Pichai: Google accepts short-term Cloud revenue loss for AI compute
Google's CEO said on a podcast that DeepMind gets computing first, and that Gemini 4 Argon used about 25 trillion tokens in 72 hours.
Google CEO Sundar Pichai said the company is willing to give up some short-term Cloud revenue in order to direct computing resources toward advanced AI development, according to aitimes.com. Pichai made the remarks in a podcast interview with host Alex Kantrowitz.
Under the arrangement he described, Google DeepMind receives the computing it needs for research and next-generation models first, before resources are planned for other parts of the business. Pichai said the computing currently in use reflects decisions he made in 2024.
Pichai also gave usage figures for the company's latest model, Gemini 4 Argon, which he said consumed about 25 trillion tokens in its first 72 hours — three times the figure recorded at the launch of Gemini 3.0.
On internal impact, Pichai said development work saw productivity rise 30% and rollbacks fall by a third within four weeks of adoption. He described recursive self-improvement, in which AI helps build the next generation of AI, as still being at an early stage.
Longer-term context
Pichai said Google moved to an "AI First" approach in 2015, and noted that around the same time last year the rapid growth of coding agents had not yet begun in earnest.
Quick answers
What did Sundar Pichai say Google is sacrificing for AI?
He said Google is giving up some short-term Cloud revenue so that computing can be prioritized for advanced AI development.
How many tokens did Gemini 4 Argon use at launch?
Pichai said the model used about 25 trillion tokens in its first 72 hours, three times the figure at the Gemini 3.0 launch.
When did Google adopt an "AI First" strategy?
Pichai said the company shifted to "AI First" in 2015.