By Stephen Nellis
SAN FRANCISCO, Oct 27 (Reuters) - Oracle said on Tuesday it is expanding its cloud AI chip deals with Nvidia while also diversifying to add offerings from Nvidia rivals such as Alphabet's Google and Cerebras Systems.
Oracle and Nvidia said the cloud firm will expand to "multi-gigawatt scale" with Nvidia's newest generation of "Vera Rubin" AI servers. The companies did not disclose the value of the deal, though Nvidia CEO Jensen Huang has previously said that Vera Rubin systems can
generate up to $40 billion in revenue per gigawatt for the chipmaker.
Oracle is also diversifying its offerings, saying that even the Nvidia-based systems will skip some of Nvidia's networking chips in favor of Oracle's own in-house "Acceleron" networking chips. Oracle on Tuesday said it will begin offering Google's tensor processing unit (TPU) chips for the first time.
"Different AI workloads have different compute needs," Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, said in a statement. "That choice gives (customers) the flexibility to choose the right infrastructure for the performance, throughput and economics of each AI workload.”
In addition to Google chips, Oracle plans to offer chips from Intel-backed SambaNova Systems and Cerebras Systems, both of which make chips for what is known as inference, when AI systems respond to requests from human users. While Nvidia dominates the market for training AI systems, the inference market is more competitive and diverse.
"That's a massive expansion of our customer base of tier-one customers," Cerebras CEO Andrew Feldman said of the Oracle deal in an interview with Reuters. "We're being deployed in other people's data centers at large scale."
Oracle, which plans to hold an investor day this week, did not disclose how it plans to finance the chip purchases. During an earnings call last month, Oracle said customers are increasingly footing the bill for chips and maintained its capital spending target of about $90 billion to $95 billion for the 2027 fiscal year.
(Reporting by Stephen Nellis in San Francisco; Editing by Will Dunham)
