OpenAI Signs $10 Billion Deal with Cerebras for Massive Inference Power

OpenAI has reached a multi-year agreement with chip startup Cerebras Systems that provides for the deployment of 750 megawatts of computing power based on Cerebras systems. The partnership is valued at over ten billion dollars and extends over three years, with implementation scheduled in multiple phases through 2028. This deployment will roll out in multiple stages beginning in 2026.

According to both companies, the agreement marks the world’s largest deployment of high-speed AI inference. OpenAI CEO Sam Altman is personally involved as an investor in Cerebras, and the two companies have been holding regular discussions about a possible collaboration since 2017.

Inference is not about training new AI models, but about running existing ones – that is, when users send requests via ChatGPT and these should be answered as quickly and well as possible.

Speed as a Competitive Advantage

Cerebras develops AI processors at wafer scale that, according to the company, generate answers significantly faster than GPU-based systems. The technology integrates massive computing capacity, memory, and bandwidth on a single large-format chip, thereby eliminating bottlenecks in conventional hardware.

For applications such as coding agents or language chats, large language models on Cerebras systems deliver results up to 15 times faster than on comparable GPU platforms. This speed fundamentally transforms the user experience and enables real-time interactions that promote longer engagement times and more productive workflows.

OpenAI faces significant challenges in procuring sufficient computing capacity. The company has more than 900 million weekly users and has repeatedly pointed to a severe shortage of computing resources.

The integration of Cerebras technology is targeted specifically for workloads where low latency is critical. An infrastructure executive at OpenAI explained that the company’s compute strategy aims to optimally align different systems with specific tasks. Negotiations between the two companies began in the fall and led to the signing of a term sheet by Thanksgiving. Generally, they know each other well; since 2017 there have been regular meetings between the two companies.

Financial Dimension and Market Context

According to information from company circles, Cerebras is in discussions about a funding round of one billion dollars that would value the company at 22 billion dollars – nearly a tripling compared to the previous valuation. The startup, founded about ten years ago, has faced challenges in the past:

In preparation for an IPO in 2024, the company disclosed that the majority of its revenue came from a single customer, G42, based in Abu Dhabi. Cerebras withdrew the planned IPO and instead raised 1.1 billion dollars privately, which corresponded to a valuation of 8.1 billion dollars. In the meantime, the company has concluded new agreements with IBM and Meta, thereby diversifying its customer base – now also with OpenAI.

OpenAI is pursuing a broad-based chip strategy to break free from the dominance of individual suppliers. The company is developing a custom chip jointly with Broadcom and has signed an agreement for the use of AMD’s new MI450 chip. Nvidia, the leading provider of AI chips, had reached a preliminary agreement with OpenAI in September for the sale of up to ten gigawatts of chips, which has not yet been finalized. In parallel, Nvidia concluded a 20 billion dollar licensing agreement with inference startup Groq in December. The market for specialized inference chips is gaining momentum as AI companies seek fast and cost-efficient alternatives to conventional GPU systems.

OpenAI’s financial commitments are increasingly raising questions among investors. The company generated approximately 13 billion dollars in revenue last year, while it signed cloud contracts with Oracle, Microsoft, and Amazon totaling approximately 600 billion dollars. Sam Altman emphasizes that the company will finance these commitments, staggered over several years, through future revenue growth. OpenAI is in an early phase of another massive funding round that could value the company at 830 billion dollars before the new investment. The financing is intended to prepare for a planned IPO and support the company’s extensive growth plans.