Bitcoin is a chain of decisions · The cost of proof-of-work
Bitcoin Mining and the Grid: A Consumer That Can Switch Off
Mining is described both as a threat to power systems and as their ideal helper. Both claims are too convenient. A miner can shut down thousands of machines faster than a factory, but those machines first require a large electrical connection. Benefit or harm comes not from the technology’s name but from location, generation source, contract, and behavior during scarcity.
Why Bitcoin needs electricity
Miners vary block-header data to find a hash below the target. One attempt says almost nothing about the next, so advantage comes from the number of computations. Electricity pays for the physical work behind proof-of-work.
A winning block may earn a reward while all other attempts create no separate product. Yet calling every losing attempt useless is also incomplete: their aggregate cost is part of what makes rewriting the history expensive.
The scale, without convenient minimization
The Cambridge Centre for Alternative Finance’s 2025 study estimated annual network consumption at roughly 138 TWh, about 0.54% of global electricity use. This is an estimate, not one meter reading: the location and operating schedule of every miner are unknown.
Hardware efficiency improved by about 24% year over year, yet total consumption still rose. Cheaper computation attracts new hashrate and network difficulty adjusts. Better joules per terahash therefore do not guarantee a smaller total load.
The energy source matters more than one total
The same mining unit can run on hydro, gas, coal, nuclear power, or a grid mixture. Identical electrical consumption can carry very different emissions and local-system consequences.
The Cambridge survey reported 52.4% sustainable sources, including renewables and nuclear power. Its sample represented firms with about 48% of network hashrate and may overrepresent North America. It is an important industry snapshot, not a precise map of every machine.
Why mining really is flexible
ASIC machines can be switched off in separate groups without destroying a half-finished physical product. Other participants continue finding blocks, and difficulty eventually adjusts if hashrate stays away.
For a grid, that is an unusual property in a large customer. When demand and wholesale prices surge, an operator can curtail mining within minutes and leave more capacity for other loads. The machines can return later.
What happened in Texas
ERCOT created a voluntary curtailment program for large flexible customers and explicitly named bitcoin mining facilities. Participants can reduce consumption during high-demand periods; some sites also react to high wholesale prices without a special instruction.
That is a real demand-response mechanism, not a marketing metaphor. It does not mean previously consumed electricity is “returned to the grid.” Curtailment simply removes part of demand at that moment, making it easier for available generation to serve everyone else.
Why curtailment is not guaranteed
The decision is economic. When expected mining revenue is high, an operator may continue through a higher electricity price. A recent analysis of ERCOT found exactly this relationship: aggregate load was less responsive when hashprice was high.
The 2026 study is still a preprint. It does not erase observed flexibility, but it warns planners not to count all connected mining as an equally reliable demand-response reserve.
Large loads create risks of their own
A new site needs lines, substations, and available generation. If it arrives where the network is already constrained, it can deepen peaks, require new infrastructure, or affect local prices.
NERC identifies less obvious issues with large electronic loads: forecasting errors, abrupt simultaneous loss of load, and possible oscillation risks on weak portions of the grid. ERCOT is therefore tightening interconnection studies and visibility requirements for large sites.
“Surplus energy” is a claim to prove
Mining can sit next to generation that is hard to transmit or buy power during periods of excess wind or water. In that specific setting, a flexible customer may create revenue for energy that would otherwise be curtailed.
The word “surplus” cannot be applied to an entire industry. If a miner runs during scarcity, extends a polluting plant’s life, or competes for constrained capacity, the effect is different. Hourly source and contract data matter more than an annual average.
The honest conclusion
Mining is a large and unusually controllable consumer. It can participate in demand response better than many industrial processes, but it does not generate electricity or make a grid resilient merely by connecting.
The useful question is not whether mining is always good or bad. Ask where the power comes from, what infrastructure had to be built, when the site curtails, and who pays the consequences. Only then does “flexible” have real meaning.
A flexible load is still a load. Its value appears only where curtailment genuinely helps the system.
Sources and verification
Educational overview. Energy-use estimates carry uncertainty, and the effect of a particular site depends on its local power system. The 2026 hashprice study is identified as a preprint and may change after peer review.
← Back to the rubric