Global data center electricity consumption will nearly double from 485 TWh in 2025 to roughly 950 TWh by 2030, according to the International Energy Agency. That figure exceeds three times the entire annual electricity demand of the United Kingdom. Artificial intelligence is the primary driver, and the scale of capital now flowing into this infrastructure is without modern precedent.
Why AI Reshapes the Energy Equation
In 2024, data centers accounted for roughly 1.5% of global electricity, or about 415 TWh, per IEA data. That share looks modest until you examine the growth rate: 15% per year, more than four times faster than any other sector. The capital commitment matches the trajectory. Big tech capex surpassed $400 billion in 2025, with a further 75% increase expected in 2026, according to IEA projections. That puts tech infrastructure spending ahead of global oil and gas investment combined. All coverage of digital infrastructure is collected in our data center section, where the link to the blockchain world grows tighter by the month.
KEY FIGURES
Data center consumption 2025........... 485 TWh
Projection 2030.................... approx. 950 TWh
Share of global electricity 2030..... approx. 3%
Annual growth 2024-2030........... approx. 15%
Big tech capex 2025................ over $400 billion
Expected capex increase in 2026...... +75%
Source: IEA, Key Questions on Energy and AI, April 2026
How Much Energy Does AI Actually Consume?
Functionally, a single text query costs very little. The problem is scale and task type. Training a frontier model consumes tens of GWh. A large AI-dedicated data center draws as much power as 100,000 homes, and the largest facilities currently under construction target twenty times that. The IEA warns that agentic queries and video generation can cost hundreds or thousands of times more energy per request than simple text. The more AI agents go live, the faster the bill climbs. This dynamic is already visible with agentic AI entering banking through GPT-5.5, which multiplies model calls with every transaction. Ongoing analysis is available in our AI section.
Global Data Center Electricity Consumption (TWh)
Global data center electricity consumption, in TWh
Source: IEA (data 2024-2025, projections 2030-2035) · April 2026
Source: IEA (data 2024-2025, projections 2030-2035) · April 2026
The IEA has placed this issue at the center of the global energy debate: → read recent posts from @IEA on X.
Global electricity demand grew more than 2x as fast as overall energy demand in 2025, with electrification & AI expanding strongly
— International Energy Agency (@IEA) June 1, 2026
Energy demand growth in the US reached one of its fastest rates this century while growth in emerging economies slowed 👉 https://t.co/U5a5fJsS6H pic.twitter.com/F7JduiaFxq
Who Controls the Power, and at What Cost
The energy mix behind data centers tells a complicated story. In the United States, natural gas supplies over 40% of data center electricity, followed by renewables at 24%, nuclear at roughly 20%, and coal at 15%, per IEA figures. In China, coal dominates at close to 70%. The contradiction is hard to miss: companies with net-zero pledges are burning gas today while waiting on nuclear for tomorrow. That tension is driving the race by Microsoft, Google, and Meta toward reactors and small modular reactors. An entire adjacent industry is also transforming: Bitcoin miners are converting their facilities to AI data center operations, because hosting AI compute generates up to three times more revenue per megawatt than pure mining. Signed contracts in this space already exceed $65 billion.
For anyone tracking the mining sector, this is a structural shift. The question is no longer whether there will be enough power. It's who owns it.
What to Watch in the Coming Months
The defining test has a name: Stargate. OpenAI's joint project with Oracle and SoftBank targets 10 gigawatts and half a trillion dollars in investment, with the Abilene, Texas campus already running Nvidia chips. Half a trillion on a single compute infrastructure signals the scale of what is at stake. The real bottleneck won't be software. It'll be megawatts, grid capacity, and permitting timelines. The IEA makes one counterintuitive point clearly: despite the pace of growth, data centers will still account for only around 3% of global electricity demand in 2030. Significant and fast-growing, but not the energy catastrophe that some headlines imply.
