Business

Data Centre Energy Crisis: How AI Computing and Crypto Mining Compete for Power

Illustration of data centre with mining rigs and cooling systems
Data centres are consuming an increasing share of global electricity. AXT News

Global data centre electricity consumption exceeded 1,000 terawatt-hours in 2025 for the first time, according to the International Energy Agency. This figure is roughly equivalent to the entire electricity consumption of Japan and represents approximately 4% of global electricity use. The two fastest-growing sources of demand are artificial intelligence computing and cryptocurrency mining.

The Scale of the Problem

A single large AI training cluster - the kind used by companies like OpenAI, Google, and Meta to develop language models - can consume 50 to 100 megawatts of power. For context, 100 megawatts is enough to power approximately 80,000 homes. A major Bitcoin mining facility typically consumes between 30 and 200 megawatts.

Microsoft, Google, and Amazon have collectively committed to building data centres with a combined power demand exceeding 5 gigawatts in the United States alone over the next three years. This has created competition with local communities for grid capacity, particularly in regions like northern Virginia, central Texas, and parts of Ireland where data centres are concentrated.

Crypto Mining's Energy Footprint

Bitcoin mining consumes approximately 150 terawatt-hours per year globally, according to the Cambridge Bitcoin Electricity Consumption Index. This figure has remained relatively stable since the 2024 halving, which reduced mining rewards and forced less efficient operations to shut down.

The geographic distribution of mining has shifted significantly. China's 2021 ban on cryptocurrency mining pushed operations to the United States (estimated 38% of global hash rate), Kazakhstan (13%), Russia (11%), and Canada (7%). Within the US, Texas is the largest mining hub, attracted by deregulated electricity markets, cheap natural gas, and bitcoin-friendly state policies.

The Conversion Trend: Mining to AI

Some cryptocurrency mining companies have begun converting their facilities to serve AI computing workloads. The economics are straightforward: a GPU used for AI inference or training can generate more revenue than the same hardware used for crypto mining at current difficulty levels and Bitcoin prices.

Marathon Digital, one of the largest publicly traded Bitcoin miners, announced in late 2025 that it would dedicate approximately 20% of its computing capacity to AI cloud services. Core Scientific, which emerged from bankruptcy in 2024, has signed a 12-year agreement with CoreWeave to host AI computing infrastructure at its former mining sites.

Renewable Energy and Nuclear

The energy demands of data centres have accelerated interest in new power sources. Microsoft signed an agreement to restart a unit at the Three Mile Island nuclear plant in Pennsylvania specifically to power its data centre operations. Amazon has purchased a data centre campus adjacent to a nuclear power plant in Pennsylvania.

In Scandinavia, cryptocurrency miners and AI companies have been attracted by abundant hydroelectric power. Iceland, with its geothermal energy resources, hosts several mining operations that can claim virtually carbon-free electricity. These examples suggest that the data centre industry could become a significant driver of clean energy investment.

Regulatory Responses

Governments are beginning to respond to the energy demands of data centres. Ireland introduced a moratorium on new data centre connections to the grid in 2022, which was partially lifted in 2024 with conditions requiring operators to provide their own electricity generation. Singapore has similarly limited new data centre construction pending energy reviews.

In the United States, the Department of Energy has proposed new efficiency standards for data centre cooling systems, which account for approximately 30-40% of a facility's total energy consumption. The Environmental Protection Agency has also expanded its Energy Star certification programme to include AI-specific computing equipment.