Wire flash
IEA: AI Data Center Power Demand to Double by 2030, Emissions Up 70%
Editorial responsibility
- No named human review is recorded for this page.
- Source reporting is collected, normalized, translated or condensed automatically when needed.
- Automatically published source-backed update
According to the International Energy Agency's World Energy Outlook 2026, global AI data center power demand is projected to double by 2030, driven by approximately $4 trillion in infrastructure spending. The report indicates that while 120GW of capacity currently exists, an additional 120GW is expected by the decade's end, with 75% dedicated to AI. This surge includes the rapid expansion of 93 large 'AI factories,' which could add a power load equivalent to a second Great Britain. Although renewable energy dominates power purchase agreements at nearly 50%, actual generation relies heavily on coal and natural gas (around 60%) for reliability, leading to a projected 70% rise in emissions from these facilities by 2030. The IEA notes that grid upgrades required to accommodate this growth may increase utility bills for consumers, though existing capacity usage could theoretically lower costs. Significant bottlenecks remain, as renewable plants and transmission lines take years to complete, often outpacing data center construction timelines.
Source report
The rapid expansion of AI data centers is placing unprecedented stress on electricity grids worldwide. Many regions are now facing demand from new projects that could exceed the current peak load capacity of their existing systems.
Capacity Projections and "AI Factories"
According to the International Energy Agency's (IEA) World Energy Outlook 2026 report, the landscape of global data center capacity is shifting dramatically:
- Current Capacity: Approximately 120 GW is currently available across all types of data centers globally.
- Future Growth: An additional 120 GW is expected to be completed by the end of the decade.
- 75% of this new capacity will be dedicated specifically to artificial intelligence.
- "AI Factories": The IEA is tracking 93 large-scale AI facilities currently under construction or being expanded, identified via satellite imagery.
- Capacity at these sites has surged from under 2 GW to 17 GW in just two years.
- Upon completion, these sites are projected to reach a total of 50 GW.
- This increase is equivalent to adding a second Great Britain's worth of power demand.
Economic Impact on Utility Bills
The massive power consumption required by these facilities has sparked opposition in local communities, with residents citing fears of increased utility prices. However, the IEA's analysis offers a nuanced view:
- No Clear Correlation Yet: The agency states there is currently no clear relationship between data center capacity growth and rising electricity prices.
- Potential for Lower Costs: If a data center utilizes existing grid capacity and pays its fair share of costs, it could potentially lower utility bills for all customers.
- Risk of Higher Costs: Conversely, if grid operators must upgrade infrastructure to accommodate these large consumers, monthly bills could rise if those upgrade costs are distributed across all users.
Emissions and Energy Sources
While renewable energy dominates data center power purchase agreements (accounting for nearly 50%), actual power generation relies heavily on fossil fuels.
- Generation Mix: Coal and natural gas sources account for approximately 60% of actual power generation.
- Reliability Drivers: This reliance on fossil fuels is driven by the need for reliability; gas and coal plants often activate when renewable sources cannot meet immediate demand.
- On-Site Generation: Some projects are deploying independent power sources. Notably, Elon Musk's Colossus 2 in Southaven, Mississippi, is part of a broader trend where roughly 180 GW of planned or under-construction capacity exists in the U.S.
- Emission Forecast: Despite the growth in renewables, the IEA expects emissions from coal and natural gas used by data centers to rise by 70% by 2030.
Investment Scale vs. Infrastructure Bottlenecks
Investment in AI data centers is projected to reach $32 trillion by 2050, surpassing combined infrastructure spending for railways, electrification, and the internet. Despite this capital influx, the power grid remains a critical bottleneck:
| Infrastructure Component | Completion Timeframe | | :--- | :--- | | Renewable Power Plants | 1 – 5 years | | Massive Data Center Projects | ~1 – 5 years (if not delayed by moratoriums) | | Transmission Works | 3 – 5 times longer than other components |
Due to these delays, some AI hyperscalers are exploring alternative solutions, such as installing jet engines on-site. However, even these alternatives face supply chain constraints, particularly regarding turbine blade shortages.
Related Industry Updates
- Cooling Technologies: The current state of play in data center cooling.
- Hardware Innovation: The status of custom AI ASIC development.
- Regulatory Landscape: How changing U.S. AI chip rules impact the global market.
- GTC 2026 Insights: Ian Buck (VP of Hyperscale and HPC) discusses shelving CPX and shipping LPUs this year.
- CPU Demand: A surge in data center CPU demand driven by AI agents.
Jowi Morales is a tech enthusiast with years of industry experience. He has been writing for various tech publications since 2021, focusing on tech hardware and consumer electronics.
Source
Latest from Tom's HardwareNeutral / independent
Part of this Story
IEA Projects Global Data Center Power Demand to Double by 2030