AMD to power OpenAI with $10B+ chip deal for 6GW compute
▼ Summary
– AMD has signed a multi-year chip supply deal with OpenAI to provide 6 gigawatts of compute capacity across multiple generations of its Instinct GPUs, starting with the MI450 series.
– The first gigawatt of compute capacity will be delivered in the second half of 2026, with AMD claiming the MI450 series will outperform Nvidia’s comparable offerings through hardware and software improvements.
– As part of the agreement, AMD has given OpenAI the option to buy up to 160 million shares (a 10% stake), with vesting tied to the deployment of compute capacity and AMD’s stock price milestones.
– AMD’s stock price surged nearly 35% following the deal announcement, with the final stock option tranche vesting when shares reach $600.
– This partnership is part of OpenAI’s broader strategy to secure chip supply for its AI infrastructure buildout, including multiple recent deals with companies like Nvidia, Broadcom, and Samsung.
In a landmark agreement set to reshape the artificial intelligence hardware landscape, AMD has secured a multi-year chip supply contract with OpenAI that promises to generate tens of billions in revenue. This strategic partnership significantly boosts AMD’s position within the competitive AI sector, providing a massive infusion of computational resources for one of the world’s leading AI research organizations.
The deal commits AMD to supplying a staggering 6 gigawatts of compute capacity to OpenAI, a scale of power comparable to that required by approximately 4.5 million households. This immense capacity will be delivered through several generations of AMD’s Instinct GPU accelerators, beginning with the forthcoming Instinct MI450 series. OpenAI is slated to receive the initial one gigawatt of this power in the latter half of 2026, coinciding with the new chip’s planned deployment.
AMD asserts that its MI450 GPUs will surpass the performance of competing products from Nvidia, specifically the Rubin CPX platform. This anticipated performance lead is expected to stem from a combination of hardware and software enhancements, many of which are being developed in close collaboration with OpenAI’s engineering teams. The existing MI355X and MI300X series from AMD are already employed by OpenAI for certain tasks, valued particularly for their substantial memory capacity and high bandwidth, which are critical for running large language model inference.
A notable component of the arrangement involves a substantial equity incentive. AMD has granted OpenAI an option to purchase up to 160 million shares, representing a ten percent ownership stake in the chipmaker. The vesting of these shares is directly linked to the fulfillment of the compute agreement. The first portion will vest upon the delivery of the initial one gigawatt of capacity, with subsequent portions unlocking as OpenAI procures the remaining compute power up to the full six gigawatts.
Furthermore, the final tranche of this stock option is contingent upon AMD’s share price reaching a milestone of $600 per share. This clause underscores the confidence both companies have in the partnership’s potential to drive significant market value. Following the announcement, AMD’s stock experienced a dramatic surge, opening at $222.24 on Monday, a jump of nearly 35 percent from its previous closing price of $164.67.
This agreement is a key element of OpenAI’s broader strategy to lock in long-term chip supplies and partnerships as it aggressively expands its AI infrastructure. This initiative includes the development of five new “Stargate” data centers with a combined planned capacity of seven gigawatts.
Dr. Lisa Su, AMD’s Chair and CEO, expressed enthusiasm for the collaboration, stating, “We are thrilled to partner with OpenAI to deliver AI compute at massive scale. 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.”
Echoing this sentiment, OpenAI CEO Sam Altman described the partnership as “a major step in building the compute capacity to realize AI’s full potential.” This deal is the latest in a series of recent strategic moves by OpenAI to secure its computational foundation. Just last month, the company entered into a monumental agreement with Nvidia involving a potential $100 billion investment and a supply of at least 10 gigawatts of compute. OpenAI also inked a $10 billion custom AI chip development deal with Broadcom and announced an expansion of its Stargate collaboration with Oracle and SoftBank.
Adding to this flurry of activity, OpenAI recently finalized agreements with Samsung Electronics and SK Hynix to source DRAM memory chips for its Stargate project and to construct new data centers in South Korea.
(Source: TechCrunch)
