How OpenAI’s deal with AMD could change the AI chip market
Techbehemoth and OpenAI yesterday struck a multi-billion-dollar partnership with chipmaker AMD, committing to purchase up to six gigawatts of the company’s next-generation processors to power new AI data centres.
The deal, which will start with AMD’s new ‘MI450’ chip in the second half of 2026, marks a key step in OpenI’s strategy to diversify its hardware supply and reduce dependence on industry behemoth Nvidia.
Under the agreement, the ChatGPT maker will acquire the chips either directly or through cloud partners such as Microsoft and Oracle.
AMD boss Lisa Su described the partnership as a milestone for the company, estimating it could generate tens of billions of dollars in revenue annually in the next five years.
Meanwhile, Sam Altman’s OpenAI will receive performance-based warrants for up to 160m AMD shares, roughly 10 per cent of the chipmaker, contingent on hitting share price targets and deployment milestones.
Following the news, shares in AMD rose by over 30 per cent.
Diversification and scale
The timing of the AMD agreement follows an even larger deal between OpenAI and Nvidia, under which the AI start-up will deploy at least 10 gigawatts of Nvidia-powered data centres, with Nvidia investing up to $100bn (£74.49bn) in OpenAI.
Analysts suggest that the AMD partnership is part of OpenAI’s broader strategy to secure enough high-performance computing power to meet the soaring demand for its AI LLM models, while also maintaining a resilient supply chain.
Matt Britzman, senior equity analyst at Hargreaves Lansdown, said: “AMD has achieved a breakthrough with OpenAI, demonstrating that its next-generation chips are a viable alternative for the industry’s largest AI workloads. However, Nvidia still maintains an edge because its offering spans the full data centre stack, from chips to networking and software”.
The agreement has been structured to ensure that OpenAI actively deploys AMD hardware, rather than simply taking the stock warrants.
Vesting occurs in tranches, beginning with the initial one-gigawatt installation, and scaling up to the full six-gigawatt commitment.
This approach aligns the interests of both firms: OpenAI secures long-term access to newly developed AI processors, while AMD gains a committed client for multiple chip generations.
Implications for the AI hardware sector
Historically, Nvidia has dominated the AI accelerator market, with estimates suggesting it controls more than 90 per cent of data centre GPU shipments.
The AMD-OpenAI partnership offers AMD a high-profile entry into this space, providing it with a foothold among the world’s leading AI developers.
“This deal is significant for AMD’s strategy in AI infrastructure” said Jean Hu, AMD’s chief financial officer. “It creates meaningful strategic alignment and is expected to deliver substantial revenue while helping accelerate OpenAI’s AI buildout”.
For OpenAI, the arrangement offers financial flexibility. Instead of paying upfront for equity stakes in a supplier, the performance-based warrant allows the company to acquire potential ownership in AMD through the achievement of deployment milestones.
The structure also demonstrates how AI infrastructure agreements are evolving beyond traditional vendor relationships, with customers gaining influence over technology roadmaps and potential financial stakes in their suppliers.
OpenAI is simultaneously pursuing a multi-pronged hardware strategy.
In addition to partnerships with AMD and Nvidia, the company is also developing its own custom chips in collaboration with Broadcom and has committed to purchasing $300bn in computing capacity from Oracle over the next five years.
This diversified approach is designed to mitigate risks of supplier concentration, optimise hardware for specific workloads, and ensure sufficient compute capacity for future AI development.
A measured challenge to Nvidia
Despite AMD’s entry into the AI data centre space, analysts caution that Nvidia’s market position remains strong.
The company continues to face robust demand for its GPUs and generates significantly higher revenue, with an expected $206bn for fiscal year 2025, compared with AMD’s $32.8bn.
While the AMD deal validates its technology and expands its market reach, it does not immediately threaten Nvidia’s dominance.
OpenAI chief Sam Altman emphasised that the AMD agreement complements, rather than replaces, its existing Nvidia partnerships.
“This is about building the compute capacity needed to realise AI’s full potential”, he said. “Working alongside AMD allows us to scale AI tools that can benefit people everywhere”.
Analysts describe the partnership as a “validation moment” for AMD, signalling the increasing financial and operational sophistication in AI infrastructure deals.
By incorporating equity incentives, technical collaboration, and staged deployments, the arrangement exemplifies how the sector is evolving to meet the extraordinary capital and technological demands of modern AI.