October 27, 2026 — Oracle is expanding its AI chip supply deal with Nvidia to “multi-gigawatt scale” on Vera Rubin servers, while also offering Google’s TPU chips for the first time and adding inference chips from Cerebras and SambaNova. Why it matters: Oracle is turning its cloud into a multi-vendor chip marketplace instead of an Nvidia-only shop.
What happened
Oracle and Nvidia said on Tuesday that Oracle 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. Nvidia CEO Jensen Huang has previously said Vera Rubin systems can generate up to $40 billion in revenue per gigawatt for the chipmaker.
Even inside the Nvidia systems, Oracle is asserting more control: Reuters reported the Nvidia-based systems will skip some of Nvidia’s networking chips in favor of Oracle’s own in-house “Acceleron” networking chips.
The diversification is the story
Oracle said it will begin offering Google’s tensor processing unit (TPU) chips for the first time. It also plans to offer chips from SambaNova Systems and Cerebras Systems, two makers of chips built for inference — the stage where AI systems respond to user requests. Nvidia dominates AI training, but the inference market is more competitive and diverse.
“Different AI workloads have different compute needs,” Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, said in a statement, adding that the choice gives customers flexibility on performance, throughput and economics. Cerebras CEO Andrew Feldman told Reuters the Oracle deal is a massive expansion of its tier-one customer base, with Cerebras being deployed in other companies’ data centers at large scale.
Oracle did not disclose how it plans to finance the chip purchases. On an earnings call last month it maintained a capital spending target of about $90 billion to $95 billion for its 2027 fiscal year, and said customers are increasingly footing the bill for chips. Oracle plans to hold an investor day this week.
Why it matters
For anyone renting AI compute, more chip choice in one cloud means real leverage: training on Nvidia, inference on cheaper specialist chips, and Google TPUs as a third option, all under one contract. For Nvidia’s rivals, an Oracle-scale deployment is the kind of reference win that has been hard to land. The open question is financing — Oracle is spending at a historic pace, and it has not yet said how these purchases get paid for.
Frequently asked questions
What did Oracle announce with Nvidia?
An expanded supply deal taking Oracle to multi-gigawatt scale on Nvidia’s Vera Rubin AI servers. No deal value was disclosed.
Is Oracle offering non-Nvidia chips?
Yes. Oracle will offer Google TPU chips for the first time, plus inference chips from Cerebras and SambaNova, and will use its own Acceleron networking chips in some Nvidia-based systems.
How much is Oracle spending on AI infrastructure?
Oracle has maintained a capital spending target of about $90 billion to $95 billion for its 2027 fiscal year, and did not disclose financing for these chip purchases.
Sources: Reuters