Google Partners With Oracle to Expand Access to Custom AI Chips

Alphabet agreed to offer its custom artificial intelligence chips through Oracle’s cloud infrastructure starting in 2027, marking the first time customers can access the processors on a competing platform.

Under the arrangement announced on Tuesday at the Oracle AI World conference in Las Vegas, Oracle Cloud Infrastructure will host Google’s eighth-generation Tensor Processing Units.

Oracle will provide the power, cooling and data center networking required for the processors, while Google will supply its hardware alongside associated software tools.

The deal enables Google to expand sales of its custom processors beyond its own data centers as cloud computing providers compete to meet rising corporate demand for artificial intelligence hardware.

“Different AI workloads have different compute needs,” said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, adding that the partnership gives customers access to multiple computing architectures in one cloud platform.

Google executives said offering the processors on external infrastructure provides greater flexibility for businesses running complex software models.

“The next generation of AI will require a diverse range of compute,” said Amin Vahdat, senior vice-president and chief technologist for AI and infrastructure at Google.

Google has developed custom chips since 2013 to challenge Nvidia, which currently supplies most of the graphics processing units used for artificial intelligence training.

Financial analysts said opening up hardware distribution could help Google capture broader market share in the rapidly growing market for specialized hardware.

“They're a very strong competitor, and it's very possible that they could eventually take market share from Nvidia,” said Chris Ballard, managing director of Check Capital Management.

Artificial intelligence startup Anthropic previously agreed to use up to 1 million Google chips, while regulatory filings showed it plans to spend $114.5 billion on Google compute services through 2033, Business Insider reported.

ChatGPT creator OpenAI also plans to purchase processors from Google, according to reports by CNBC and Business Insider, as technology companies seek alternative suppliers to Nvidia.

Google’s newest processor line includes the TPU 8t for model training and the TPU 8i for inference tasks, with the training chip delivering nearly three times the compute performance of its predecessor.

Google and Oracle did not disclose pricing terms or specific deployment capacities for the partnership, which is scheduled to launch in calendar year 2027.