Subscribe
Datacenters

TensorWave to supply OpenAI with 457 MW of AMD-powered AI infrastructure

Wednesday, November 18, 2026 · 6 min read

TensorWave will provide OpenAI with 457 megawatts of dedicated AI infrastructure powered by AMD Instinct MI455X GPUs, in the first publicly announced neocloud deal backing the OpenAI-AMD partnership.

Bullish$AMD

📊 Scorecard: our Bullish call on $AMD is scored from $503.41, the last close before it ran. Its first result comes after the next close.

TensorWave to supply OpenAI with 457 MW of AMD-powered AI infrastructure

The News

TensorWave announced on November 16 a six-year agreement to provide OpenAI with 457 megawatts of dedicated AI infrastructure capacity across multiple datacenter sites in the United States, powered by AMD Instinct MI455X graphics processors in AMD Helios rack-scale systems. The capacity will be single-tenant, meaning OpenAI gets the whole system to itself rather than sharing it with other customers.

The first capacity is due to come online in the first half of 2027, with full deployment by early 2028. It is the largest deployment contract TensorWave has signed to date, and the first publicly announced deal by a neocloud — a smaller cloud provider built for AI — supporting the OpenAI-AMD strategic partnership announced in October 2025, which covers 6 gigawatts of AMD Instinct GPUs.

The value of the deal, the site locations and the financing were not disclosed. TensorWave says it has secured more than 2 gigawatts of long-term datacenter capacity and raised a $350 million Series B round in June 2026. Separately, TensorWave's previously announced 222-megawatt lease at Fermi America's Project Matador in Texas was terminated on November 10, Fermi confirmed in its third-quarter results and 10-Q filed November 12; the AMD credit-support agreement tied to the lease was terminated with it.

OpenAI is buying its compute from more suppliers, and AMD just became one of its biggest routes in.

Why It Matters

Four hundred fifty-seven megawatts is a serious amount of computing power. At roughly a kilowatt-plus per accelerator plus overhead, 457 megawatts represents tens of thousands of AMD's newest chips working for OpenAI — a meaningful slice of the 6-gigawatt partnership the two companies announced last year, and a real test of AMD's hardware at frontier scale.

The deal also shows how OpenAI is spreading its bets. Instead of relying on a single cloud provider, it is buying capacity from specialist neoclouds like TensorWave alongside the big hyperscalers. That diversifies supply, gives OpenAI more negotiating leverage, and gives AMD a second route into OpenAI's infrastructure beyond its direct partnership. It also spreads risk: if one supplier has an outage or a delay, OpenAI can shift work elsewhere rather than waiting.

For AMD, the read-through is the point. AMD is not a party to this deal — TensorWave is buying the chips and selling the capacity — but every megawatt of AMD-powered infrastructure that OpenAI rents is a megawatt that did not go to Nvidia. If TensorWave delivers on time and the systems run well, it becomes a proof point that AMD's rack-scale systems can run frontier AI workloads.

Analyst's Take

Our call is Bullish. A 457-megawatt, six-year commitment from OpenAI is a vote of confidence in both TensorWave and AMD's hardware, and the first-half-2027 start date means revenue and proof points arrive quickly rather than in some distant future. TensorWave has already secured more than 2 gigawatts of long-term capacity, so this deal is a large but manageable share of what it can deliver. For AMD, every megawatt of its chips running OpenAI workloads is a megawatt that did not go to Nvidia.

The risks are execution and disclosure. The deal's value, locations and financing are all undisclosed, so the market cannot judge how profitable it is for TensorWave or how much of the money reaches AMD. The termination of TensorWave's 222-megawatt Texas lease, confirmed by Fermi in its November 12 filing, is also a reminder that announced capacity and delivered capacity are not the same thing.

Still, the direction is clear: OpenAI wants more compute from more suppliers, and AMD is now firmly in that mix. If the Helios rack systems perform well in OpenAI's hands, expect more neocloud deals like this one — and expect Nvidia's rivals to keep chipping away at its lead. Reader poll: Can AMD-powered clouds challenge Nvidia? Vote: AMD gains ground, Nvidia stays ahead, Too early to tell.

Are you bullish or bearish on this story?

By The Numbers

457 MWCapacity for OpenAI
6 yearsAgreement term
H1 2027First capacity online
6 GWOpenAI-AMD partnership size
$350MTensorWave Series B, June 2026
$503.41AMD close, Tue Nov 17 (-0.17%)

FAQ

What is a neocloud?

A neocloud is a smaller, newer cloud provider built specifically for AI work, rather than a general-purpose giant like Amazon or Microsoft. Neoclouds rent out large clusters of graphics processors to AI companies that need computing power fast.

What is a GPU?

A GPU, or graphics processing unit, is a computer chip originally designed for video games that turned out to be very good at the maths behind AI. Training and running AI models needs thousands of GPUs working together in datacenters.

What does single-tenant mean?

Single-tenant means one customer gets the whole system to itself, rather than sharing it with others. OpenAI will be the only user of this 457 megawatts of capacity, which gives it control over security and performance.

👁

Source: Business Wire

Get the daily briefing. AI news with a markets lens — tickers, bullish/bearish reads, zero hype. Subscribe free on Substack

More on $AMD

All $AMD coverage →

More from AI Agent Update