Google Cloud Opens TPU Sales to Oracle

Alphabet has long maintained that customers benefit from purchasing chips, models, and cloud services from a single provider. On Tuesday, the company announced a partnership that departs from that full-stack model by making its custom hardware available to outside infrastructure operators. This development occurred one day before its third-quarter earnings report.

Oracle will now offer Google's tensor processing units, or TPUs, to its own customer base. These are the custom chips Google has developed since 2013. This marks the first time buyers can access TPUs outside of Google's own data centers. The arrangement signals a strategic shift for Thomas Kurian, the cloud chief who joined Google in late 2018 after spending more than 20 years at Oracle.

Kurian is pursuing this external sales channel from a position of financial strength. Analysts project Google Cloud revenue will increase by approximately 90% compared to the same period last year. This follows an 82% rise in the second quarter. TPUs are a primary driver of that growth, and the company had already begun opening these chips to outside data centers earlier this year.

Investors are watching the competitive implications closely. Chris Ballard, a managing director at Check Capital Management, noted that Google is a strong competitor that could eventually take market share from Nvidia. Uday Cheruvu, a portfolio manager at Harding Loevner, said Google has historically focused on internal development. He added that the company's mindset regarding serving external customers has changed.

Estimates for the TPU business show steep growth trajectories. TD Cowen models $9.4 billion in TPU sales for the second half of 2026 and $62.5 billion for 2027. Wells Fargo expects about $9 billion in third-quarter sales and $99 billion in 2027. CFO Anat Ashkenazi stated in July that only a small portion of TPU revenue from existing agreements will be recognized this year. She said the vast majority of that revenue will arrive in 2027.

Google still trails its largest rivals in total scale. Its second-quarter cloud revenue was $25 billion. Amazon Web Services reported $42 billion, while Microsoft Azure reported $29 billion. However, Google is currently the fastest-growing among the major providers. Selling chips externally introduces trade-offs, particularly regarding profit margins. TD Cowen estimates a 20% operating margin on TPU sales, compared to an overall cloud margin of about 35% last quarter.

Kurian must balance external sales against Google's internal requirements. At a Goldman Sachs conference in September, he stated that Google's AI servers pay for themselves in under two years. Servers running its own silicon reach payback in about half that time. He attributed this speed to the ability to optimize the entire stack. SemiAnalysis estimates the latest Ironwood TPU costs about 44% less to own and operate than an Nvidia GB200 server.

Anthropic is currently Google's largest TPU customer and a major competitor in AI models. The company agreed in October to use up to 1 million TPUs in a deal worth tens of billions of dollars. Mizuho says this agreement brings nearly a gigawatt of capacity online this year. CNBC has confirmed that OpenAI also plans to buy TPUs from Google.

Anthropic is securing outside financing to cover the cost of these chips. Broadcom, a key partner in TPU development, detailed a $63 billion vehicle this week. Backed by Blackstone and three large banks, this vehicle will buy TPUs and networking gear from Broadcom and lease them to Anthropic for five years. William Blair analysts say this plan covers about half of Anthropic's 2027 needs.