Technology / Business

Add as a preferred source on Google

Google Gives Google DeepMind Compute Priority as 2027 Cloud Capacity Tightens

Pichai said infrastructure planning runs two to three years ahead. Google Cloud reported $24.8 billion in second-quarter revenue, up 82% year over year.

Updated 2 min read
Cooling equipment and power infrastructure stand at a data center campus in Council Bluffs, Iowa. / TokenPost.ai (macro)
Cooling equipment and power infrastructure stand at a data center campus in Council Bluffs, Iowa. / TokenPost.ai (macro)

Google and Alphabet CEO Sundar Pichai said Google gives Google DeepMind’s frontier research first claim on computing capacity, a choice that can limit near-term sales at Google Cloud.

Pichai made the comments in an interview with Big Technology Podcast host Alex Kantrowitz, published Oct. 20. He said Google DeepMind’s research and frontier-model development receive first call on the company’s computing resources. Google then plans for its other businesses. Google Cloud’s capacity plans account for forecasts, signs of customer demand and existing commitments, he said.

The allocation challenge reflects the time it takes to build infrastructure, Pichai said. Today’s computing supply reflects choices he made in 2024, and Google plans two to three years ahead. A new customer seeking capacity for 2027 would find it “incredibly difficult” to secure, he said. Google also leased short-term computing capacity in the preceding year to help some customers.

Pichai said the shortage extended across the industry, and companies would have preferred to plan for twice the computing supply. He described Google’s allocation choice as “forgoing short-term revenues.”

Alphabet reported Google Cloud revenue of $24.8 billion for the quarter ended June 30, 2026, up 82% from a year earlier. Pichai said Cloud backlog had reached $514 billion in his second-quarter remarks. The figures provide context for a business whose computing resources Pichai said must be balanced against Google’s research priorities.

Asked about lessons from railroad and fiber buildouts, Pichai said bottlenecks or economic shocks could lead to overbuilding. He said he expects planned computing capacity to find uses over time. Pichai also said frontier research can help Google anticipate capabilities it may need to prepare for.

Pichai did not estimate the revenue forgone, the computing capacity assigned to research or the number of Cloud customers affected. The interview identified no specific upcoming allocation decision or date when additional capacity would be available.

Related articles

View all