build 457 megawatts of computing power for artificial intelligence (AI) for OpenAI. The equipment will be spread across several data centre sites in the United States. It will use only AMD’s new Instinct MI455X graphics processors. The companies announced the deal on Monday, one day before Nvidia was due to share its quarterly financial results.

TensorWave has chosen a different path from most companies in the AI cloud industry. Most providers build their systems around Nvidia chips. But TensorWave chose to use only processors made by Nvidia’s competitor, AMD. This choice has now put the company in a position to provide computing power to one of the world's largest AI developers.

All 457 megawatts of computing power will be set aside for OpenAI. This means OpenAI will have its own access to the systems instead of sharing them with other customers. The systems will use AMD’s Helios technology, which connects many computer parts in large racks. Neither company has shared how much the deal is worth or where the data centres will be built.

The size of the deal is easier to understand when compared with another project. Airtel Africa is building a data centre at Tatu City, outside Nairobi. It is expected to be the largest data centre in East Africa, with a planned capacity of 44 megawatts. TensorWave's project for OpenAI is more than ten times as large.

This is TensorWave's biggest project so far. Its president and co-founder, Piotr Tomasik, said the 457-megawatt deal shows that AMD technology can meet the computing power and reliability needs of large AI companies.

The deal is also part of a much bigger agreement between OpenAI and AMD: their six-gigawatt strategic partnership. TensorWave says it is the first specialist cloud provider publicly announced as supporting this partnership. These specialist providers, also called neoclouds, focus on supplying the computing power needed to build and run AI systems.

For OpenAI, the deal gives it more access to AMD-powered computing systems as demand for AI services grows. “As demand for AI continues to grow, expanding access to reliable, large-scale compute is critical,” said Uday Ruddarraju, OpenAI’s chief technology officer for compute.

AMD also has a financial interest in TensorWave's growth. Its investment arm, AMD Ventures, helped lead TensorWave's $350 million Series B funding round in June. Investment firm Magnetar also helped lead the round. The funding round valued TensorWave at $1.55 billion.

TensorWave had already built a large AMD-powered computing operation before signing the OpenAI deal. Its existing systems include a group of computers used to train AI, with 8,192 Instinct MI325X processors. TensorWave also says it has secured more than two gigawatts of long-term data centre capacity. Its customers include Fireworks AI and Luma AI.

Vamsi Boppana, AMD’s senior vice-president for AI, called TensorWave “a strong partner in scaling AMD Instinct infrastructure.” The deal comes as AMD tries to win a bigger share of a market where Nvidia has long been the leading supplier of AI chips.

The timing highlights the competition between the two companies. Nvidia is due to report its financial results for the third quarter of its financial year on Tuesday. This makes the TensorWave announcement another important development to watch as competition for AI computing power grows.

The first computing systems covered by the deal are expected to start working in the first half of 2027. The full project is planned to be finished by early 2028.