IEA Scenario Ties AI Boom to Doubling of Data Center Power Use by 2030
The IEA's World Energy Outlook 2026 links AI-driven demand to global data center electricity use doubling to 1,000 TWh by 2030 — with grids, not chips, now the binding constraint on AI growth.

Why it matters
- IEA's World Energy Outlook 2026, released November 19, 2026, shows data center electricity use doubling to 1,000 TWh by 2030 in its Stated Policies Scenario.
- Data center capacity approaches three times its current size by 2035 across the report's scenarios.
- Grid spending rises 75% over the next decade even in the Current Policies Scenario.
- Data center electricity demand grew 17% in 2025 versus 3% growth in global electricity demand, per an earlier IEA analysis.
- Conditional SMR power agreements for data center operators grew from 25 GW at end-2024 to 45 GW.
The International Energy Agency's World Energy Outlook 2026, released November 19, 2026, links AI-driven demand to global data center electricity consumption doubling to 1,000 terawatt-hours by 2030. That doubling sits inside the report's Stated Policies Scenario — a conditional pathway, not a forecast. The distinction matters, because the number is already shaping how utilities, regulators and hyperscalers plan the next decade of electricity infrastructure.
The report's scenarios show data center capacity approaching three times its present size by 2035 as AI and digital services expand. And according to the IEA, the bottleneck has shifted. Grids, equipment supply chains and, in some markets, public acceptance now constrain AI infrastructure growth more than computing itself.
For anyone building AI infrastructure, the message is blunt: power-system delivery has become a core constraint on expansion.
What exactly does the scenario say?
The IEA runs two main frameworks here. The Stated Policies Scenario incorporates policy measures formally proposed but not yet adopted. The Current Policies Scenario looks only at existing settings, with no further policy action.
The full outlook states plainly that none of its scenarios is a forecast. The 1,000-terawatt-hour figure is a conditional pathway built on specific assumptions about which policies get implemented.
That framing cuts both ways. It protects the agency from being wrong about a fast-moving sector. It also means the doubling could land lower if policies stall — or higher if AI adoption outruns the assumptions.
The takeaway in plain words: the IEA treats AI's power appetite as a planning problem, not a prophecy.
Why is grid investment the bigger story?
The St Policies Scenario includes 26 million kilometers of new power lines by 2035. Another 16 million kilometers of existing lines need replacement or refurbishment. Renewables supply half of global generation by 2035 in the same scenario, so the network buildout serves a changing generation mix, not just growing consumption.
Grid spending rises 75% over the next decade even in the Current Policies Scenario — the case where governments do nothing new. That investment requirement belongs to a much broader electrification wave. The full report projects more than 10,000 terawatt-hours of additional global electricity demand over the same period, with electric vehicles, cooling, heat pumps, industrial motors and data centers all contributing.
Data centers are one driver among several, but they're the most concentrated and the fastest-moving. The energy-security chapter projects 40% to 50% growth in peak electricity demand by 2035 across scenarios. Adequacy, the report argues, requires new capacity and robust grids. Flexibility calls for more storage, demand response and diverse generation.
The takeaway: AI's grid impact is real, but it arrives inside a system already rebuilding itself for electrification.
Can efficiency gains outrun demand growth?
Not so far. An earlier IEA analysis found that data center electricity demand grew 17% in 2025, against 3% growth in global electricity demand. That's a fivefold gap in growth rates.
The same analysis documented rapid declines in power consumption per AI task. But rising adoption and energy-intensive uses such as AI agents outweighed those efficiency gains at the system level.
This is the Jevons paradox at industrial scale: each unit of AI work gets cheaper in energy terms, and total consumption climbs anyway because there's so much more work. It's why "more efficient computing" and "fast growth in total electricity demand" can both be true at once.
The takeaway: efficiency per task is improving, but the IEA's data shows it can't hold total demand down.
How is the industry actually responding?
Procurement has moved fast. The earlier IEA analysis reported that technology companies accounted for about 40% of corporate renewable power-purchase agreements signed in 2025.
Nuclear is the more striking number. The pipeline of conditional agreements for data center operators to buy electricity from small modular reactor projects grew from 25 gigawatts at the end of 2024 to 45 gigawatts when that analysis was published. That's an 80% jump in conditional commitments within roughly a year.
Onsite natural gas is a third route, aimed at working around slow grid connections. Many such projects remain at early stages. The IEA's earlier analysis described rapid, large swings in AI data center demand that can stretch the technical capabilities of onsite gas plants. That makes battery storage increasingly important for reliable operation — and potentially for grid support, given suitable incentives.
Supply chains are the quiet constraint behind all of this. The earlier analysis found tightening supply of gas turbines, transformers, advanced chips and IT components, while planning and regulatory delays held up grid connections and other approvals.
The takeaway: developers are signing renewables, SMR deals and gas projects because the queue for grid connections is too slow.
Does more data center demand mean higher electricity prices?
Not necessarily — and the IEA is careful on this point. The earlier analysis found that demand growth doesn't automatically raise prices when the right policies and infrastructure investment are in place. Smart grid integration and incentives for flexible operation are among the available responses.
The challenge is concentration. Large, clustered data center loads can trigger investment in generation and networks, creating particular affordability challenges for the regions hosting them. Onsite generation and batteries offer partial workarounds, but the IEA's position is that grids, electrical equipment and flexible operation remain system-wide requirements.
What does the IEA say about AI as a tool, not just a load?
The report identifies a dual role for AI in the energy system. It could help optimize electricity systems and reduce costs. At the same time, current data center demand is amplifying strains in some markets and supply chains.
That framing matters for policy. AI isn't purely a burden on the grid in the IEA's reading — it's also a potential instrument for managing the very system it stresses. Which side dominates depends on deployment choices and market design, not on the technology itself.
Why this matters beyond the energy sector
The stakes run in two directions. For the AI industry, the IEA's analysis reframes the scarce resource: compute supply is expanding, but transformers, cables, copper and grid connections aren't keeping pace. For energy planners, data centers are the fastest-growing and least predictable line item in a decade that already demands 26 million kilometers of new transmission and a 75% rise in grid spending.
The report's scenarios suggest the binding question for AI growth through 2035 won't be whether chips get faster. It'll be whether the electricity system can be built quickly enough — and the IEA's own numbers show that system is already behind.
Source: unite.ai
More from Sophie Lindqvist
Show full bio
Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.
823 articles
Related articles
- OpenAI Tells White House AI Growth Hinges on Energy and Infrastructure
- Senate takes up permitting bill that would make AI data centers pay for their power
- Data Center Boom Faces a Bigger Risk Than Water or Power: Finance
- Google's Emissions Jumped 50% in 2024 to Feed Generative AI
- Jet Engines Pulled From Retired Aircraft Now Power AI Data Centers