Cerebras and Oracle have announced plans to offer Cerebras inference through Oracle Cloud Infrastructure in 2027. The proposed service would let AI labs, businesses and developers use wafer-scale computing for applications that generate responses from trained models. The companies are targeting interactive applications and agents that make repeated model calls, with availability still a future deployment milestone.
Cerebras, an AI computing company, and cloud provider Oracle announced their partnership on October 27. They plan to make Cerebras inference available through Oracle Cloud Infrastructure in calendar year 2027. Inference is the stage in which a trained model generates a response to a request. The proposed service is aimed at AI labs, enterprises and developers building interactive applications.[1], [2]
Cerebras builds its processors at the scale of a silicon wafer, rather than using the wafer only as a surface from which many smaller chips are cut. Its Wafer-Scale Engine combines computing and memory on that processor. The company describes reducing data movement and communication between chips as the design’s approach to generating model output for long reasoning sequences and multi-step work.[1]
The partners point to AI agents, software that can reason and take actions through several steps. Such workflows make repeated calls to a model, including while checking earlier work or choosing an action. The companies aim to provide another computing option for these response-sensitive applications, building on Oracle’s existing infrastructure and inference capabilities. Availability remains planned for 2027; Oracle did not disclose how much Cerebras hardware it would use.[1], [2]
