OpenAI’s ambitions are no longer just about building smarter models - they’re about building the infrastructure to run them. In the past few months, the company has signed some of the largest hardware and cloud deals in tech history, locking in billions in future compute capacity as it races to stay ahead of rivals.

Its most eye-catching move is a multi-year agreement with AMD to supply up to six gigawatts of AI chips, giving OpenAI the option to acquire roughly 10% of AMD through warrants tied to performance milestones. Analysts estimate the deal could generate over $100 billion in long-term revenue for AMD, while giving OpenAI crucial leverage against its dependency on Nvidia.

“We need as much computing power as we can possibly get.”

- Greg Brockman, OpenAI President

That comes alongside a $30 billion-a-year cloud deal with Oracle - the largest in AI history - securing more than 4.5 gigawatts of data-centre capacity under the “Stargate” project. The partnership will underpin OpenAI’s global expansion, including a planned $25 billion data-centre complex in Argentina.

At the same time, Nvidia has deepened its role in OpenAI’s ecosystem, supplying GPUs across a proposed 10-gigawatt roadmap and reportedly committing billions in shared infrastructure and development investment. The company’s dominance still defines the upper limit of what OpenAI can deploy - but for the first time, OpenAI is building serious redundancy.

Inside the company, resource allocation has become a strategic flashpoint. Brockman described the internal process of dividing scarce GPU resources as “pain and suffering.”

“These interwoven deals may bolster short-term valuations but raise concerns about market transparency and sustainability.”

- Financial Times

OpenAI’s web of partnerships - AMD for chips, Oracle for cloud, Nvidia for hardware, and SoftBank for capital - has created a complex, trillion-dollar supply network built on trust, timing, and sheer computational ambition. Whether that pays off will depend on its ability to turn raw power into profit before the next wave of AI competition arrives.