✉ The Friday AI Brief: the week's 5 best AI stories, tools & comparisons — in your inbox every Friday morning.

AMD logo

OpenAI Signs On for 457 Megawatts of AMD-Powered Capacity With TensorWave

OpenAI has signed a six-year agreement with neocloud provider TensorWave for 457 megawatts of dedicated AI capacity powered entirely by AMD chips — the first publicly named piece of OpenAI’s giant AMD buildout. Why it matters: OpenAI is locking in compute years ahead, and for the first time a specialist cloud, not a hyperscaler, is delivering a named slice of its AMD partnership.

November 16, 2026

What happened

TensorWave announced on November 16 that it will provide OpenAI with 457 megawatts of IT capacity across multiple U.S. data center sites. The infrastructure runs on AMD Instinct MI455X GPUs inside AMD’s Helios rack-scale systems, and it will be operated as dedicated, single-tenant capacity — OpenAI gets the machines to itself rather than sharing a public cloud. TensorWave says this is its largest deployment contract to date. The dollar value of the deal was not disclosed.

The details

The first capacity is expected to come online in the first half of 2027, with the full deployment targeted for early 2028. Helios racks combine MI455X GPUs with 6th-generation AMD Epyc “Venice” server CPUs, AMD’s ROCm software and Pensando networking in a single rack-scale design; AMD rates one rack at up to 2.9 exaflops of peak FP4 performance with 31 terabytes of HBM4 memory. TensorWave, which is backed by AMD Ventures and raised a $350 million Series B in June 2026, says it is the first neocloud publicly announced as an infrastructure provider supporting the AMD–OpenAI agreement.

That parent agreement, announced in October 2025, covers up to 6 gigawatts of AMD Instinct GPUs over multiple years. TensorWave’s 457 megawatts is one of the first concrete, named deliveries against it.

Why it matters

Frontier AI now runs on contracted power. Deals like this are measured in megawatts and years, not servers and months, and they are signed long before the capacity exists — nothing here reaches a user until 2027. The other signal is who is doing the delivering: a relatively young, AMD-only cloud provider just won a central role in OpenAI’s supply chain, which says the market for AI compute is broadening beyond Nvidia hardware and the big hyperscalers. For everyone watching ChatGPT’s capacity and pricing, this is the kind of unglamorous infrastructure deal that decides how much headroom OpenAI has two years from now. Everything AI, tested — and increasingly, everything AI, pre-booked.

Frequently asked questions

How much capacity is TensorWave providing OpenAI?

457 megawatts of IT capacity across multiple U.S. data center sites, under a six-year agreement announced November 16, 2026.

What hardware will it run on?

AMD Instinct MI455X GPUs in AMD Helios rack-scale systems, operated as dedicated, single-tenant capacity for OpenAI.

When does the capacity come online?

The first capacity is expected in the first half of 2027, with full deployment targeted for early 2028.

How much is the deal worth?

The companies did not disclose a dollar value.

Sources: TensorWave announcement (Business Wire), Data Center Dynamics

Leave a Comment

Your email address will not be published. Required fields are marked *

Get the 5 best AI tools every week

Top AI news, tools, and prompts — one short email. Free, unsubscribe anytime.

Run a newsletter of your own? Monetize and grow it with SparkLoop →

Scroll to Top