- Electricity availability is overtaking chips as the main AI bottleneck.
- The speed of grid investment will determine future AI leadership.
Energy in the Artificial Intelligence race is fast becoming the defining advantage for global technology leadership. Artificial intelligence depends on computing power. However, computing power depends on electricity. Therefore, reliable and affordable energy now shapes the future of AI competitiveness.
Data-centre investments already signal this shift. Increasingly, electricity availability determines where AI infrastructure can scale. As Albert O. Hirschman argued, economic power lies in controlling industrial choke points. Today, energy has become that choke point for artificial intelligence.
Previously, chips defined strategic advantage. The United States restricted advanced chip exports to China. Meanwhile, China leveraged control over rare-earth materials. However, as AI workloads expand, electricity constraints now overshadow chip supply. Data centres without steady power cannot operate, regardless of the quality of their hardware.
The International Energy Agency estimates that grid bottlenecks could threaten 20 per cent of planned data-centre capacity by 2030. Consequently, constrained supply will raise electricity costs. These costs will eventually affect firms and households alike.
China has acted decisively. It has invested heavily in energy generation and transmission. Much of this investment targets renewables and grid expansion. According to the Financial Times, China is now building solar, hydropower, and long-distance transmission projects at scale. It also manufactures energy hardware domestically, which dramatically cuts solar panel prices.
As a result, China can add between 500 gigawatts and one terawatt of capacity annually. Moreover, local governments support AI data centres through electricity subsidies. Facilities using domestic chips can reduce power bills by up to 50 per cent. This approach aligns industrial policy with local execution.
By contrast, the United States faces mounting challenges. China added 429 gigawatts of capacity in 2024. Meanwhile, the United States added far less. Although American firms plan massive data centres, grid expansion lags behind demand.
For example, proposed US data centres may require ten gigawatts of power. This demand rivals New York City’s summer peak load. However, transmission infrastructure takes nearly a decade to complete. Therefore, investment timelines remain misaligned.
Wholesale electricity prices already reflect strain. Bloomberg data shows that prices near data centre hubs are rising sharply. In some regions, costs increased by over 260 per cent in five years. Consequently, the United States needs urgent grid reform.
Europe also enters the picture. Energy in the AI race presents Europe with an opportunity. The European Union leads in the development of clean energy technology. Its electricity system remains highly interconnected. Moreover, EU policy treats grids as strategic assets.
However, Europe faces execution delays. Grid projects often exceed ten years. Permitting consumes much of that time. Current investment covers only a fraction of the required capacity. Hundreds of gigawatts of offshore projects still await grid connection.
Ultimately, energy will decide the AI race. China aligns energy, chips, and execution. The United States risks complacency despite its leadership in innovation. Europe holds potential but needs faster delivery.