According to a preliminary study reported by theEnergyMag, annual electricity demand for Bitcoin mining will increase to approximately 190 terawatt-hours in December 2025, a 38% increase from 138 terawatt-hours in June 2024.
Alexander Neumuller of the Cambridge Center for Alternative Finance announced the numbers at the Energy Investor Forum in Dallas. Cambridge plans to publish the second edition of its Digital Mining Industry Report in late 2026.
The study also found that hydropower has overtaken natural gas as the largest single energy source for Bitcoin mining. Low-carbon electricity supplies 59.4% of the reported mining mix, up from 52.4% in the previous survey. However, total estimated greenhouse gas emissions still increased by 20%, from about 40 million to 48 million tonnes of carbon dioxide equivalent.
Hydropower accounts for the largest share of mine electricity
The 2025 Cambridge Digital Mining Report found that natural gas provided 38.2% of the electricity for surveyed miners, making it the largest single source of electricity at the time. Renewables provided a total of 42.6%, with nuclear power adding 9.8%. The share of coal fell to 8.9%, down from the initial estimate of 36.6% in 2022.
A preliminary update changes that order. Cambridge has not released a complete breakdown of each resource, but hydropower now exceeds natural gas. Neumuller linked some of the changes to increased research coverage in hydro-rich markets such as Ethiopia. Ethiopia has expanded its Bitcoin mining operations with the help of low-cost electricity from the Grand Ethiopian Renaissance Dam.
Electricity demand increases faster than emissions
The network’s annual electricity usage increased by approximately 52 TWh between the two reference points. Annualized demand measures how much electricity Bitcoin mining would use in one year if December 2025 rates continued. This does not mean that the miner consumed exactly 190 TWh during the 2025 calendar year.
Emissions grew more slowly than electricity demand, as miners reported using lower carbon electricity mixes. Still, Cambridge estimates it has increased from about 40 million tonnes of CO₂ equivalent to 48 million tonnes. Although the cleaner mix slowed the rate of growth in emissions, it was unable to offset the increase in overall electricity consumption.
More mining machines joined the network during the measurement period, increasing the total computing power. New hardware can perform more computations per unit of power, but the efficiency gains haven’t fully countered the increase in hashrate. The Cambridge Bitcoin Power Consumption Index tracks how prices, transaction fees, mining equipment, and network difficulty change estimated power demand over time.
Preliminary figures include survey limits
Cambridge based its new estimates primarily on responses from mining companies, which account for just over half of the global Bitcoin hashrate. The broader scope allows researchers to obtain a larger sample than the original report. However, some figures may be revised in the final publication after Cambridge completes further checks.
The 2025 report also warned that participation in surveys could skew geographic estimates. US companies accounted for the majority of responses, potentially overstating the US’ share of global mining activity. The recent increase in reported hydropower generation may partially reflect the increased coverage of miners in Ethiopia and other markets that rely heavily on hydropower generation.
A previous study from the University of Cambridge used a survey-based approach to estimate emissions at 39.8 million tonnes. Another location-based model yielded a much higher estimate of 69.6 million tonnes. This gap indicates that results are dependent on assumptions regarding mining location, power contracts, grid configuration, and use of stranded or flared energy.
Miners are exploring AI, but adoption remains limited
The new study also investigated whether Bitcoin miners are shifting power capacity toward artificial intelligence and high-performance computing. Approximately 10% of respondents said they have already allocated some level of authority to AI or accelerated computing. More than 40% of the remaining miners said they were actively considering this option.
Neumuller cautioned that “the intention to investigate does not guarantee implementation.” AI data centers require expensive networking, cooling, and reliable systems that basic Bitcoin mining sites may not have. If electricity prices rise, miners can quickly reduce the Bitcoin load, but AI customers typically require stable power and stronger service guarantees.
Still, nearly nine in 10 respondents expect AI and HPC to become more diverse in the coming years. As reported by crypto.news, publicly traded miners have already announced more than $70 billion in AI and HPC contracts in search of more stable revenue outside of Bitcoin production.
This change is already being reflected in the performance of some companies. TeraWulf generated more revenue from HPC hosting than Bitcoin mining in the first quarter of 2026. HPC services reported $21 million, while digital asset mining accounted for less than $13 million.
The Cambridge findings show that two changes are occurring simultaneously. Bitcoin mining uses more electricity, but hydropower and other low-carbon sources account for a larger share. At the same time, mining companies are evaluating whether their power connections and sites can support AI services. Cambridge’s full report is expected to provide a detailed energy breakdown and final methodology in late 2026.
