Oracle and Nvidia announced plans to expand Oracle Cloud Infrastructure's Vera Rubin computing systems to multi-gigawatt scale, while making Vera Rubin NVL72 systems and Vera CPUs generally available on OCI.
Oracle also announced plans to offer Google's Tensor Processing Units and inference accelerators from Cerebras and SambaNova, while incorporating its proprietary Acceleron networking technology.
The companies did not disclose the total financial value of the expanded Nvidia arrangement or the complete deployment schedule.
Oracle's previously stated fiscal 2027 capital-expenditure target remains approximately $90 billion–$95 billion. NVIDIA Investor Relations
Primary sources: Nvidia–Oracle Official Announcement — October 27 · Oracle–Cerebras Announcement
Why it matters strategically
Oracle is simultaneously expanding its relationship with the dominant supplier of AI training hardware and introducing alternatives for different computing workloads.
This suggests that the competitive structure of AI infrastructure may become increasingly differentiated.
Training large models requires substantial computing resources.
Inference workloads can have different performance requirements, including latency, throughput, memory utilization, and energy efficiency.
The most economically attractive hardware for one workload may not be optimal for another.
Consequently, the ability to support multiple hardware architectures could become a competitive advantage.
