How fast will data center electricity demand actually grow?

The International Energy Agency's central projection for global data center electricity demand is that it will roughly double by 2030: from 485 TWh in 2025 to 950 TWh, which it estimates would account for "around 3% of global electricity demand" by that date. The report notes this projection "remains close to the trajectory set out in the IEA's 2025 report," and that within it, "electricity consumption from AI-focused data centres grows much faster than overall data centre electricity consumption, tripling in this period." The scale of the underlying buildout is large: the IEA reports that capital expenditure by the largest technology companies "exceeded USD 400 billion in 2025" and "is expected to jump by another 75% in 2026," with the capex of "just five technology companies" now larger than global investment in oil and natural gas production. Growth has already accelerated: data center electricity demand "grew by 17% in 2025," while AI-focused data center consumption alone surged "50%."

The IEA ties this demand path to a buildout it can observe directly: its "unique satellite-based tracking shows that 'artificial intelligence (AI) factories'... have more than tripled in capacity in the past 18 months," and roughly one-fifth of US data center projects relying on onsite natural gas power "have started land clearing or construction." The agency flags this growth as running ahead of the systems meant to support it: "bottlenecks across energy supply chains and advanced chip manufacturing have tightened," a high-bandwidth memory shortage "has developed over the past six months and is anticipated to persist through at least the end of 2027," and the report treats these bottlenecks as a moderating force, reducing "the likelihood of more aggressive near-term scenarios, despite booming investment." It also notes the buildout now exceeds company balance sheets, making the pace of growth "sensitive to market sentiment" and financing conditions.

This macro-level demand path sits alongside Epoch AI's facility-level breakdown in What actually drives AI data center costs?, which finds energy is a small fraction of a single data center's annualized cost ($0.6 billion of $8.5 billion) — a reminder that the IEA's doubling of grid-level electricity demand is a supply-side physical constraint distinct from, and not contradicted by, the finding that servers rather than power dominate a facility's dollar cost. The IEA's bottleneck framing — energy supply chains and chip manufacturing tightening together — also echoes the compute-supply concern in Will AI inference demand outpace supply capacity soon?, though that note's bottleneck is inference capacity rather than electricity specifically.

The report is explicit that its own demand path depends on "three rapidly evolving and uncertain trends" it cannot fully resolve — efficiency, uptake and changing model capabilities — and that "not all projects will come to fruition," so the 950 TWh figure is a central projection, not a forecast with narrow error bars. It does not quantify how much of the realized buildout depends on the financing conditions it flags as a sensitivity, nor does it say whether a slower capex trajectory after 2026 would pull the 2030 figure down proportionally.