Overview
- OpenAI confirmed an agreement for Cerebras to provide 750 megawatts of compute through 2028, with people close to the company estimating the deal’s value at more than $10 billion.
- OpenAI executive Sachin Katti said Cerebras adds a dedicated low-latency inference option, promising faster responses and a foundation for scaling real-time AI.
- The contract diversifies revenue after G42 represented 87% of Cerebras’ sales in the first half of 2024, according to prior disclosures.
- Cerebras is reported to be in talks to raise nearly $10 billion at an implied valuation of about $22 billion, a development that has not been finalized.
- The company targets inference with its wafer-scale CS-3 systems powered by WSE-3, with reported benchmarks indicating strong Llama 3 70B performance and lower power use versus Nvidia Blackwell-based setups.