A Bitcoin energy estimate is a model, not a meter reading

Inherited archive entry · retained from the earlier edition. Its factual, legal and policy claims have not all been reverified for 19 September 2026. Research scope.

An estimate, built from a profitability model

The Cambridge Centre for Alternative Finance's published methodology describes its Bitcoin electricity index as a 'hybrid top-down approach': rather than measuring electricity use directly, it infers a plausible hardware mix from network hashrate and then asks which of over 100 known ASIC mining devices would still be profitable to run, on the assumption that 'miners are rational economic agents who only operate their devices for as long as they are profitable.' Total power demand follows from that inferred hardware mix, not from a meter reading anywhere on the network. The index's own overview page is explicit about this: the figure is 'an estimate' produced by a model, not a direct measurement of consumption.

What the model is standing in for

The quantity being modelled is real: Bitcoin's mining process requires hardware to iterate through nonce values, generating a hash for each attempt, until one is found below the network's difficulty target — a search whose cost scales with difficulty and is genuinely proportional to energy spent. What the methodology cannot observe is which specific machines are doing that search, where they are, what electricity they pay, or how efficiently they are cooled. It substitutes assumptions for each: a weighted profile of hardware released since 2013, weighted down as it ages and dropped after five years; a global electricity price near $0.05 per kilowatt-hour; and a power-usage-effectiveness factor for data-centre overhead ranging from 1.01 to 1.20 across its three scenarios.

Why the range is wide, and how to read one headline number

The methodology produces a lower bound (most efficient hardware, minimal overhead), an upper bound (least efficient hardware still profitable, higher overhead) and a best-guess figure between them, and states plainly that it 'depends on electricity cost estimates' that vary substantially by region and are not independently verified for the actual fleet. A single headline number reported without its bound discards the part of the estimate that discloses how uncertain it is. The documentation frames the best-guess line as more realistic than either extreme, not as a measured fact standing beside two hypothetical ones — the three numbers are one estimate with a stated error range, not three competing claims.

Questions to carry forward

• Is a cited energy figure the best-guess estimate, or has the range been dropped in favour of one number?
• What electricity price and hardware-age assumptions does the estimate rely on, and how sensitive is the headline figure to them?
• Does the source distinguish a modelled estimate from a directly metered measurement anywhere in its presentation?

None of this makes the exercise meaningless; a disciplined model beats no estimate at all. It does mean every figure produced this way inherits the assumptions built into it, and a reader who drops the range has quietly discarded the honesty the methodology built in.