Google Cloud Revenue Forecast to Jump 90% as TPU Sales Expand Beyond Internal Use

Analysts project Google Cloud revenue will increase by approximately 90% from the same period last year, following an 82% rise in the second quarter. This acceleration is largely attributed to the expanding sales of tensor processing units, or TPUs, which Google has developed in-house since 2013.

Alphabet announced on Tuesday that Oracle will offer access to these custom chips, marking the first time customers can utilize TPUs outside of Google's own data centers. The move signals a strategic shift under cloud chief Thomas Kurian, who joined Google in late 2018 after a two-decade tenure at Oracle. By partnering with external infrastructure providers, Kurian aims to broaden the market for hardware that was previously restricted to Google's internal ecosystem.

Financial forecasts for the TPU business vary by institution but consistently indicate rapid expansion. TD Cowen models $9.4 billion in TPU sales for the second half of this year, with total sales reaching $62.5 billion in 2027. Wells Fargo expects third-quarter sales of roughly $9 billion and projects $99 billion in revenue for next year. Chief Financial Officer Anat Ashkenazi noted in July that while some revenue from existing agreements will be recognized this year, the majority is scheduled to arrive in 2027.

Despite this growth, Google still trails its primary competitors in total cloud revenue. In the second quarter, Google Cloud generated $25 billion, compared to $42 billion for Amazon Web Services and $29 billion for Microsoft Azure. However, Google is currently growing at a faster rate. Analysts suggest the company is well-positioned to capture market share from Nvidia, the long-dominant player in AI chips. Uday Cheruvu of Harding Loevner stated that Google's previous focus on internal solutions has changed, allowing it to better serve outside customers.

Expanding chip sales to external partners introduces financial trade-offs. TD Cowen estimates an operating margin of 20% on TPU sales, which is lower than the 35% overall cloud margin recorded last quarter. Cheruvu does not expect Google's cloud operating margin to exceed the mid-30s. Kurian must balance external customer growth with internal efficiency. At a September Goldman Sachs conference, he explained that servers running Google's own silicon reach payback in about half the time of standard AI servers, citing optimization across the entire software and hardware stack.

Google's largest TPU customer is Anthropic, a company whose Claude models compete directly with Google's Gemini. Anthropic agreed last October to use up to 1 million TPUs in a deal valued in the tens of billions of dollars. Mizuho estimates this agreement brings nearly a gigawatt of compute capacity online this year. CNBC also confirmed that OpenAI plans to purchase TPUs from Google, following an initial report from The Information.

To finance its chip purchases, Anthropic is utilizing a $63 billion vehicle backed by Blackstone and three major banks, detailed this week by Broadcom. The financing allows Anthropic to lease TPUs and networking gear from Broadcom for five years. William Blair analysts indicate this arrangement covers approximately half of Anthropic's projected 2027 compute needs. In its IPO filing on Monday, Anthropic stated it uses silicon from Google, Amazon, Nvidia, and AMD to select the best hardware for specific workloads.