OpenAI's $10 Billion Bet: Cerebras Deal Secures 750MW to Rival Nvidia

OpenAI isn't just renting chips anymore; it's securing an empire. In a staggering deal worth over $10 billion, AI chipmaker Cerebras Systems will deliver 750 megawatts of computing power through 2028. This massive commitment, announced via a blog post on Wednesday, marks a significant shift in the AI hardware landscape as the creator of ChatGPT seeks to diversify its infrastructure beyond Nvidia's dominant GPUs.

Inside the OpenAI Cerebras $10 Billion Chip Deal

For Cerebras, this partnership is a lifeline for revenue diversification. In the first half of 2024, the United Arab Emirates' G42 accounted for a whopping 87% of the company's revenue. CEO Andrew Feldman noted that winning a second major customer like OpenAI is the natural progression of keeping their first big client happy. From OpenAI's perspective, the attraction lies in Cerebras' low-latency inference solution, which promises faster AI responses and more natural human-AI interactions as the company scales to millions more users.

A Long-Awaited Rivalry with Nvidia

The relationship between Sam Altman and Cerebras dates back to 2017, according to emails surfacing from recent litigation. Interestingly, Elon Musk reportedly tried to acquire the startup in 2018 for Tesla. Now, Cerebras stands as a formidable challenger to Nvidia, which recently hit a $5 trillion market cap. By optimizing its open-weight models like gpt-oss for Cerebras silicon, OpenAI is building a more resilient, multi-vendor hardware strategy that includes AMD and Nvidia.

IPO Ambitions and Future Expansion

Despite a brief setback in October 2024 when it withdrew its IPO paperwork, Cerebras is preparing for a public comeback. After securing $1.1 billion in funding that valued the company at $8.1 billion, the chipmaker plans to re-file with its new, stronger financial outlook. The company's revenue in the second quarter of 2024 reached nearly $70 million, a massive jump from $6 million just a year prior. With data centers spanning the U.S. and abroad, Feldman's firm is set to expand its footprint significantly to meet OpenAI's massive compute requirements.