Google is reserving computing capacity for Google DeepMind’s advanced AI research, potentially giving up short-term revenue from cloud customers.
Google Cloud generated $24.8 billion in revenue in the second quarter of 2026, up 82% year on year, while its contracted backlog reached about $513.9 billion.
Pichai says expanding capacity is challenging because infrastructure decisions take years to translate into available computing power, forcing Google to balance customer demand with AI development.
Alphabet CEO Sundar Pichai has acknowledged that the company is sacrificing potential short-term revenue from its cloud business to reserve computing capacity for developing advanced artificial intelligence models. Speaking recently on the Big Technology Podcast with Alex Kantrowitz, Pichai said Google was making computing-resource trade-offs to support frontier AI development, even when this meant foregoing immediate commercial opportunities.
The strategy highlights the challenge facing technology companies as they balance demand from paying cloud customers with the enormous computing requirements of increasingly sophisticated AI systems.
Google Deepmind Gets Priority As Demand For Computing Power Rises
Pichai said Google’s priority was to ensure that its AI research division, Google DeepMind, had access to the computing resources required to develop leading-edge models. He described that commitment as central to the company’s strategy, with other businesses having to plan around the resources needed for frontier research. The approach reflects Google’s belief that remaining competitive in advanced AI could strengthen its products and business over the longer term, even if it limits cloud revenue in the short term.
The decision comes as Google Cloud experiences rapid growth. The business generated $24.8 billion in revenue in the second quarter of 2026, an increase of 82% year on year, according to Alphabet’s results. Its contracted revenue backlog stood at approximately $513.9 billion at the end of June, reflecting strong demand for cloud services. However, Pichai said the company was constrained by the computing capacity available to it, and that accommodating customers seeking additional capacity for 2027 would be “incredibly difficult”.
AI Ambitions Intensify Pressure On Infrastructure
The shortage is not limited to Google. Developing and operating advanced AI models requires large quantities of specialised chips, servers, electricity and data-centre capacity. Expanding that infrastructure takes time, as companies must make investment decisions well before new computing resources become available. Pichai said the capacity Google has today reflects decisions made in 2024, underlining the long lead times involved in meeting demand.
Alphabet has signalled that it will continue investing heavily in infrastructure to support AI and cloud growth. The company’s second-quarter results showed that cloud demand was being driven by enterprise AI solutions and infrastructure, alongside its core cloud services. The challenge is to expand capacity quickly enough to serve customers while ensuring its researchers can continue developing advanced models.
Pichai did not disclose how much revenue Google Cloud was foregoing because of these allocation decisions. The financial trade-off therefore remains difficult to quantify, even as his comments make clear that Google is willing to prioritise its long-term AI ambitions over some immediate sales.
