Google CEO Sundar Pichai says cloud forgoes revenue to keep AI at the frontier

Searches for “google ceo” rose in India on Wednesday, 21 October, as two pieces of news about Google’s present and former chief executives circulated: a wide-ranging interview with Sundar Pichai, and a report on Eric Schmidt’s stake in the AI company Anthropic.

In the interview, published on Tuesday by the Big Technology Podcast, Pichai said Google is willing to give up some near-term cloud revenue so that its most advanced AI research gets first call on computing power. He spoke to journalist Alex Kantrowitz at Google’s Mountain View headquarters.

Compute for research comes first

“Our cloud business is at any given moment in time forgoing short-term revenues because we are making compute trade-offs to support frontier AI development,” Pichai said. He described adequate compute for Google DeepMind’s research and model building as “non-negotiable”, with the company’s other businesses planned around it.

He said the shortage of computing capacity is industry-wide. A customer asking now for capacity in 2027 would find that “incredibly difficult”, he said, adding that Google had leased short-term capacity last year to help some customers. Because data centres take years to build, the capacity Google has today reflects decisions he took in 2024.

Asked whether the AI build-out could overshoot demand, Pichai said it was possible. Shortages of components such as memory, power or land, or shocks to the wider economy, could cause problems, he said, but he was confident the compute Google is planning would be put to use over time.

Gemini 4 Argon and AI building AI

Pichai rejected suggestions that Google had settled for cheaper, lighter “Flash” models instead of competing at the cutting edge. Google rolled out its Gemini 4 Argon model globally on 13 October, according to The Verge, as summarised by Techmeme. Pichai said Argon handled close to 25 trillion tokens in its first 72 hours, roughly three times what Gemini 3.0 did, and that Google was short of capacity to serve it. He acknowledged that some areas, including parts of coding, still need work. The usage figures are Google’s own and have not been independently verified.

On recursive self-improvement, where AI systems help develop better AI, Pichai said agents now assist researchers in parts of model development, but the industry is “pretty far” from the end state. He admitted there are “clear gaps” in alignment research and said Google relies on real-time monitoring among other safeguards. He also said the volume of code changes submitted by Google engineers had risen by about 30% since Argon’s launch, with about a third fewer rollbacks.

Anthropic, Apple and space

Pichai said Google invested in Anthropic early because its founding team was clearly capable of frontier work. He described the arrangement under which Apple uses Gemini for Siri as “working well”, but declined to say whether it would expand. On data centres in orbit, he said Google was two to three years into the effort, had sent its TPU chips into space and was awaiting a first “hello world” signal from them.

A former CEO in the news

Separately, former Google chief executive Eric Schmidt holds about 1% of Anthropic, according to a Bloomberg estimate reported by TechCrunch on Tuesday. Schmidt, who stepped down as Google CEO in 2011, put tens of millions of dollars into Anthropic’s 2021 Series A round, when PitchBook data valued the company at $461 million.

TechCrunch noted that Anthropic is widely expected to list before the end of the year. If the company were valued at about $2 trillion, the highest of the market estimates, Schmidt’s holding could lift his fortune by up to $20 billion, which Bloomberg puts at $57.5 billion. Neither Schmidt nor Anthropic responded to TechCrunch’s request for comment.

Sources: Big Technology, TechCrunch, Techmeme (summarising The Verge)