Oracle Cloud adds Google TPU acceleration and ChatGPT integration

On Tuesday local time (October 27, 2026), Oracle announced that it is expanding its partnership with NVIDIA on cloud AI chips. By leveraging NVIDIA's latest-generation Vera Rubin servers, the company plans to scale its cloud deployments to the multi-gigawatt level. For the first time, Oracle will also offer Google's TPU compute through its cloud platform, alongside chips from Cerebras Systems and SambaNova Systems. Even on NVIDIA-based systems, Oracle will replace some NVIDIA networking silicon with its in-house Acceleron networking chips. Oracle is maintaining its capital expenditure target of roughly $90 billion to $95 billion for fiscal 2027.

The expanded collaboration also covers Oracle Cloud Infrastructure (OCI). The companies plan to grow OCI's superclusters to multi-gigawatt scale based on the NVIDIA Vera Rubin NVL72 system, aiming to support growing demand for AI training, inference, and agent workloads. The NVIDIA Vera Rubin NVL72 system and NVIDIA Vera CPU are now generally available on OCI. Customers can scale CPU and GPU resources independently, aligning GPU-heavy frontier model training and inference with CPU-heavy tasks that agents execute between model calls. This lets compute configurations match workloads more closely, reducing idle capacity and improving infrastructure utilization.

According to SiliconANGLE, citing CNBC, Oracle is also bringing Google's TPU AI accelerators to its public cloud, marking the first time Google has supplied those chips to another hyperscaler. Oracle will additionally add Cerebras Systems' wafer-scale accelerators and OpenAI's Jalapeño inference processor, and it will introduce ChatGPT integrations. As of the latest reports, these expansion and partnership plans remain announcements without a further rollout timeline.