• Oracle becomes the first cloud vendor (outside of Google) to sell TPUs
• TPU 8t and 8i, launched earlier this year, will be added to the platform
• Google's TPUs are specifically optimized for AI workloads

Oracle has confirmed it will be making Google Cloud's TPUs available through its own OCI platform in a bid to expand the range of AI accelerators available to customers. In doing so, customers will be able to choose which hardware they use to train and run AI models.

Announced at the company's annual AI World conference, the partnership gives Oracle's customers access to Google's purpose-built AI accelerators without them having to go through Google Cloud, while also offering another option instead of Nvidia's GPUs.

It marks an important development for Google's hardware, because it's the first time a hyperscaler other than Google Cloud has sold its TPUs.

Oracle becomes the first third-party cloud vendor to sell Google's TPUs

Google's Tensor Processing Units (TPUs) are custom-designed chips designed to accelerate machine learning workloads, spanning both model training and AI inference, but unlike general-purpose GPUs, they've been optimized specifically for artificial intelligence.

While Nvidia remains the dominant supplier of AI accelerators, helping it to maintain its position as the world's most valuable company with a market cap of $5.265 trillion, its facing increasing pressure from hyperscalers' own hardware and custom silicon endeavours.

OCI President Mahesh Thiagarajan explained that "different AI workloads have different compute needs," hence the decision to offer another option to customers.

"The next generation of AI will require a diverse range of compute," Google AI & Infrastructure SVP and Chief Technologist Amin Vahdat added.

OCI is launching the partnership with Google's eighth-generation TPUs – 8t (for training) and 8i (for inference). Both were launched by Google Cloud earlier this year, and will become available via OCI in 2027.