Google CEO says Pichai says people "love" AI Overviews and keep coming back to search more

Google suggests AI is the main driver behind that cloud growth, pointing to a sharp increase in token usage. That's a reasonable proxy for overall AI activity, but it's a weak signal for actual usefulness, which is what will really matter going forward.

To meet the demand, Google is also rethinking how it sells its TPUs, the company's in-house AI chips. Until now they've only been available through Google Cloud, but the company will start shipping them directly to select customers' data centers.

AI is showing up across the business

Google says revenue from generative AI models is up 800 percent year-over-year and that deal sizes keep growing. Paying monthly active users on Gemini Enterprise, the company's B2B product, rose 40 percent in a single quarter. The consumer Gemini apps also had their best quarter yet.

The effect is showing up in Search too. Google Search, which plenty of observers had written off as a declining business, grew 19 percent to $60.4 billion. Ads chief Philipp Schindler credited Gemini for helping Google better understand what users actually mean when they type a query. That lets Google serve relevant ads against longer, more complex searches that used to be hard to monetize.

"People love our AI experiences like AI Mode and AI Overviews, and they're coming back to search more," Pichai said. Those AI-generated answers have drawn plenty of criticism, since Google uses them to hold onto traffic that would otherwise go to outside websites, effectively capturing publisher revenue without paying much for the underlying content.

Ads will eventually come to the Gemini chatbot, but for now the focus is AI Mode. The cost of serving AI Overviews and "core AI responses" in AI Mode has dropped 30 percent, thanks to hardware and engineering gains.

Google still isn't breaking out the AI numbers

For all the growth percentages, Google still won't share specific revenue figures for its AI business. It's also unclear how much of the booked workload comes from circular deals with AI startups like Anthropic, where companies take Google's money and spend it right back on Google services.

The bigger question is whether customers are getting real economic value from these AI services. That's what will ultimately decide whether Google and the other hyperscalers can justify their investments over the long run.