Google Exports Its AI Chips Through Oracle as Cloud Revenue Soars
Alphabet is leasing its custom TPUs to outside partners, a shift that analysts say could challenge Nvidia's dominance in the AI hardware market.
By Dominic Foster ·
Google is selling its custom artificial intelligence chips to customers outside its own data centers for the first time. The company announced on Tuesday that Oracle will offer Google's tensor processing units, or TPUs, to its enterprise clients. This marks a departure from Google's long-standing strategy of selling its chips, models, and cloud services as a single, integrated package.
The move comes ahead of Alphabet's third-quarter earnings report. Analysts project Google Cloud revenue will jump approximately 90 percent from a year earlier, following an 82 percent increase in the second quarter. Thomas Kurian, the executive leading Google Cloud, has spent more than 20 years at Oracle before joining Google in late 2018. He is now using that background to expand the company's hardware reach beyond its own infrastructure.
Market data shows Google Cloud generated $25 billion in revenue during the second quarter. That figure trailed Amazon Web Services, which reported $42 billion, and Microsoft Azure, which earned $29 billion. However, Google is expanding its share at a faster rate than its rivals. Chris Ballard, a managing director at Check Capital Management, noted that the company's chip business is growing rapidly. His firm held about $236 million in Alphabet stock as of June 30.
Financial models suggest the TPU business is scaling quickly. TD Cowen estimates TPU sales will reach $9.4 billion in the second half of this year and $62.5 billion in 2027. Wells Fargo projects approximately $9 billion in third-quarter TPU sales and $99 billion in revenue for next year. Anat Ashkenazi, Google's chief financial officer, stated in July that most TPU revenue from current agreements will be recognized in 2027 rather than this year.
Anthropic, the developer of the Claude AI models, is the largest customer for these chips. The company agreed in October to use up to 1 million TPUs in a deal valued in the tens of billions of dollars. OpenAI is also planning to purchase TPUs from Google. Anthropic recently filed for an initial public offering, stating it uses silicon from Google, Amazon, Nvidia, and Advanced Micro Devices to select the most efficient hardware for specific workloads.
To support these sales, Google is balancing external revenue with its own infrastructure needs. Kurian said at a conference in September that servers running Google's own silicon reach cost recovery in roughly half the time of standard AI servers. SemiAnalysis estimates that Google's latest Ironwood TPU costs about 44 percent less to own and operate than an Nvidia GB200 server. However, selling chips externally may lower profit margins. TD Cowen estimates a 20 percent operating margin on TPU sales, compared with the roughly 35 percent margin Google Cloud posted last quarter.
