Overview
- Oracle confirmed a multi-gigawatt deployment of Nvidia’s latest Vera Rubin AI servers to scale its cloud training capacity.
- Oracle will be the first hyperscale cloud provider outside Google to offer Google’s TPU family, with service rollout planned for 2027.
- The company will add Intel-backed SambaNova and Cerebras accelerators targeted at inference workloads to broaden customer options.
- Oracle said it will use its self-developed Acceleron networking ASIC in some Nvidia-based systems instead of Nvidia network chips.
- The firms did not disclose financial terms and Oracle projects about $90–95 billion in chip-related spending for fiscal 2027 while saying customers will increasingly shoulder chip procurement costs.