OpenAI signed a multiyear computing partnership with Cerebras Systems worth over $10 billion, securing 750 megawatts of processing capacity through 2028. The deal, announced Wednesday, will integrate Cerebras's specialized AI processors into OpenAI's infrastructure to accelerate response times across ChatGPT and other AI models.
Cerebras will deliver the computing power in multiple tranches, with capacity coming online gradually through 2028. The Sunnyvale-based chipmaker builds wafer-scale processors that combine compute, memory, and bandwidth on single chips to reduce bottlenecks in AI inference workloads.
"This partnership will make ChatGPT not just the most capable but also the fastest AI platform in the world," said Greg Brockman, OpenAI co-founder and president. "Cerebras adds a dedicated low-latency inference solution to our platform," said Sachin Katti, who works on compute infrastructure at OpenAI. "That means faster responses, more natural interactions, and a stronger foundation to scale real-time AI to many more people."
The arrangement represents a strategic diversification for both companies. OpenAI expands its compute portfolio beyond Nvidia and AMD GPUs, while Cerebras reduces dependence on its largest customer, UAE-based G42, which accounted for 87% of revenue in early 2024.
Cerebras is reportedly seeking $1 billion in new funding at a $22 billion valuation ahead of a planned IPO later this year. The company withdrew its initial IPO filing in October after securing $1.1 billion in pre-IPO funding that valued it at $8.1 billion.
OpenAI has been exploring collaboration with Cerebras since 2017, when it reportedly considered teaming up with Tesla to acquire the chipmaker. Recent testing showed Cerebras hardware running OpenAI's GPT-OSS-120B model 15 times faster than conventional systems.
The deal follows OpenAI's other massive infrastructure commitments, including six gigawatts from AMD and ten gigawatts from Nvidia. Nvidia announced a $100 billion investment in OpenAI infrastructure in September, while AMD committed its GPU capacity in October.
Cerebras CEO Andrew Feldman compared the partnership's potential impact to broadband transforming internet usage. "Just as broadband transformed the Internet, real-time inference will transform AI, enabling entirely new ways to build and interact with AI models," he said.
Both companies maintain sovereign compute programs for countries willing to pay for dedicated AI infrastructure. Cerebras houses a significant portion of its software team in Toronto, while OpenAI has expressed interest in participating in Canada's sovereign AI initiatives.
The computing capacity will support OpenAI's rapid infrastructure expansion as demand for power-hungry AI tools continues unabated. The company is also developing its own AI chips with Broadcom and advising Microsoft on its Maia chip design.
Industry analysts view the partnership as part of an unprecedented bet on sustained AI demand, with technology companies committing hundreds of billions to computing infrastructure. OpenAI's deals and investments reportedly lay groundwork for a potential IPO that could value the company at up to $1 trillion. The company continues to navigate legal challenges, including Elon Musk's lawsuit against OpenAI CEO Sam Altman that is scheduled for trial on April 27.
