Oracle brings Cerebras inference chips into its AI cloud

Oracle is adding a new engine to its AI cloud. The company announced a partnership with Cerebras Systems that will bring Cerebras’ AI inference technology into Oracle Cloud Infrastructure, better known as OCI.

The news landed on October 27, 2026, at Oracle AI World in Las Vegas. Customers will have to wait a bit, though. Availability is expected in 2027, so this is a reservation, not a delivery.

What Oracle and Cerebras actually agreed to

Under the deal, OCI customers will be able to tap Cerebras’ wafer-scale inference. That customer list includes AI labs, enterprises, and developers.

Cerebras’ pitch rests on its Wafer-Scale Engine. The processor puts compute and memory together on a single chip, which the company positions as the key to ultra-low latency inference.

Oracle is framing the addition as a complement to what it already sells. Its AI inference lineup already includes Dedicated AI Clusters. Cerebras brings what the companies describe as high-speed, deterministic options built for complex, multi-step AI workflows.

The executives make their case

Both companies leaned on the same theme: customers want faster AI, especially for agents. Cerebras CEO Andrew Feldman tied the deal directly to that demand.

“Bringing Cerebras fast inference to Oracle Cloud Infrastructure gives AI labs, enterprises, and developers the speed to build agents that do more, faster.”

Oracle’s side struck a similar note, with an emphasis on choice. Mahesh Thiagarajan, president of OCI, cast Cerebras as one more tool in a larger kit.

“Bringing Cerebras’ inference to Oracle Cloud Infrastructure will add another ultrafast and deterministic option to our broad compute portfolio.”

Why agents change the math

The focus on multi-step workflows reflects how AI usage is shifting. A simple chatbot answers one question and stops. An agent might plan, search, write, check its own work, and try again, all before a human sees anything.

Each of those steps is its own inference call. Small delays stack up fast. A pause that feels trivial on one request becomes a real drag when an agent strings dozens of them together.

What this means for Oracle, Cerebras, and cloud buyers

There are plenty of open questions. Neither company disclosed capacity, pricing, or deployment scale. Those details will determine whether this becomes a core offering or a niche one.

Timing is the other variable to watch. With availability expected in 2027, the AI market has a long runway to shift before customers get their hands on the service.