Spending and sites

For a long time, software could grow without much attention to the buildings and power behind it. That is no longer a safe assumption for AI.

The International Energy Agency estimates that capital expenditure by five large technology companies exceeded $400 billion in 2025, and expects it to rise by another 75% in 2026. That spending is larger than global investment in oil and natural gas production. The IEA’s satellite tracking also finds that data centres built specifically for AI more than tripled in capacity over 18 months. Not every announced project will be built. The ones that proceed still have to wait on equipment, sites and connections.

Electricity makes the same point in physical units. The IEA’s updated projection has data-centre electricity use rising from about 485 terawatt-hours in 2025 to about 950 terawatt-hours in 2030, around 3% of global electricity demand by then. Use at AI-focused sites is expected to grow faster still, roughly tripling over the same period. The Agency also says bottlenecks in electricity supply, grid connections, chips, memory and planning make the more aggressive near-term scenarios less likely.

The timing difference

The timing difference is the substance. A software release can reach users in weeks. A grid connection, a turbine, a memory supply line or a data-centre hall takes years, and much of the equipment is already spoken for. Orders for new gas-fired plants in the United States reached a 25-year high in 2025, with data-centre demand part of the reason, and that demand is limiting turbine availability elsewhere.

Global energy investment is expected to reach $3.4 trillion in 2026. About $2.2 trillion of that is for renewables, nuclear, grids, storage, low-emissions fuels, efficiency and electrification, and about $1.2 trillion for oil, gas and coal. Electricity is already the centre of the spending. The digital build-out sits inside that larger reallocation, not beside it.

Through 2027, the pace of digital demand will depend on how quickly the physical systems underneath it can be permitted, equipped and connected.

Sources

IEA, Key Questions on Energy and AI, executive summary. Five technology companies’ capital expenditure exceeded $400 billion in 2025 and is expected to jump another 75% in 2026, exceeding global investment in oil and gas production. AI data-centre capacity more than tripled in 18 months on satellite tracking. Data-centre electricity from 485 TWh in 2025 to 950 TWh in 2030, about 3% of global electricity. AI-focused consumption expected to triple. Bottlenecks reduce the likelihood of more aggressive near-term scenarios.
https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary

IEA, World Energy Investment 2026. Global energy investment expected at $3.4 trillion in 2026. About $2.2 trillion to renewables, nuclear, grids, storage, low-emissions fuels, efficiency and electrification. About $1.2 trillion to oil, natural gas and coal.
https://www.iea.org/news/impacts-of-middle-east-conflict-set-to-reshape-energy-investment-plans-as-disruptions-put-focus-on-security