OpenAI and Advanced Micro Devices (AMD) have formed a landmark partnership that sent AMD stock up over 30% on Monday. The deal could give AI firm of Sam Altman a 10% interest in the chipmaker, one of the largest-ever deployment deals for GPUs in the history of the AI industry.
The agreement has OpenAI utilizing 6 gigawatts worth of AMD Instinct graphics processing units spread across multiple generations of gear and a span of years. The utilization commences with a first-year implementation of 1 gigawatt for second half 2026.
OpenAI Partners with AMD to Secure Massive Compute Power for Scalability
OpenAI President Greg Brockman didn’t shy away when defining the need for this alliance. In a talk with CNBC, he commented on a particular aspect that the firm isn’t yet strong enough to ship numerous features within ChatGPT and other offerings, which earn revenue primarily due to a lack of sufficient available compute power.
“We have to do this,” Brockman said. “This is so central to our goal if we’re ever going to be scalable to all of humanity, this is what needs to occur.”
The sentiment underscores a critical challenge facing the AI industry: as models become more sophisticated and user demand grows, the infrastructure requirements are becoming astronomical.
Though OpenAI referred to the transaction as being worth billions, the company refused to quote a number. The financial mechanics, however, are interesting. AMD issued OpenAI a warrant for up to 160 million common stock shares with both volume deployment and share-price-of-AMD-based-vesting provisions.
The first tranche vests with the first full gigawatt deployment. Additional tranches unlock as OpenAI scales toward the full 6-gigawatt commitment and meets key technical and commercial milestones. If OpenAI exercises the full warrant, it would acquire approximately 10% ownership in AMD based on current outstanding shares.
It is a commercial success story for AMD, as much as a vote of confidence for its next-generation Instinct strategy. For years, AMD lagged behind Nvidia in the AI accelerator market, but it now has a premium flagship customer at the epicenter of the generative AI explosion.
OpenAI’s $1 Trillion Week: AMD, Nvidia, and the Closed-Loop AI Economy
AMD chief executive Lisa Su told CNBC recently that AI is on a 10-year path, and such partnerships help hold the ecosystem together to create the best technologies. “Ultimately, you need the foundation compute to do that,” she added.
The transaction constitutes what Su termed a “true win-win allowing the world’s greatest AI buildout and a step forward for the whole AI ecosystem.”
This AMD partnership comes on the heels of less than two weeks since OpenAI revealed a landmark $100-billion equity-and-supply agreement with Nvidia. That agreement, which had witnessed Nvidia gaining a majority ownership stake in OpenAI, will deliver 10 gigawatts of capacity within the total 23-gigawatt infrastructure plan of OpenAI.
With just the Nvidia and AMD agreements, OpenAI spent roughly $1 trillion on new build-out costs for just a little more than two weeks. At $50 billion per gigawatt construction cost, the scale of capital deployment is staggering.
OpenAI is even negotiating with Broadcom for creating custom chips for its subsequent generations of models, diversifying its approach to hardware even further.
The arrangement belies the increasing circularity of the AI industry’s corporate economy, which sees capital, equity, and compute circulated between the same small number of companies that create and instantiate the tech. Nvidia finances purchasing its own chips.
Oracle contributes to building out the sites. AMD and Broadcom step into the supply role. OpenAI supplies the anchor for demand.
It’s a closed ecosystem that some observers fear might experience real stress if any part of the chain breaks down.
How is OpenAI Securing Its AI Infrastructure Future?
Under its Stargate program, OpenAI is fast becoming one of the industry’s most ardent infrastructure developers. The company’s very first site, which is located in Abilene, Texas, is already operational using Nvidia chips, with the building still expanding capacity.
Future constructions at its plants in New Mexico, Ohio, and Middle West are likely to have a wide array of suppliers, such as AMD, which should make OpenAI less dependent on a sole provider and more flexible.
With AI expanding at a meteoric rate, such deals like this between AMD and OpenAI will be the norm, rather than the exception. The issue isn’t whether companies are going to invest hugely in infrastructure, but how quickly they’ll be able to spin out the compute horsepower needed to keep pace with the rate of innovation.
