Oracle Expands Nvidia AI Chip Deal, Adds Google TPUs and Cerebras to Cloud Offerings

Oracle said Tuesday it is expanding its cloud AI chip partnership with Nvidia to "multi-gigawatt scale" using Nvidia's next-generation "Vera Rubin" AI servers, while diversifying its offerings to include chips from Nvidia rivals Google and Cerebras Systems. The companies did not disclose the deal's value, though Nvidia CEO Jensen Huang has said Vera Rubin systems can generate up to $40 billion in revenue per gigawatt. Oracle will also for the first time offer Google's tensor processing unit chips, alongside offerings from Intel-backed SambaNova Systems and Cerebras, targeting the more competitive inference market. Oracle said even its Nvidia-based systems will use its in-house "Acceleron" networking chips instead of some Nvidia networking silicon. The cloud firm, which holds an investor day this week, did not disclose how it will finance the purchases; it has maintained a capital spending target of about $90 billion to $95 billion for fiscal 2027.