Most companies would do almost anything for more revenue. Google is choosing to pass on some of it.
In an interview on October 20, 2026, Google CEO Sundar Pichai said Google Cloud is giving up short-term revenue so it can back frontier AI development. Limited compute is being steered mainly toward Google DeepMind’s most advanced projects rather than sold to the highest bidder.
This matters because compute has become the scarcest input in tech. When you can’t build data centers fast enough, every chip becomes a choice: sell it today, or use it to build something bigger tomorrow.
The trade-off, by the numbers
Google Cloud is not struggling. It is turning away money while growing fast.
The division’s revenue rose 82% year-over-year to $24.8 billion in Q2 2026. Its operating margin came in at 35.6% for the same quarter.
Then there’s the backlog. Google Cloud reported a backlog of $514 billion as of October 2026.
That is the core tension Pichai described. Every unit of compute handed to DeepMind is a unit that could have been rented to an outside cloud customer for immediate revenue.
Spending its way out of scarcity
Alphabet is guiding for $195–205 billion in capital expenditures for 2026. Most of that is earmarked for AI infrastructure.
That spending is meant to serve two hungry groups at once. One is Google’s internal AI work. The other is external cloud customers who want access to AI capabilities.
Until that new capacity comes online, someone has to decide who gets what. According to the research findings, compute scarcity has caused allocation conflicts inside Google, and co-founder Sergey Brin has reportedly been involved in those decisions.
A strategy a decade in the making
Pichai set Google on an AI-first course with a directive in 2015.
Google Cloud is not being run purely as a standalone profit engine. It is also serving as the infrastructure backbone for DeepMind’s research ambitions.
What this means
For investors, the most notable detail is that Google is making this trade from a position of strength. Growth of 82% and a margin of 35.6% give the company room to prioritize research without the cloud business looking weak.
The $514 billion backlog also works as a cushion. Committed business can help smooth out results during a period of heavy infrastructure spending and internal compute rationing.
The capex guidance is the other key metric to track. Spending of $195–205 billion in a single year is a massive commitment. Its success depends on whether new capacity arrives fast enough to ease the internal tug-of-war.
Cloud buyers who need capacity now may weigh how much priority they get relative to Google’s own research. In a market where compute is scarce everywhere, that kind of allocation choice could affect how customers think about their providers.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.