OpenAI Signs 457 MW AMD GPU Deal With Neocloud TensorWave

TensorWave has signed a six-year agreement with OpenAI to deliver 457 megawatts of dedicated AI infrastructure across multiple US data center sites, the AMD-focused cloud provider said on 16 November. It will run AMD Instinct MI455X GPUs in systems built for AMD’s Helios racks, with first capacity expected in the first half of 2027.

457 megawatts equals 7.6% of the 6 gigawatts that AMD and OpenAI agreed in October 2025, by our arithmetic. TensorWave says the deal builds on that partnership but does not say whether its capacity counts toward the 6 gigawatts. No contract value was disclosed.

Business Pill · GPUs explained

An evergreen Business Engineer explainer on GPUs. It teaches the general idea only and does not discuss TensorWave, OpenAI or AMD.

The key insight: As we read it, the release puts a size and a schedule on one supplier inside the AMD-OpenAI plan: 457 megawatts, first capacity in the first half of 2027, full deployment by early 2028, over six years. The price, the financing and whether it counts toward the 6 gigawatts are the parts it leaves out.

What TensorWave Announced

TensorWave calls the agreement its largest deployment to date. It says the capacity will be “dedicated, single-tenant” and will connect AMD Helios racks through Ethernet networking for both training and inference.

The company expects to bring the first capacity online in the first half of 2027 and to complete the full deployment by early 2028. The release does not name the data center sites, the power suppliers or the financing.

TensorWave describes itself as “the first publicly announced neocloud infrastructure provider supporting the collaboration” between OpenAI and AMD. A neocloud is a specialist GPU cloud that rents compute rather than selling software.

Megawatts of AI compute capacity stated by AMD (6 October 2025) and TensorWave (16 November 2026). 457 / 6,000
Megawatts of AI compute capacity stated by AMD (6 October 2025) and TensorWave (16 November 2026). 457 / 6,000 = 7.6% is our arithmetic; TensorWave does not say whether its 457 MW counts toward the 6 GW. Sources: TensorWave and AMD releases.

How It Fits the AMD-OpenAI Agreement

In AMD’s October 2025 release, the two companies announced “a 6 gigawatt agreement to power OpenAI’s next-generation AI infrastructure across multiple generations of AMD Instinct GPUs”. The first 1 gigawatt, of MI450 GPUs, was set to begin in the second half of 2026.

AMD also issued OpenAI a warrant for up to 160 million AMD shares. Per the release, the first tranche vests with the initial 1 gigawatt deployment and further tranches vest as purchases scale up to 6 gigawatts.

TensorWave’s 457 megawatts is 45.7% of the size of that first 1 gigawatt step, by our arithmetic. Its schedule starts later: the first half of 2027, against the second half of 2026 for AMD’s first deployment.

What AMD and TensorWave Said

Vamsi Boppana, AMD’s senior vice president for AI, said in the release that TensorWave “has been a strong partner in scaling AMD Instinct infrastructure”. OpenAI is not quoted in TensorWave’s release.

Piotr Tomasik, TensorWave’s president and co-founder, 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”.

TensorWave says it has more than 2 gigawatts of long-term data center capacity secured. On that figure, the OpenAI deal is under 23% of it, by our arithmetic.

How It Compares With Other Compute Deals

AMD’s data center unit is still a fraction of Nvidia’s, as our AMD data center comparison shows. Our read of short GPU deal pricing sets out what neoclouds charge per megawatt by contract length.

Financing is the other half of these builds. Our Blackstone leaseback piece covers a 647 MW deal funded half by debt, and our AI compute contracts tracker lists who rents compute from whom.

The Structural Read

As we read the two releases, OpenAI is getting AMD capacity through a neocloud as well as through the direct AMD agreement; TensorWave says it is the first publicly announced neocloud supporting that collaboration.

As we read the schedule, TensorWave’s first capacity is due in the first half of 2027, while AMD said its first 1 gigawatt for OpenAI was set to begin in the second half of 2026.

Piotr Tomasik, president and co-founder, TensorWave

“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.”

Three Implications

FOR AMD Its MI455X GPUs and Helios racks are the hardware named for the 457 megawatts, per TensorWave’s release.

FOR OPENAI The capacity is dedicated and single-tenant, for training and inference, per the release; OpenAI is not quoted in it.

FOR NEOCLOUDS TensorWave says it has more than 2 gigawatts of long-term data center capacity secured; this deal is its largest to date, it says.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Capex Map.

The framework sets out to map the financing of AI: “How the money flows, where it cracks, what it buys, who can absorb it, and what happens when revenue can’t catch up.”

As we read it, TensorWave sits in the neocloud layer of that map, between the chip maker and the model lab, and its release gives no contract value or financing. The framework was published before this deal and does not discuss it.

What Is Not Established

TensorWave did not disclose the contract value, the price per megawatt, the site locations or the financing. It is not stated whether the 457 megawatts counts toward the 6 gigawatts in the AMD-OpenAI agreement, or when the six-year term starts. The release carries no statement from OpenAI. We did not contact TensorWave, OpenAI or AMD for this piece.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The AI Capex Map

A Business Engineer framework on how the AI capex cycle is financed, where it cracks and who can absorb it.

Read the Map of AI →

The Bottom Line

TensorWave says it will deliver 457 megawatts of AMD MI455X capacity to OpenAI under a six-year agreement, with first capacity expected in the first half of 2027. That equals 7.6% of the 6 gigawatts in the AMD-OpenAI agreement, by our arithmetic, with no contract value disclosed.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. We read TensorWave’s release of 16 November 2026 on its website in full and AMD’s 6 October 2025 release on the AMD-OpenAI agreement. The story was flagged by a Wall St Engine post on X. Megawatt figures are as stated by the companies; the percentages are our arithmetic. We did not contact TensorWave, OpenAI or AMD for this piece.

Sources: TensorWave: TensorWave to Provide OpenAI with 457 Megawatts of AMD-Powered AI Infrastructure (16 Nov 2026) · AMD: AMD and OpenAI Announce Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs (6 Oct 2025) · Wall St Engine on X (16 Nov 2026)

Scroll to Top
FourWeekMBA