One signal

Update What changed

Oracle plans to integrate Cerebras wafer-scale processors into its cloud data centres, with the service expected in the first half of 2027. That is a more specific timetable for the Cerebras hardware Oracle has already said it will add to Oracle Cloud Infrastructure.

The Yahoo Finance article, originally published by Stocktwits, says the offering is aimed at AI inference. It describes Cerebras’ wafer-scale design as a way to reduce data movement compared with conventional multi-chip systems; the article attributes claims about faster token generation to the companies.

Cerebras co-founder and chief executive Andrew Feldman told Stocktwits that faster inference could let agents complete more steps without making users wait.

Sources and evidence

• Oracle Expands Cloud Infrastructure Arsenal With Google, OpenAI, And Cerebras Partnerships - Yahoo Finance: Yahoo Finance reports that Oracle plans to integrate Cerebras wafer-scale processors into its cloud data centres for AI inference in the first half of 2027.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

Oracle plans to start deploying Google’s AI accelerators in its cloud next year, making it the first hyperscaler to receive the chips from Google, SiliconANGLE reports. Oracle says it will gradually add more TPU capacity.

The report says the clusters will use Google’s optical circuit switching to move data between TPUs, alongside copper-based connections and Oracle’s Acceleron networking technology. Google’s optical system uses mirrors to direct light between servers, avoiding an extra conversion from light to electricity and back that can use more power.

SiliconANGLE also details Oracle’s new AI Workspace integrations, which let ChatGPT work with Oracle applications and coordinate tasks across services.

Sources and evidence

• Oracle brings Google’s TPUs to its cloud, adds ChatGPT integrations - SiliconANGLE: SiliconANGLE reports that Oracle plans to begin deploying Google TPUs in its cloud next year, with Google’s optical circuit switching and Oracle’s Acceleron networking, alongside newly detailed ChatGPT integrations for Oracle applications.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

Oracle is expanding its AI infrastructure beyond the chip options already discussed, with plans for multi-gigawatt NVIDIA Vera Rubin capacity and new Intel, AMD and SambaNova offerings. Analytics India Magazine reports that Oracle says Vera Rubin systems and CPUs are already available on Oracle Cloud Infrastructure for training and inference.

The report adds that Oracle plans to bring Google TPUs to OCI from 2027, while Cerebras inference technology is also planned for the platform that year. The Cerebras service is aimed at developers and businesses building applications with frequent model calls, including AI agents.

There are also planned Intel Diamond Rapids-based OCI X14 offerings, AMD Venice-powered Compute E7 instances, and SambaNova SN50 systems for inference. The report says these Intel offerings do not yet have a firm availability date.

Sources and evidence

• Oracle Expands AI Compute Options With NVIDIA, Google TPUs and Cerebras - Analytics India Magazine: Analytics India Magazine reports that Oracle is broadening OCI’s AI infrastructure with multi-gigawatt NVIDIA Vera Rubin capacity, planned Intel, AMD, SambaNova, Google TPU and Cerebras offerings, and related AI products and data services.

Independent WittyWires Watcher; not an official account or feed.