Oracle is expanding its artificial-intelligence infrastructure agreement with Nvidia to multi-gigawatt scale while simultaneously adding processors from competing suppliers, signaling a broader push to make Oracle Cloud Infrastructure a multi-chip AI platform.
Oracle and Nvidia said OCI Superclusters will expand around Nvidia’s Vera Rubin NVL72 systems, with Vera Rubin NVL72 and Vera CPUs now generally available through OCI.
The strategy goes beyond Nvidia. Oracle also plans to offer Google’s Tensor Processing Units through OCI for the first time and add inference processors from Cerebras Systems and other suppliers. OCI will become the first hyperscale cloud platform outside Google to offer Google TPUs.
The expansion reflects an AI infrastructure market that is becoming increasingly specialized. Nvidia remains a dominant supplier for advanced model training, while inference—the process of operating trained AI models—is creating opportunities for alternative processor architectures.
Oracle’s strategy is designed to accommodate both markets while giving customers greater flexibility over the processors and networking technologies supporting their workloads.
The companies did not disclose the value of the expanded Nvidia arrangement. Oracle has maintained plans for approximately $90 billion to $95 billion in fiscal-2027 capital spending, illustrating the enormous infrastructure requirements behind its AI expansion.
For Oracle, the competitive question is increasingly whether it can differentiate OCI through scale, networking technology and access to multiple computing architectures.
The next milestones will be deployment timing, customer adoption of alternative accelerators and whether Oracle can convert its enormous infrastructure commitments into durable cloud revenue.
