Google reports rapid growth in demand for artificial intelligence infrastructure and plans significant scaling to meet future needs.

Scale targets and operational constraints

According to a Google Cloud vice-president, the company needs to expand capacity by 1000 times within 4–5 years while keeping current cost and power envelopes stable.

Google currently observes that deployed compute resources are approximately doubling every six months, a pace that strains procurement and deployment processes across the industry.

«scale by 1,000 times in 4–5 years»

— said a Google Cloud vice-president.

Hardware bottlenecks and internal alternatives

Shortages of Nvidia GPUs are cited as a limiting factor for the rollout of new AI services and developer tools such as Veo.

To reduce reliance on external accelerators, Google is prioritising its own accelerators, notably the TPU seventh generation, which are custom silicon designed for machine learning workloads.

Industry outlook

Google warns that demand is outpacing infrastructure growth, and expects 2026 to be a particularly challenging year for the technology supply chain and capacity planning.

• Rapid demand increase forces cloud providers to reassess procurement timelines and energy budgets.
• Adoption of custom accelerators aims to diversify supply sources and improve long‑term predictability.
• Scaling efforts focus on maintaining unit economics and power efficiency while expanding capacity.

Google’s plan reflects a broader industry trend toward aggressive expansion of AI compute, requiring coordination between hardware vendors, cloud operators and energy planners to meet projected needs.