Dive Brief:
- OpenAI cut a deal with AMD to deploy six gigawatts of AI compute capacity in a massive infrastructure push unveiled Monday. The companies will share technical expertise and align product roadmaps across multiple generations of AMD Instinct GPUs, according to a joint announcement.
- “This partnership is a major step in building the compute capacity needed to realize AI’s full potential,” Sam Altman, co-founder and CEO of OpenAI, said in the release. The companies plan to activate one gigawatt of AMD Instinct MI450 series GPUs by the second half of next year.
- OpenAI has invested heavily in the compute needed to power large language model technologies through the $500 billion Stargate infrastructure initiative. The company agreed to purchase up to 160 million shares of AMD stock as specific milestones of the Monday deal are achieved, yielding billions in revenue for AMD, the chipmaker’s EVP, CFO and Treasurer Jean Hu said in the announcement.
Dive Insight:
As enterprises pursue elusive returns on generative AI investments, big tech is pouring resources into a building boom where capacity is measured in gigawatts and investments in the hundreds of billions of dollars.
Last year, data center spending reached nearly half a trillion dollars, fueled by hyperscale investments in AI infrastructure, according to Dell’Oro Group research. The spending spree continued through the first half of this year, boosting data center investments in Q1 and Q2 by 53% and 43%, respectively, to $134 billion and $158 billion.
OpenAI has been a major player throughout.
The scope and cost of the model builder’s involvement in the White House-endorsed Stargate undertaking expanded last month, with the company pledging to bring seven gigawatts of AI capacity online over the next three years at five data center sites in partnership with Oracle and Softbank. Earlier in the year, OpenAI and Oracle began work on $300 billion in buildouts aimed at delivering an additional 4.5 gigawatts over the next five years.
GPU cloud provider CoreWeave is intertwined in the broader buildouts with OpenAI. In the last six months, OpenAI has nearly doubled its commitments for CoreWeave compute capacity to $22.4 billion, the companies announced on Sept. 25.
The AMD deal arrives two weeks after OpenAI entered into a strategic alliance with GPU giant Nvidia to deploy 10 gigawatts of AI processing capacity, which is also set to begin to come online during the second half of 2026. Nvidia will invest up to $100 billion in OpenAI as part of the alliance.
AI buildouts have been a boon for Nvidia. The company largely eclipsed AMD as the primary supplier of GPUs, growing its revenues nearly eightfold in three years and climbing to the top rung in Gartner’s global semiconductor market share rankings for the first time last year. AMD fell one spot to No. 8 last year, compared to 2023.
As GPU demand surged, AMD invested almost $5 billion to acquire cloud infrastructure provider ZT Systems last year. The chipmaker saw revenues increase 32% year over year to a record $7.7 billion in Q2 2025, boosted in part by a 14% year-over-year increase in data center segment revenue, which reached $3.3 billion.
AMD expects the OpenAI partnership to “enable additional revenue from existing and new customers at scale,” Chair and CEO Lisa Su said Monday, during a virtual press conference. The deal has the potential to generate over $100 billion in revenue over the next several years, Su added.
