Oracle Expands Nvidia Partnership While Adding Google TPU and Cerebras Chips
Oracle announced an expanded cloud AI chip partnership with Nvidia, planning large-scale deployment of Vera Rubin servers. To address diverse AI workload needs, Oracle is diversifying its supply chain by becoming the first hyperscaler to offer Google TPUs, alongside inference chips from Cerebras and SambaNova, with services expected to roll out starting in 2027.
Full Report
IT Home reported on October 28 that Oracle announced on Tuesday (local time) it will expand its cloud AI chip partnership with Nvidia while further advancing supply diversification by introducing products from Nvidia’s competitors, including chips from Google (a subsidiary of Alphabet) and Cerebras Systems.
Oracle and Nvidia stated that Oracle will deploy Nvidia’s latest-generation Vera Rubin AI servers at scale, reaching a deployment size in the "multiple gigawatts" range. The specific financial terms of the deal were not disclosed. However, Nvidia CEO Jensen Huang previously indicated that each gigawatt of Vera Rubin system deployments could generate up to $40 billion in revenue for Nvidia (Note: IT Home exchange rate conversion approximates this to 267.35 billion RMB).
Oracle is also driving supply diversification. The company noted that even systems based on Nvidia chips will move away from using certain Nvidia networking components, opting instead for Oracle’s proprietary Acceleron network chips.
Additionally, Oracle announced it will offer Google’s Tensor Processing Units (TPUs) to customers for the first time. Oracle will become the first hyperscale cloud provider outside of Google to offer TPUs. Both companies stated that the service is expected to launch in 2027 and will scale gradually over several years, supporting large model training, inference, and agentic applications. Google’s eighth-generation TPUs include the TPU 8t for training and the TPU 8i for inference.
Mahesh Thiagarajan, President of Oracle Cloud Infrastructure, said in a statement: "Different AI workloads have different computational requirements." He noted that providing more choices allows customers to flexibly select infrastructure tailored to the specific performance, throughput, and cost-efficiency needs of various AI workloads.
Beyond Google’s chips, Oracle plans to offer chips from SambaNova Systems (backed by Intel) and Cerebras Systems. Both companies specialize in AI inference chips—the computational phase used when AI systems generate responses after receiving requests from human users.
Currently, Nvidia dominates the AI model training market, but the competitive landscape for AI inference is more diverse, with a wider array of participants.
Andrew Feldman, CEO of Cerebras, told Reuters regarding the partnership with Oracle: "This will significantly expand our tier-one customer base." He added: "Our products are entering other companies’ data centers through large-scale deployments."
Oracle plans to hold an investor day this week but has not yet disclosed how it will finance these chip procurements.
During last month’s earnings call, Oracle stated that customers are increasingly bearing the costs of chip procurement. The company projects related expenditures to reach approximately $90 billion to $95 billion in fiscal year 2027 (current exchange rate approximates this to 601.538 billion to 634.956 billion RMB).