Oracle, Cerebras target 2027 for OCI inference service

Oracle and Cerebras plan to make Cerebras AI Inference available through Oracle Cloud Infrastructure in calendar 2027. The planned service would add a wafer-scale, non-GPU inference option to OCI for AI labs, enterprises and developers building latency-sensitive applications.

The calendar-year window is the only timing the companies provided. They did not disclose regional availability, pricing, commercial terms, supported models, capacity, service interfaces or service-level commitments.

Cerebras said the planned OCI service would use its Wafer-Scale Engine. The company describes it as a single wafer-scale processor that integrates compute and memory, reducing data movement and communication between separate chips that can slow inference. Oracle Cloud Infrastructure President Mahesh Thiagarajan characterized the service as another inference choice in Oracle’s compute portfolio.

For now, the announcement describes product direction rather than a production launch. Its disclaimer says the companies are not committing to deliver specific functionality and may change development, release timing and pricing at their discretion. The release included no measured production benchmark for Cerebras Inference on OCI and identified no customer already using the planned service in production.