By Chinonyerem Emmanuel
AI chipmaker Cerebras has struck a monumental deal with OpenAI, securing 750 megawatts of computing power through 2028, according to a blog post from the company behind ChatGPT released Wednesday. Sources close to the matter estimate the agreement to be worth over $10 billion, marking one of the largest AI infrastructure deals in recent history.
The partnership represents a strategic shift for Cerebras, which has been heavily reliant on the United Arab Emirates’ G42, a single client that accounted for 87% of the company’s revenue in the first half of 2024. By collaborating with OpenAI, Cerebras is diversifying its customer base while positioning itself as a serious competitor in the rapidly growing generative AI market.
Cerebras is known for building massive, high-performance processors designed to train and run generative AI models at scale. This makes it a direct challenger to Nvidia, the dominant supplier of AI chips for cloud providers such as Amazon and Microsoft, which rent Nvidia’s graphics cards to clients by the hour.
Nvidia’s dominance in AI hardware helped it become the first company to reach a $5 trillion market capitalization in October, driven by investor enthusiasm for the AI boom.
In a parallel move, Cerebras’ rival Groq disclosed in December that Nvidia had signed a non-exclusive licensing agreement, which included the transfer of some employees to Nvidia. While Groq’s cloud service was not part of the agreement, the deal, reportedly worth $20 billion in cash, marked Nvidia’s largest transaction to date.
OpenAI highlighted the benefits of the Cerebras partnership in its blog, emphasizing improvements in speed and efficiency. “Cerebras adds a dedicated low-latency inference solution to our platform,” wrote Sachin Katti, who leads compute infrastructure at OpenAI. “This translates to faster responses, more natural interactions, and a stronger foundation to scale real-time AI to millions of people.”
The deal builds on months of prior collaboration, as OpenAI worked with Cerebras to ensure its GPT-OSS open-weight models could operate seamlessly on Cerebras hardware, alongside chips from AMD and Nvidia.
Cerebras has experienced rapid growth in recent years. In September 2024, the company filed for an initial public offering (IPO), revealing second-quarter revenue of nearly $70 million, a significant jump from $6 million in the same period a year earlier. However, its net loss also grew, reaching $51 million, up from $26 million in 2023.
Unlike most top-tier tech IPOs, Cerebras’ prospectus excluded major investment banks and used an auditor outside the so-called Big Four accounting firms.
The company later withdrew the filing in October after raising $1.1 billion in funding, which valued the company at $8.1 billion. CEO Andrew Feldman explained on LinkedIn that the withdrawal was to update financials and strategy in light of the rapidly evolving AI market.
Cerebras’ growing client roster now includes notable names such as Cognition, Hugging Face, IBM, and Nasdaq. In March 2025, the Committee on Foreign Investment in the United States (CFIUS) approved Cerebras’ request to sell shares to G42, reinforcing the company’s global strategic ambitions.
The deal with OpenAI signals a new era for Cerebras as it seeks to challenge industry giants, expand its market presence, and power the next generation of AI applications at unprecedented speed and scale.
Source: CNBC
Original article
Comments