AI Energy Use Could Surpass Bitcoin Mining by 2025

Artificial intelligence is on track to consume more electricity than Bitcoin mining by the end of 2025, according to recent research. This surge in energy demand is expected to account for nearly half of all electricity used by data centers globally. Despite advances in energy efficiency, the rapid growth in AI model size and deployment is driving unprecedented power consumption.

Alex de Vries-Gao, a researcher at Vrije Universiteit Amsterdam, highlights that AI already consumes up to 20% of data center electricity. However, precise measurement is difficult due to limited transparency from tech companies about AI-specific energy use. De Vries-Gao used a triangulation method combining chip production data, analyst estimates, and earnings reports to estimate AI’s growing power needs.

The trend reflects a 'bigger is better' mindset among tech giants, who continuously scale up AI models to stay competitive, increasing resource demands. This has fueled a boom in data center construction, especially in the United States, which leads globally in data center numbers. Energy providers are responding by planning new gas-fired and nuclear power plants to meet this demand.

The environmental impact is compounded by challenges in tracking AI’s energy footprint. While major companies report overall emissions, they rarely disclose AI-specific data, making it difficult to assess true environmental costs. This opacity hinders efforts to implement targeted sustainability measures.

De Vries-Gao’s estimates suggest AI’s electricity consumption in 2024 matched that of the Netherlands, with projections to reach the scale of the UK by 2025, consuming around 23 gigawatts. A separate report forecasts a 25% increase in U.S. electricity demand by 2030 driven largely by AI, traditional data centers, and cryptocurrency mining.

The complexity of AI’s energy impact is influenced by factors such as model size, query types, and the energy mix of local power grids. For example, identical AI queries can produce vastly different carbon footprints depending on whether data centers use renewable or fossil fuel energy sources.

The future of AI energy consumption hinges on whether the industry embraces efficiency over scale. Some models, like DeepSeek’s, demonstrate that AI can operate with significantly less power, challenging the prevailing trend of ever-larger, energy-intensive models. However, the Jevons paradox warns that increased efficiency might lead to greater overall usage.

Transparency and measurement are critical to managing AI’s environmental footprint. Without clear data, it’s nearly impossible to develop effective policies or technologies to curb energy consumption. The AI sector’s energy challenge is a call to action for both industry leaders and policymakers to prioritize sustainable innovation.