OpenAI has closed a multi-year agreement with AMD to incorporate new technologies into its infrastructure. Next-generation Instinct GPU, a move set to reconfigure the computing landscape for AI. The agreement includes the deployment of up to 6 gigawatts of capacity over several years and across several generations of hardware.
The plan starts in the second half of 2026 with a first installation of 1 gigawatt based on chips AMD Instinct MI450. In addition, the alliance includes an option for OpenAI to participate in AMD that, upon meeting certain milestones, could be raised to around 10% of the capital, which demonstrates the close alignment of interests between both companies.
What exactly does the agreement include?
In terms of infrastructure, OpenAI will gradually incorporate hundreds of thousands of GPUs AMD equivalent to 6 GW, with the aim of powering training and, especially, training tasks. inference on a large scale. The next batch of MI450 It will be the starting point for the deployment, which will be carried out in phases.
The scale of the project is considerable: 6 GW of computing power is a figure that, according to rough estimates, can be compared with the consumption of several million homesAlthough the volume is spread over time, the quantum leap strengthens OpenAI's ability to scale next-generation services and models.
AMD and OpenAI frame the collaboration as a multigenerational bet, combining hardware and software to optimize performance, costs, and deployment times. This technical fit aims to accelerate the adoption of the Instinct platform in workloads of Generative AI high traffic.
- Calendar: start in 2nd semester of 2026 with 1 GW initial.
- Product: introduction of Instinct MI450, with further iterations planned.
- Main use: reinforcement of large-scale inference and support for new models.
From AMD's side, the deal validates its roadmap against a market in which Nvidia dominates with around 90% fee in AI accelerators, and opens the door to adopting its solutions in large-scale installations.
The financial structure: warrants and objectives
As part of the agreement, AMD has issued a license to OpenAI. warrants by up 160 million shares at a symbolic price of $0,01 by title. The award will be made in tranches subject to previously agreed technical and financial goals.
Milestones include: first shipment of MI450 in the second half of 2026 and AMD share price targets that scale up to 600 dollars per share in the final stretch. If all requirements were met, OpenAI could reach approximately 10% of the manufacturer chip.
On the commercial front, AMD projects that the alliance will be able to report tens of billions in annual revenue, counting direct demand and the so-called "trickle-down effect" on other customers; in fact, executives contemplate overcome the 100.000 million accumulated in four years because of that added boost.
This capital component seeks to narrow the strategic alignment between supplier and customer, while facilitating OpenAI's access to supply and AMD's consolidation of its presence in large data centers.
Market impact and stock market reaction
The news triggered strong movements on the stock market: AMD shares even rose in value. above 20%, with intraday peaks of around 30% and an increase in market value close to 80.000 million of dollars in moments of greatest enthusiasm.
For analysts, OpenAI's commitment acts as confidence vote in AMD's technology, while not altering Nvidia's leadership, which continues to sell virtually everything it is capable of manufacturing. Even so, the agreement places AMD as credible alternative in large AI infrastructure contracts.
Executives from both companies have stressed that the agreement is an important step for speed up computing large-scale AI, although the market is closely monitoring the sustainability of the investment pace and the future profitability of data center deployment.
In parallel, Nvidia registered slight declines in the announcement session, a movement interpreted as profit taking in a segment where the computing demand remains tense.
Why OpenAI is diversifying: large-scale computing and multiple vendors
The movement fits into a strategy of multiple suppliers adopted by OpenAI. A few weeks ago, it closed a commitment with Nvidia of at least 10 gigawatts in systems and an announced investment of 100.000 million of dollars. With AMD, the company strengthens its access to chips and reduces its exposure to a single supplier.
According to internal objectives, Sam Altman has proposed to achieve the order of 250 gigawatts of total computing capacity by 2033. In this context, agreements such as the one signed with AMD help ensure supply and scalability for the next generation of models.
OpenAI also works with Broadcom in the development of its own processors and maintains infrastructure alliances with Oracle infrastructures , among other partners. This framework seeks to combine capacity, cost, and speed of deployment, key factors in a rapidly growing market.
In financial terms, the company generated around 4.300 million of dollars of revenue in the first half of 2025, with cash consumption close to 2.500 million during that period, according to reports. Hence, the company is exploring formulas for creative financing to accompany its expansion.
Sources close to the matter indicate that the agreement with AMD does not alter the computing plans already underway nor OpenAI's relationship with ecosystem and other partners, suggesting operational continuity while available capacity is expanded.
With this alliance, OpenAI gains access to more capacity and more options to scale its services, while AMD gains traction as a leading provider of AI accelerators. The numbers—6 GW, warrants for 160 million shares, and multi-billion-dollar revenue expectations—reflect the scale of a deal that could set the pace for the industry for years to come.
