Key Takeaways
How much energy does bitcoin mining use compared to AI data centers?
Both industries consume enormous amounts of power. Only one of them gets vilified for it. The comparison is more complicated than the headlines suggest, and the answer changes depending on whether you are looking at today’s numbers or where both industries are heading.
Bitcoin mining consumed roughly 138 TWh of electricity, about 0.5% of global consumption, and generated an estimated 39.8 million metric tons of CO2 equivalent. Those are large numbers in isolation. In context, they represent a network already drawing 52.4% of its power from zero-emission sources, up from 37.6% in 2022, with coal accounting for just 8.9% of the mix. [1]
Data centres tell a different story. The carbon footprint of AI systems alone reached an estimated 32.6 to 79.7 million metric tons of CO2 in 2025, on a steeper growth curve. Total data centre electricity consumption is projected to reach 945 TWh by 2030, nearly matching the entire current output of the US nuclear fleet. The hyperscalers buy renewable energy certificates to offset reported emissions, but their Scope 2 disclosures are rising faster than their clean energy procurement. [2]
The water comparison is not close. AI data centres consumed an estimated 560 billion liters of water globally in 2025 for cooling. A single large AI data centre can consume 5 million gallons of water per day. Bitcoin mining, including the air-cooled and hydro-cooled facilities BlockOps operates, draws negligible water by comparison.
What can bitcoin mining do that an AI data center cannot?
Turn itself off. This is the environmental argument for bitcoin mining that rarely gets made: flexible load is a genuine grid service.
Bitcoin miners can curtail 100% of their power draw within minutes and restart cleanly when conditions improve. No other industrial load does this. In West Texas, bitcoin mining operations absorbed 1.3 TWh of curtailed wind energy that would have otherwise been wasted. During Winter Storm Elliott, Texas bitcoin miners curtailed over 1.5 GW within minutes, stabilizing the grid. That kind of demand flexibility does not appear in any carbon accounting. What it costs the miner is a separate question, covered in do you get billed while your miner is curtailed.
AI data centres, by design, cannot do this. Inference SLAs are live. Models run continuously. The load is as rigid as it gets. As renewables scale and grid intermittency increases, the value of flexible demand grows. Bitcoin provides it. AI cannot.
Which one is worse for the environment?
Neither industry is clean. Bitcoin mining has a meaningfully better renewable mix than its reputation suggests and provides real grid value that does not show up on any ESG scorecard. AI data centres produce a larger absolute carbon footprint on a faster growth trajectory, and carry a severe water liability that is barely discussed publicly.
Over five years, the gap widens in bitcoin’s favor on the environmental axis. Bitcoin’s energy mix continues shifting toward renewables as miners chase the cheapest available power, which increasingly means stranded solar and wind. AI’s water consumption problem gets worse as inference demand scales and cooling requirements intensify.
The operator who bridges both workloads, running bitcoin on flexible power and deploying AI on the same infrastructure, produces the most efficient environmental profile possible. Bitcoin absorbs curtailed renewables. AI occupies the baseload. Neither leaves megawatts wasted. We looked at how that convergence is playing out in bitcoin mining and AI infrastructure.
BlockOps is building exactly that infrastructure in Arkansas. We own and operate all five facilities, so the power contracts, the cooling and the technicians all sit under one company. What to verify about any operator before you ship equipment is in our hosting due diligence checklist.