Oracle has broadened its artificial intelligence infrastructure by scaling up its procurement pact with NVIDIA and incorporating processors from Google and other rivals into its cloud platform. This strategy provides Oracle clients with a wider array of choices for executing AI workloads amid surging demand for computing capacity.
Revealed on Tuesday (October 27, 2026), the broadened contract enables Oracle to roll out NVIDIA’s cutting-edge Vera Rubin AI server architectures at a magnitude quantified in multiple gigawatts. While the financial terms were kept private, the scaling underscores Oracle’s deepening capital commitment toward hardware supporting advanced AI programs.
Concurrently, Oracle is diversifying its hardware vendors. The enterprise intends to make Google’s Tensor Processing Units (TPUs) accessible via Oracle Cloud Infrastructure (OCI) for the inaugural time, alongside processors originating from Cerebras Systems and SambaNova Systems.
NVIDIA Deal Expands as AI Computing Demand Grows
NVIDIA continues to function as a primary provider of semiconductors utilized in training massive AI models. Its forthcoming Vera Rubin architecture is engineered to manage intensive processing operations, encompassing both model training and inference.
Oracle and NVIDIA noted that their widened arrangement would hit “multi-gigawatt scale,” though neither organization divulged a precise timeline for rollout or a total agreement price tag. NVIDIA Chief Executive Officer Jensen Huang has previously indicated that Vera Rubin setups could yield as much as USD 40 billion in revenue per gigawatt for NVIDIA.
Additionally, Oracle is cutting back its dependence on NVIDIA for specific networking elements. The cloud service provider intends to deploy its proprietary Acceleron networking chips inside select configurations rather than depending entirely upon NVIDIA’s proprietary networking gear.
Google TPUs and Cerebras Join Oracle Cloud
Oracle’s decision to integrate Google TPUs delivers an alternative avenue for AI processing. These customized microprocessors speed up machine-learning tasks and suit both the training of models and the execution of AI programs.
Google and Oracle announced that the TPUs are slated to arrive on OCI in 2027, with the rollout projected to grow over multiple years. This milestone establishes Oracle as the primary prominent cloud service provider outside of Google to provide the company’s TPUs natively through its proprietary network.
Oracle additionally intends to supply microprocessors from Cerebras and SambaNova, businesses specializing in AI inference. Inference represents the phase where a trained AI model produces answers to user queries. Contrasting with AI training—where NVIDIA holds a dominant foothold—the inference sector features a broader array of rival hardware creators.
Also Read: Oracle AI Updates: What’s New in 2026
Oracle Maintains Heavy AI Infrastructure Spending
Oracle has projected capital expenditures ranging between roughly USD 90 billion and USD 95 billion for its 2027 fiscal term. The enterprise has mentioned that clients are increasingly underwriting the semiconductors necessitated by their tasks, though it has omitted details regarding how it will fund these newest acquisitions.
Through the integration of NVIDIA platforms with hardware from Google, Cerebras, and SambaNova, Oracle aims to grant users enhanced adaptability in weighing performance, processing scale, and expenses.
This tactical shift likewise highlights a wider industry evolution: cloud operators are broadening their underlying frameworks while investigating substitutes to relying on a sole dominant silicon manufacturer. For Oracle, the primary hurdle will involve fulfilling escalating needs without permitting the expenditures associated with establishing AI capacity to outpace incoming client revenue.
