Google is expanding its artificial intelligence chip business through external cloud partners, including Oracle, as it seeks to challenge Nvidia’s dominance in a market benefiting from growing demand for computing power.
Alphabet, Google’s parent company, says that Oracle would offer access to its tensor processing units, or TPUs, custom-designed chips the company has been developing since 2013. The move comes as Google seeks to reach customers beyond its own cloud infrastructure and broaden the market for its AI hardware.
The partnership marks a shift in Google’s strategy after years of promoting an integrated offering of its chips, AI models and cloud services. It also reflects the growing importance of partnerships in an AI infrastructure market where access to computing capacity has become a competitive advantage.
“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, referring to Google’s position in chips.
Google began developing its own silicon in 2013 but only started opening its TPUs to external customers this year. The company made the chips available through outside data centres in the second quarter, ahead of the Oracle announcement.
Thomas Kurian, Google’s cloud chief, joined the company in 2018 after spending more than 20 years at Oracle. The strategy shift comes as Google Cloud expands rapidly, with analysts expecting revenue growth of about 90% year over year, following an 82% increase in the second quarter.
“Google was great at doing its own thing, but it was never able to serve customers,” said Uday Cheruvu, a portfolio manager at Harding Loevner, which owns Alphabet shares. “That mindset has changed.”
Estimates for Google’s TPU business vary widely, but analysts expect substantial growth as demand for AI computing capacity increases. TD Cowen forecasts $9.4 billion in TPU sales in the second half of 2026 and $62.5 billion in 2027. Wells Fargo estimates about $9 billion in third-quarter sales and $99 billion next year.
Alphabet Chief Financial Officer Anat Ashkenazi said in July that only a small portion of revenue from existing TPU agreements would be recognised in 2026, with the “vast majority” expected in 2027. She also said Google planned to rely more heavily on external capacity in the third quarter as it expanded its own infrastructure.
Google Cloud generated $25 billion in revenue in the second quarter, compared with $42 billion at Amazon Web Services and $29 billion at Microsoft Azure. Google, however, has been growing faster than its larger cloud rivals.
“It’s almost like the bigger the pie is getting, the faster they’re accelerating into it,” said Ballard, whose firm owned about $236 million in Alphabet stock as of June 30.
Expanding external TPU sales presents Google with a strategic trade-off: the company can broaden its customer base and monetise its chip technology, but it must also preserve enough computing capacity for its own AI products and cloud services.
At a Goldman Sachs conference in September, Kurian said Google’s AI servers paid for themselves in less than two years, while servers running its own silicon reached payback in roughly half that time.
“It’s because we can optimize that whole stack,” Kurian said.
Research firm SemiAnalysis estimates that Google’s latest Ironwood TPU costs about 44% less to own and operate than an Nvidia GB200 server, potentially giving Google a pricing advantage as customers seek alternatives to Nvidia’s hardware.
However, selling chips externally could generate lower margins than providing a broader package of cloud services. TD Cowen estimates that TPU sales carry an operating margin of about 20%, compared with an overall Google Cloud operating margin of approximately 35% in the latest quarter.
Cheruvu said he did not expect Google’s cloud operating margin to expand substantially beyond the mid-30% range. Google declined to comment.
Google’s biggest TPU customer is Anthropic, whose Claude AI models compete with Google’s Gemini products.
Anthropic agreed in October 2025 to use up to 1 million TPUs in a deal worth tens of billions of dollars. Mizuho estimates the agreement will bring nearly one gigawatt of computing capacity online this year, describing the relationship as Google “arming its closest model competitor.”
OpenAI is also planning to buy TPUs from Google, CNBC confirmed after The Information first reported the ChatGPT maker’s intentions on Tuesday.
The deals illustrate how demand for AI computing is creating commercial opportunities even between companies competing in the development of AI models. They also underscore the importance of financing arrangements to secure the chips, data centres and networking equipment needed to run increasingly powerful AI systems.
Broadcom, which works with Google on TPUs, this week detailed a $63 billion financing vehicle backed by Blackstone and three major banks. The vehicle will buy TPUs and networking equipment from Broadcom and lease them to Anthropic for five years.
Analysts at William Blair estimate the financing plan covers about half of Anthropic’s computing needs in 2027, suggesting the company may require additional funding arrangements to secure capacity.
In its initial public offering filing on Monday, Anthropic said it uses chips from Google, Amazon, Nvidia and Advanced Micro Devices, allowing it to “select chips based on the best price-performance and efficiency characteristics for each workload.”
Stay ahead of the Stories shaping our world. Subscribe to Impact Newswire and join our
WhatsApp Channel for updates on global tech, business, and innovation—all in one place.
Dive deeper into the future with the Cause Effect 4.0 Podcast, where we explore the ideas, trends, and technologies driving the global AI conversation.
Got a story to share? Contact Us to reach a global audience with Impact Newswire.
Discover more from Impact AI News
Subscribe to get the latest posts sent to your email.
