Oracle Adds Google TPUs and Cerebras to OCI, Expands Nvidia Deal

ORCL used its AI World event on Tuesday to widen the range of AI chips on Oracle Cloud Infrastructure (OCI), expanding its deal with Nvidia (NVDA) while adding offerings from Google and Cerebras. On the Nvidia side, Vera Rubin NVL72 systems and Vera CPUs are now generally available on OCI, and the two companies plan a multi-gigawatt expansion of OCI Superclusters built on Vera Rubin, according to Nvidia's release. Neither company disclosed a dollar value or GPU count.

The Google and Cerebras additions are separate agreements. Alphabet's (GOOGL) eighth-generation Google TPUs, the TPU 8t for training and TPU 8i for inference, will be offered through OCI starting in calendar 2027 and ramp over several years, making OCI the first hyperscaler outside Google to offer them, per Google's announcement. Cerebras (Nasdaq: CBRS) plans to bring its wafer-scale inference service to OCI, also in calendar 2027. "AI inference demands choice," said Mahesh Thiagarajan, president of OCI.

On the software side, Snowflake's (SNOW) AI Data Cloud will be offered as a native, first-party service on OCI, bought and billed through Oracle and payable with Oracle Universal Credits, though no availability date was given. "Organizations are looking for faster, simpler ways to turn their enterprise data into business value with AI," said Oracle CEO Clay Magouyrk. Oracle also expanded its OpenAI ties: a new Oracle AI Workspace lets employees act across Fusion, NetSuite and Oracle's database from inside ChatGPT, and Codex powers Oracle's App Composer for building applications.

ORCL closed up 3.5% at $130.53; earlier in the session 24/7 Wall St. noted the stock was still down about 31% for the year. What to watch: pricing and capacity for the 2027 TPU and Cerebras offerings, when the Snowflake service becomes available, and whether a multi-vendor chip lineup reduces OCI's dependence on Nvidia supply.