Cerebras jumps on new Oracle cloud deal

Cerebras Systems shares jumped by about 10% after the company announced an AI inference collaboration with Oracle Cloud Infrastructure. Shares of Cerebras Systems were up about 10%.

The partnership, announced on October 27, 2026 at Oracle AI World in Las Vegas, will integrate Cerebras’ AI inference technology into Oracle Cloud Infrastructure. The offering targets AI labs, enterprises and developers building interactive AI applications.

Cerebras’ inference capabilities are planned to become available on Oracle Cloud in calendar year 2027. Oracle plans to begin deploying Cerebras equipment in the first half of 2027, with the companies noting that timing, features and pricing may change at their discretion. As a result, customers are not expected to run Cerebras-powered workloads on Oracle’s cloud in the current quarter.

The collaboration focuses on delivering high-speed, low-latency AI inference for business and research use cases. It is described as supporting AI-driven business applications, particularly those requiring rapid response times.

Cerebras’ technology centers on a wafer-scale engine that integrates compute and memory on a single processor. This design reduces data movement and communication between separate chips, which the companies say supports faster token generation in long reasoning tasks and multi-step AI agent workflows.

Cerebras positions its architecture as a way to minimize data transfer bottlenecks in inference. The collaboration with Oracle is framed as expanding Cerebras’ market access through a large cloud network while providing Oracle customers with additional inference options.

The new deal adds Cerebras to Oracle’s existing AI capabilities, including Dedicated AI Clusters in OCI Enterprise AI. Oracle has also expanded partnerships with Nvidia and is adding inference offerings from Google and SambaNova alongside Cerebras.

Oracle executives have described Cerebras and SambaNova as providing extremely low-latency inferencing options. The arrangement is characterized as filling out Oracle’s bench of AI hardware providers rather than replacing existing partners.

Cerebras stock has been under pressure since its Nasdaq debut and recently touched its weakest level since listing. Sentiment had been affected by reports that OpenAI picked Nvidia GPUs, not Cerebras chips, to run the "Ultrafast" mode in GPT-6.1 Sol.

Cerebras nonetheless has a multi-year agreement with OpenAI valued at $10 billion to deliver 750 megawatts of computing capacity through 2028. The Oracle partnership adds another large-scale channel for its AI inference technology as cloud demand for interactive and agentic AI continues to grow.