Overview
- Oracle and Google announced the partnership at Oracle AI World in Las Vegas on Oct. 27, 2026, saying Google’s latest TPU hardware will begin appearing on Oracle Cloud Infrastructure in 2027 with a phased, multiyear scale-up.
- The initial offering will include Google’s eighth‑generation TPU 8t for model training and TPU 8i for inference, specialized chips that speed large‑model math for training and for serving model responses.
- Oracle will keep using Nvidia and other accelerator partners while adding TPUs to give customers more hardware options inside a single cloud platform.
- Multiple outlets reported that OpenAI is planning to buy TPUs from Google as it expands capacity, a report that companies involved have not publicly confirmed and that would sit alongside OpenAI’s own Jalapeno chips developed with Broadcom.
- The commercial impact will hinge on concrete details that remain unannounced — pricing, which TPU models and regions come first, and integration with existing software and orchestration tools — and those factors will drive whether TPUs on OCI become a mainstream option for enterprise AI buyers.