Cerebras’ initial compute-capacity deliveries to OpenAI fell behind schedule because of delays in data-center capacity contracts.

Cerebras’ CEO attributed the delayed initial OpenAI compute-capacity deliveries to delays in data-center capacity contracts. The supplied statement identifies a contracting-related delivery issue; it does not attribute the delay to accelerator design or manufacturing.

Management also said OpenAI business increased in Q3 compared with Q2. That reported increase is separate from the delivery shortfall, and the sources provide no customer-specific revenue amount or revised delivery date.

Another CEO-attributed message states that demand for OpenAI’s Ultrafast mode, which runs partly on Cerebras chips, substantially exceeded available capacity. The message says this led to usage-based limits, but does not quantify the unmet demand, number of affected users, or additional capacity required.