Bitcoin Power Use Jumps 38% with Greener Energy Mix: Report

Bitcoin mining’s annual electricity consumption increased by an estimated 38% in the 18 months through December 2025, outpacing the rise in associated greenhouse-gas emissions as miners drew a larger share of their power from low-carbon sources, according to preliminary Cambridge research.

The network consumed about 190 terawatt-hours of electricity on an annualized basis in December, up from 138 TWh in June 2024, Alexander Neumueller of the Cambridge Centre for Alternative Finance said during a presentation at the inaugural Energy Investors Forum in Dallas.

Estimated emissions rose by a smaller 20%, from about 40 million metric tons of carbon-dioxide equivalent to 48 million tons, he said. The divergence reflects a shift in the reported electricity mix: low-carbon sources increased to 59.4% from 52.4%, with hydropower overtaking natural gas as the largest single source.

The findings complicate both sides of the debate over Bitcoin’s environmental footprint. Mining is consuming substantially more electricity while each unit of power comes from a less carbon-intensive mix. The result is still an increase in total emissions, but one much smaller than the growth in consumption.

Neumueller, a research associate who leads the Cambridge Digital Assets Programme’s work on the climate impact of digital assets, presented the figures as an advance look at the second edition of the Cambridge Digital Mining Industry Report. He said publication was expected later in 2026.

The figures should be treated as preliminary estimates rather than a complete measurement of the network. They are based largely on a survey of mining companies representing a little more than half of global computing power, or hashrate. Participation improved from the first edition, Neumueller said, but survey-based research can overrepresent companies and regions more willing to disclose data.

Cambridge’s 2025 report, which provides the baseline for the new study, acknowledged that heavy participation by US companies likely overstated the country’s share of global activity. That report estimated annual electricity use at 138 TWh and emissions at 39.8 million tons of CO2 equivalent. It found that 52.4% of surveyed miners’ electricity came from sustainable sources, including renewables and nuclear power.

The new results suggest the network’s expansion since then has outweighed continued gains in equipment efficiency. More computing machines are competing to secure the Bitcoin network, pushing total power use higher even as newer hardware performs more calculations per unit of electricity.

The reported rise in hydropower also reflects a geographic change. Neumueller cited increased survey coverage in hydro-rich markets such as Ethiopia, which has emerged as a mining location after adding generation from the Grand Ethiopian Renaissance Dam. That shift illustrates how estimates of Bitcoin’s global energy mix depend heavily on where mining can be identified and which operators participate in a survey.

The second part of Cambridge’s research examined a different transition: whether bitcoin miners are turning their power portfolios and data-center sites over to artificial intelligence.

Only about 10% of surveyed miners said they had already allocated some power to AI or accelerated-computing services, Neumueller said. Yet more than 40% of the remainder said they were definitely exploring diversification into AI and high-performance computing, while just 10% said they definitely were not.

“Intent to look into it is not commitment to deploy,” Neumueller said.

That gap is significant for investors who have increasingly valued mining companies on their access to land and electricity rather than solely on bitcoin production. Several publicly traded miners have announced AI hosting agreements or plans to convert sites, but the Cambridge findings indicate that actual industry-wide deployment remains limited.

Respondents identified financial resilience and a stronger market position as the main reasons to diversify. Capital expenditure was the largest obstacle, followed by a desire to maintain focus on mining.

“Miners do want to have a second leg to stand on, provided it does not destabilize the first,” Neumueller said.

Converting a mining site for AI is not a simple equipment swap. Bitcoin computers can operate in basic buildings, tolerate interruptions and shut down when electricity prices rise. AI servers generally require more sophisticated networking and cooling, higher reliability, and substantially more capital. A site with an electrical connection may still lack the fiber, water, engineering design or tenant credit needed to support an AI data center.

Cambridge’s survey nevertheless found that almost nine out of 10 respondents expected diversification into AI and HPC to gain traction across the industry over the next several years. Vertical integration into power generation and the provision of grid services were the next most commonly anticipated themes.

Those priorities converge around the same scarce input. “Power is becoming a scarce commodity, and many miners do already have it,” Neumueller said.

He stopped short of presenting the pivot as inevitable. Miners possess two ingredients AI developers need—land and electricity—but turning them into a different kind of data center requires “a lot of capital and a lot of expertise,” he said.