Global Installed Power Demand From AI Data Centers Will Exceed 100 Gigawatts By 2030.

Will AI Data Center Power Demand Exceed 100 GW By 2030?

Yes, according to multiple independent forecasts, the global installed power demand for data centers, including AI, is highly likely to surpass 100 gigawatts (GW) by 2030. Our probability estimate is 60%, based on the convergence of projections from Goldman Sachs Research, the International Energy Agency (IEA), and the Electric Power Research Institute (EPRI).

What Do Goldman Sachs And IEA Predict For Data Center Power?

Goldman Sachs Research projects that global data center power demand will rise from approximately 55 GW in 2026 to 122 GW by the end of 2030. This represents a 165% increase over four years. The IEA similarly forecasts that data centers will consume around 945 TWh of electricity by 2030, approaching 3% of global electricity supply. These figures are cited directly from their 2024 and 2025 reports.

Why Is The 100 GW Threshold Considered Safe?

The 100 GW mark sits below Goldman's central estimate of 122 GW. This creates a conservative buffer. Even if AI model efficiency improves faster than expected or grid connection delays occur, the baseline growth from training large language models and inference workloads still pushes demand past 100 GW. EPRI also notes that data centers could consume between 4% and 9% of US electricity by 2030, supporting the global trend.

What Could Slow Down This Growth?

Three main factors could reduce actual demand. First, efficiency gains in AI inference, such as smaller models or specialized chips, may cut energy per query. Second, permitting and grid interconnection queues can delay construction of new facilities. Third, some hyperscale projects may be cancelled or downsized due to financing issues. However, the sheer scale of announced projects, including single campuses requiring up to 2 GW, makes a full reversal unlikely.

How Does This Affect Energy Planning And Investment?

If 100 GW is exceeded, AI's energy footprint becomes central to grid planning. Utilities will need to accelerate investments in nuclear power, renewable energy, and battery storage. This is already visible in recent power purchase agreements from major tech firms. The DOE's 2024 report on data center energy use reinforces this view, citing annual growth rates above 15%.

Frequently Asked Questions

What Is The Current Global Data Center Power Demand?

As of 2024, the global data center installed capacity is estimated at 50 to 55 GW, according to Goldman Sachs Research. This includes both traditional cloud services and AI-specific infrastructure. The IEA provides a similar range in its 2024 Electricity Mid-Year Update.

Which Source Is Most Reliable For Tracking This Metric?

The most credible sources are Goldman Sachs Research, the IEA, and EPRI. Each publishes annual or semi-annual reports with transparent methodologies. For real-time tracking, the Lawrence Berkeley National Laboratory (LBNL) and the US DOE also offer quarterly updates on US data center loads, which can be extrapolated globally.

Does This Forecast Include Only AI Or All Data Centers?

The 100 GW threshold applies to all data centers, including AI, cloud, and traditional enterprise facilities. AI is the primary growth driver, but the metric is not AI-exclusive. This makes the threshold easier to verify, as total data center power is a standard industry statistic.