OpenAI has signed a new deal with AMD as it looks to diversify its sourcing of next-generation AI chips.
The deal will see OpenAI take 6GW worth of AMD’s Instinct GPUs, which will span multiple generations and help power its future AI infrastructure. It plans to take the first 1GW in H2 2026, with the chips all set to be built on AMD’s Instinct MI450.
Under the agreement between the two companies, AMD becomes a core strategic compute partner, supplying rack-scale AI systems and future GPU generations through a co-optimised hardware and software roadmap. The companies say the collaboration deepens an existing relationship that started with MI300X and continued with the MI350X series.
Additionally, as part of the deal, AMD has issued OpenAI a warrant for up to 160 million AMD common shares. The warrant vests in tranches against specific milestones, including the initial 1GW deployment and subsequent purchases up to 6GW, and is also tied to AMD share-price targets and OpenAI’s technical and commercial milestones required to scale AMD deployments. The exercise price is $0.01. That could give OpenAI a 10% stake in the chipmaker.
“We are thrilled to partner with OpenAI to deliver AI compute at massive scale,” said Dr. Lisa Su, chair and CEO, AMD.
“This partnership brings the best of AMD and OpenAI together to create a true win-win enabling the world’s most ambitious AI buildout and advancing the entire AI ecosystem.”
“This partnership is a major step in building the compute capacity needed to realize AI’s full potential,” added Sam Altman, co-founder and CEO of OpenAI.
“AMD’s leadership in high-performance chips will enable us to accelerate progress and bring the benefits of advanced AI to everyone faster.”
OpenAI courts both chip giants
The agreement between OpenAI and AMD comes hot on the heels of the AI giant’s strategic partnership with Nvidia, which it announced last month. That will see OpenAI deploy at least 10GW of Nvidia systems for its next-generation AI infrastructure, with the initial phase of that deployment set to come online in the second half of 2026 on Nvidia’s Vera Rubin platform.
That deal also comes with a hefty investment from Nvidia, which is committing up to $100 billion to support the construction of OpenAI’s next-generation AI data centres. The company is also set to receive non-voting shares in the maker of ChatGPT as part of the deal.
It’s unsurprising to see OpenAI acquire chips from both AMD and Nvidia, after all the company is looking to grow its compute capacity exponentially. With that explosive growth obviously comes constraints on when it will actually receive chips, especially as companies such as Google, Meta and Microsoft will also all be clamouring to get their hands on the latest and greatest AI chips.
In fact, the deals signed with AMD and Nvidia don’t even get OpenAI to the capacity it is targeting – which is estimated to be around 23GW of new compute. That’s why the company has also signed a deal with Oracle to lease 4.5 GW of computing power at a cost of $30 billion.
All of these agreements mean one thing, however – a very large bill for OpenAI. In fact, the company estimates that to get to its targeted capacity it will likely need to spend over $1 trillion. That’s a lot of cash for a company that is still making a loss and achieving annual revenues of just $13 billion.
