Oracle said on Tuesday, October 27, at its AI World conference that it will offer cloud services built on Cerebras AI computers in its data centers. Bloomberg reported the plan first and the companies confirmed it. Oracle Cloud Infrastructure plans to launch the service in 2027 and did not say how much Cerebras hardware it will use. “AI inference demands choice,” said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure.

Cerebras builds each chip from a whole silicon wafer, which it says lets its systems answer AI requests much faster than rivals. It went public in May, counts OpenAI and Amazon Web Services as customers, and its shares rose about 9.5% on the news. Analysts compiled by Bloomberg expect its revenue to nearly triple next year to almost $3 billion from under $1 billion this year.

Syndicated Reuters coverage adds that Oracle is expanding its Nvidia deal to multi-gigawatt scale on Nvidia’s Vera Rubin servers and will add Google TPU access and other chips, also from 2027, with no dollar figure disclosed. Lens could not open the Reuters and Bloomberg originals and relied on The Next Web, the companies’ announcement and syndicated copies.

Investors worry Cerebras is chasing a small niche in a market where Nvidia’s products bring in close to $100 billion in sales each quarter.

Sources

Image: Oracle Campus in Austin 2018.jpg by DronePhotographer, CC0, via Wikimedia Commons. Illustrative photo; it does not show the events described.