Skip to content
Advertisement
OpenAI says its models helped AMD compress a year of chip design into five months

OpenAI says its models helped AMD compress a year of chip design into five months

FoxTPNL / Wikimedia Commons (CC BY 4.0)

Greg Brockman's claim adds a new layer to a multi-year partnership built on 6 gigawatts of AMD compute and a potential 10% equity stake

Greg Brockman, OpenAI’s co-founder, says the company’s AI models helped AMD squeeze a year’s worth of chip-design improvements into five months.

That is a bold productivity claim. It also lands in the middle of one of the largest hardware partnerships in the AI industry, which makes it more than a boast about software.

What Brockman is claiming

The claim, as Brockman framed it, is about speed. OpenAI’s models allegedly let AMD’s chip-design work advance on a timeline that would normally take twelve months, finished in just five.

Brockman did not attach a public breakdown of which design stages sped up, or by how much each one improved. For now, the figure rests on his word as an OpenAI executive. AMD’s own engineers would be the ones to confirm it.

The partnership behind the claim

The speed-up did not happen in a vacuum. OpenAI and AMD announced a multi-year strategic partnership on October 6, 2025, centered on deploying AMD Instinct GPUs across OpenAI’s infrastructure.

The scale is enormous. OpenAI committed to 6 gigawatts of compute capacity spread across multiple generations of AMD GPUs.

Advertisement

The first phase covers 1 gigawatt of MI450 series GPUs. That capacity is expected to come online in the second half of 2026.

The financial terms are just as notable. OpenAI secured warrants for up to 160 million AMD shares, which works out to nearly a 10% ownership stake.

Those warrants do not vest all at once. They unlock as specific deployment milestones are hit. In other words, OpenAI’s equity grows as AMD’s hardware actually shows up and runs.

Co-design, not just procurement

The partnership is described as a full-stack co-design arrangement. OpenAI shares detailed insights about its workloads, and those insights help shape AMD’s future product roadmaps.

In return, AMD delivers rack-scale systems such as Helios built for OpenAI’s needs.

The research behind the partnership also points to AI being used to optimize custom chip designs more broadly, including work with companies like Broadcom. Those efforts have reportedly produced significant time reductions as well.

More capacity on the way

The relationship has kept expanding. A follow-up agreement dated November 16, 2026, gives OpenAI an additional 457 megawatts of dedicated compute capacity from TensorWave.

That capacity runs on AMD MI455X GPUs. The first wave is expected to arrive in the first half of 2027.

What this means

There are caveats worth keeping in view. The five-month figure comes from OpenAI, a partner with a direct financial interest in AMD’s success. An independent account from AMD of exactly what improved would carry more weight.

The warrant structure also cuts both ways. OpenAI’s potential stake depends on milestones being met. Delays in getting the MI450 capacity online in the second half of 2026 would slow the vesting and test the partnership’s momentum.

The things to watch are concrete. First, whether the 1 gigawatt MI450 deployment arrives in the second half of 2026 as expected. Second, whether AMD itself confirms or quantifies the design acceleration Brockman described. Third, whether the TensorWave capacity lands in the first half of 2027.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.