SIAgentNews.com
← All news

TensorWave to provide OpenAI with 457MW of AMD-powered compute

The AI cloud provider will supply OpenAI with dedicated GPU capacity starting in 2027 as part of a broader partnership with AMD.

· First reported by The Next Web

Documentary photo: picture for the story 'TensorWave to provide OpenAI with 457MW of AMD-powered compute'
AI-made documentary photo for this story. No real people or products are shown.
457MWTotal computing capacity supplied

The short version

  • TensorWave will supply 457MW of capacity for a six-year period starting in the first half of 2027.
  • The infrastructure will run on AMD Instinct MI455X GPUs across multiple US data center sites.
  • TensorWave is a neocloud, a type of cloud company specifically built around GPUs for AI tasks.
  • The capacity will be dedicated exclusively to OpenAI with no other tenants sharing the hardware.

Why it matters

This deal signals a major shift toward diversifying the hardware used to train frontier AI models beyond just NVIDIA. It also highlights the growing importance of 'neoclouds' in providing the massive scale of power needed for next-generation AI.

A dedicated compute partnership

TensorWave has announced a significant agreement to supply OpenAI with 457 megawatts of computing capacity. This deal is set to run for a duration of six years, with the first portion of the capacity expected to go live in the first half of 2027. The infrastructure will be located across several different data center sites within the United States.

The agreement falls under an existing partnership between OpenAI and AMD. While both companies have worked together previously, TensorWave is the first neocloud provider to be publicly named under this specific arrangement. A neocloud is a cloud company that is specifically built around GPUs for AI applications rather than general-purpose computing.

Neither OpenAI nor TensorWave disclosed the specific financial value of the contract. They also did not name the specific locations of the data centers. However, the deployment is intended to serve OpenAI alone, meaning no other tenants will share the hardware or the capacity provided by the deal.

AMD hardware and infrastructure

The computing power provided by TensorWave will run on AMD’s Instinct MI455X GPUs. These chips are part of a broader hardware strategy by AMD to compete in the high-end AI market. TensorWave has already established itself as a specialist in this area, as it was the first public cloud to deploy AMD’s MI300X GPUs.

TensorWave has continued to expand its hardware portfolio to include the MI325X and MI355X models. This new deal with OpenAI represents the company's largest deployment to date. The company is designing the sites specifically for AMD’s Helios racks, which group GPUs into a single rack-scale system.

To ensure high-speed communication between these systems, Ethernet will be used to link the racks. This specialized architecture is designed to handle the intense demands of frontier AI models. The company expects to finish the full deployment of this infrastructure by early 2028.

The rise of the neocloud

TensorWave is a Las Vegas-based company that focuses exclusively on AMD hardware. Its growth has been supported by significant investment, including a $350 million Series B round raised in June. This funding came from investors such as Magnetar, AMD Ventures, and Nexus Venture Partners.

The company has positioned itself as a key player in the AI infrastructure space by securing more than 2 gigawatts of long-term data center capacity. This scale is necessary to meet the massive power requirements of modern AI training. By focusing on a specific hardware niche, TensorWave aims to provide reliable and fast delivery of advanced compute.

Piotr Tomasik, the president and co-founder of TensorWave, noted that the ability to deliver hundreds of megawatts of compute reliably will separate companies that can support frontier AI from those that cannot. This underscores the logistical challenge of building the physical infrastructure required for the next generation of artificial intelligence.

Strategic shifts for OpenAI

This deal comes at a time of organizational change for OpenAI. The company recently appointed Uday Ruddarraju as its chief technology officer for compute in October. Ruddarraju is now responsible for overseeing the company's data center and compute work, including this new partnership with TensorWave.

By partnering with TensorWave and AMD, OpenAI is diversifying its supply chain for the chips needed to train its models. While NVIDIA has long been the dominant provider in this space, this agreement highlights a move toward using alternative hardware to ensure sufficient capacity. Vamsi Boppana, AMD’s senior vice president for AI, called the deal a "landmark deployment" for the company.

The move allows OpenAI to access specialized infrastructure that is optimized for AMD's specific GPU architecture. This could provide a more stable path to scaling their models as the demand for compute power continues to rise. The partnership serves as a public validation of AMD's ability to provide high-performance hardware for the world's leading AI labs.

The road to 2028

The timeline for this project is clearly defined, with the initial capacity arriving in 2027. This gives the companies several years to coordinate the construction and setup of the data center sites. The complexity of building out hundreds of megawatts of power and cooling infrastructure is a significant undertaking that requires long-term planning.

As the deployment progresses toward its completion in early 2028, the industry will be watching how the MI455X GPUs perform in a production environment. This will be a critical test for AMD's ability to compete with established rivals in the frontier AI space. The success of this deal could influence how other large AI labs choose to source their hardware in the future.

For TensorWave, this contract solidifies its position as a major infrastructure provider. It moves the company from a niche hardware specialist to a primary partner for one of the most prominent AI companies in the world. The project serves as a blueprint for how neoclouds can facilitate the massive scaling of AI technology.

What to watch

  • How the MI455X GPUs perform compared to NVIDIA's latest offerings in real-world training.
  • Whether other major AI labs will sign similar deals with TensorWave or other neocloud providers.

Try it

Look up the difference between a general-purpose cloud provider (like AWS or Google Cloud) and a 'neocloud' to see how specialized AI hardware changes the infrastructure.

Sources and links

First reported by The Next Web. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news