Google Cloud Revenue More Than Doubles on TPU Demand

Google Cloud’s revenue more than doubled to $30.71 billion in the third quarter ended September, as the division highlighted demand for its Tensor Processing Units (TPUs), specialized systems for AI training and inference. Alphabet’s total revenue rose 16% to $127.6 billion from $102.3 billion a year earlier, while net income was $26.17 billion. CEO Sundar Pichai said AI infrastructure and solutions were major drivers of cloud growth and that Google was working to add capacity; he also cited strong demand for TPU systems.

Google Cloud CEO Thomas Kurian said the division had 18 product lines each generating more than $1 billion in annual revenue. He attributed quarterly growth to Google Cloud Platform’s AI and core services, with TPU sales also contributing. Google says TPU 8t offers up to 2.7x better training performance and TPU 8i offers 80% better inference price performance than the previous generation. Google and Meta recently worked together on TorchTPU, which lets developers train, serve and deploy PyTorch models on TPUs using familiar tools.

Kurian said nearly 100 customers use TPUs, and named Anthropic, Apple, Mirendil, Physical Intelligence, Citadel Securities, Jump Trading, Caltech, Cassava and Cirrascale among customers or public references. Google has entered partnerships to deliver TPU systems with Crux AI and Oracle, with more partnerships in development. Kurian described the company’s approach as improving its silicon, optimizing software for performance and workload portability, and widening access beyond Google Cloud to customers’ data centers and other providers. He also cited an agreement with Blackstone for Crux AI, a joint venture offering TPU systems.

Kurian said that over the past 12 months, more than 2,600 customers used at least 100 billion tokens on Gemini Enterprise, while nearly 75 used at least 10 trillion tokens. Nearly 90% of Fortune 100 companies use Gemini Enterprise, and customers who used it in Q3 2025 more than quadrupled their spending with Google year over year. Across its AI services, Google has sold more than 100 million knowledge-work seats. Gemini Enterprise includes an agent that can research and answer questions, write and run code, and automate processes; Kurian said it offers security and cost controls and a choice of frontier models.

Kurian cited BNP Paribas, which is deploying Gemini Enterprise to more than 65,000 employees; Tata Steel, which deployed more than 300 specialized AI agents in 9 months and cut customer service times by 50%; and Ulta Beauty, whose AI shopping assistant achieved a 3x higher conversion rate than traditional site browsing. He said Gemini use also drives demand for computing, storage, databases and cybersecurity services. Alphabet CFO Anat Ashkenaz expects TPU sales to contribute a small share of revenue by the end of this year, with that contribution rising in 2027. The article also mentions an unaddressed rumor that OpenAI plans to buy Google TPU chips.

Google Cloud’s revenue more than doubled to $30.71 billion in the third quarter ended September, as the division highlighted demand for its Tensor Processing Units (TPUs), specialized systems for AI training and inference. Alphabet’s total revenue rose 16% to $127.6 billion from $102.3 billion a year earlier, while net income was $26.17 billion. CEO Sundar Pichai said AI infrastructure and solutions were major drivers of cloud growth and that Google was working to add capacity; he also cited strong demand for TPU systems.

Google Cloud CEO Thomas Kurian said the division had 18 product lines each generating more than $1 billion in annual revenue. He attributed quarterly growth to Google Cloud Platform’s AI and core services, with TPU sales also contributing. Google says TPU 8t offers up to 2.7x better training performance and TPU 8i offers 80% better inference price performance than the previous generation. Google and Meta recently worked together on TorchTPU, which lets developers train, serve and deploy PyTorch models on TPUs using familiar tools.

Kurian said nearly 100 customers use TPUs, and named Anthropic, Apple, Mirendil, Physical Intelligence, Citadel Securities, Jump Trading, Caltech, Cassava and Cirrascale among customers or public references. Google has entered partnerships to deliver TPU systems with Crux AI and Oracle, with more partnerships in development. Kurian described the company’s approach as improving its silicon, optimizing software for performance and workload portability, and widening access beyond Google Cloud to customers’ data centers and other providers. He also cited an agreement with Blackstone for Crux AI, a joint venture offering TPU systems.

Kurian said that over the past 12 months, more than 2,600 customers used at least 100 billion tokens on Gemini Enterprise, while nearly 75 used at least 10 trillion tokens. Nearly 90% of Fortune 100 companies use Gemini Enterprise, and customers who used it in Q3 2025 more than quadrupled their spending with Google year over year. Across its AI services, Google has sold more than 100 million knowledge-work seats. Gemini Enterprise includes an agent that can research and answer questions, write and run code, and automate processes; Kurian said it offers security and cost controls and a choice of frontier models.

Kurian cited BNP Paribas, which is deploying Gemini Enterprise to more than 65,000 employees; Tata Steel, which deployed more than 300 specialized AI agents in 9 months and cut customer service times by 50%; and Ulta Beauty, whose AI shopping assistant achieved a 3x higher conversion rate than traditional site browsing. He said Gemini use also drives demand for computing, storage, databases and cybersecurity services. Alphabet CFO Anat Ashkenaz expects TPU sales to contribute a small share of revenue by the end of this year, with that contribution rising in 2027. The article also mentions an unaddressed rumor that OpenAI plans to buy Google TPU chips.