Oracle to expand Nvidia deal and add Google and Cerebras chips — Channel NewsAsia

In San Francisco, Oracle said it would expand its cooperation with Nvidia in cloud infrastructure for artificial intelligence and add chips from Google, Cerebras Systems and SambaNova Systems to its offerings. As Channel NewsAsia reports, on October 27 Oracle and Nvidia announced plans to scale cloud infrastructure to a multi-gigawatt level using Nvidia's next-generation Vera Rubin servers.

Expansion of cooperation with Nvidia

The companies did not disclose the value of the agreement. Earlier, Nvidia CEO Jensen Huang said that Vera Rubin systems could generate up to $40 billion in revenue for the manufacturer per gigawatt of capacity.

Oracle also said it would use its own Acceleron networking chips in Nvidia-based systems instead of some Nvidia networking chips. Oracle Cloud Infrastructure President Mahesh Thiagarajan explained that different AI tasks have different computing resource requirements, while the choice of infrastructure allows customers to take performance, bandwidth and economic factors into account.

More current news is available on the UA.News Telegram channel Telegram.

New chips for cloud services

Oracle will offer customers Google's tensor processing units, or TPUs, for the first time. In addition, the company plans to provide chips from Intel-backed SambaNova Systems and Cerebras Systems. Both companies produce chips for inference, the stage of AI system operation when they generate responses to user queries.

Nvidia dominates the market for chips used to train artificial intelligence systems, while the inference segment is more competitive and diverse. Oracle did not say how it would finance chip purchases. During its quarterly earnings call last month, the company said customers were increasingly paying for chips and reaffirmed its target capital expenditures for fiscal 2027 of $90 billion to $95 billion.