Bitcoin (BTC) sustainability report

Bitcoin is present on the following networks: Bitcoin.

Bitcoin settles on a single transaction history through proof of work, the consensus design the network introduced. Participation is open: no registration, no approval, and no holding of the native asset is required to take part in block production. Participants who do produce blocks collect pending transactions, arrange them into a candidate block, and then hash that block's header repeatedly — applying the SHA-256 function twice — while varying a counter field, until the resulting value falls below a numeric threshold carried in the header itself. There is no shortcut to finding such a value; the expected number of attempts is what makes a block costly to create. That threshold is recalculated every 2,016 blocks from how long the preceding 2,016 took, which holds the long-run average spacing between blocks near ten minutes however much hardware is applied to the search.

Each participant running full validation software checks every incoming block on its own account: the signatures, the absence of double spending, the block's internal structure, and the correctness of the newly issued amount. A block that violates any rule is discarded regardless of how much computation stands behind it, so block producers cannot force an invalid state onto the network. When two valid blocks arrive extending the same parent, the network briefly carries two candidate histories, and the ambiguity clears as later blocks arrive, because participants follow whichever branch has accumulated the most total work. Settlement is consequently a matter of growing confidence rather than an instant guarantee: rewriting a past block would mean redoing its work and outpacing everyone else's ongoing effort, which becomes harder with every block layered above it.

Changes to the validation rules happen rarely and require wide coordination among operators, wallet software and block producers. Distinct from those rules is relay policy — which transactions an individual node forwards and holds in memory — which each operator sets locally. Competing node implementations with different policy defaults run side by side on the same network without splitting it, precisely because they judge blocks by the same validation rules.

Bitcoin is present on the following networks: Bitcoin.

Block producers are compensated from two sources. The first is a protocol-issued amount created in the first transaction of each block, which is the network's only mechanism for bringing new units into existence. That amount is cut in half every 210,000 blocks, roughly every four years; the reduction that took effect in April 2024 set it at 3.125 units per block, and the next is anticipated around 2028. The schedule terminates after a finite number of reductions, at which point block producers are paid out of fees alone. Newly issued amounts cannot be spent until a further hundred blocks have been built on top, which limits the damage a short-lived reorganization can do.

The second source is transaction fees. The protocol sets no price. A sender's fee is simply the difference between the value its inputs carry and the value its outputs assign, and senders choose it relative to the space their transaction occupies. Because each block can hold only a bounded amount of data, producers order pending transactions by fee per unit of space and fill the block from the top, so the clearing rate rises when demand outstrips capacity and falls when it does not. Witness data is counted at a discount against the block limit, which lowers the effective cost of the transaction formats that use it. A sender who has underbid can replace the transaction with a better-paying version, or have a recipient attach a well-paid follow-on transaction that drags the original along. Fees pass entirely to the block producer; none is destroyed.

There is no penalty machinery of the kind stake-based networks use — no confiscation of a bond, no suspension, no removal from a set — because there is no registered set of participants to act against. The deterrent is purely economic: computation spent on a block that turns out to be invalid or that loses the race earns nothing at all. Participants who validate without producing blocks receive no protocol payment and bear their own costs. Nothing is charged for storage over time, and script execution carries no separate metered charge; a more elaborate script is paid for through the additional space it occupies.

Bitcoin is present on the following networks: Bitcoin.

The consumption figure for this network rests on a top-down model built from the economics of block production rather than on a count of machines. The starting point is the hash function the protocol uses, which restricts the relevant hardware to purpose-built SHA-256 devices; those devices are cataloged with their rated hashing throughput and rated power draw. Revenue per unit of hashing is then derived from what the chain itself shows — issuance and fees actually paid out over the period. Set against an assumed cost of running a machine, chiefly electricity and facility overhead, that revenue defines a break-even efficiency. Devices too inefficient to cover their running cost at the observed revenue are treated as switched off and left out of the fleet.

The remaining devices are fitted to the network's observed aggregate hashing rate, weighted by plausible deployment shares, and their power draw is summed and extended across the reporting period. An allowance is added for the overhead of the facilities that house them, cooling most of all. Where the same physical hardware submits its work to more than one chain sharing the algorithm at the same time, that overlap is accounted for so a single unit of electricity is not charged twice. Where infrastructure for associated off-chain payment channels falls within the scope of the assessment, its consumption is added to the base-layer total. Apportioning a network total to an individual asset carried on the network is done in proportion to observed on-chain transfer activity for that asset.

None of this is metered. No facility is instrumented and no operator reports its bill; every element is an inference from publicly visible chain data and published hardware specifications. The result is most sensitive to three assumptions — the efficiency mix actually deployed, the electricity price used to set the break-even point, and the overhead multiplier — and a plausible range around any of them moves the outcome materially. Where evidence is thin, the choices made lean toward the higher estimate rather than the lower one, so that the figure is more likely to overstate impact than to understate it. Estimates are revised as observation improves.

Bitcoin is present on the following networks: Bitcoin.

The renewable share is not a property of the protocol; it follows from where the hardware securing the network physically sits and what the grids in those places generate. Locating that hardware is therefore the first step. It is approached through publicly observable signals: addresses and routing information advertised by reachable peers, the operating regions that block-producing pools disclose about themselves, and measurements gathered by open network crawlers together with purpose-built ones. The picture is partial by nature, since the machines doing the hashing are not the machines that announce themselves on the peer network, and operators have reasons of their own not to publish locations.

Where direct observation leaves gaps, the geographic profile of a network built on the same hash function and a comparable reward structure is substituted, on the reasoning that similar economics attract operators to similar places — cheap power, cool climates, and permissive siting rules. The resulting distribution is expressed as shares of estimated consumption by country or region.

Those shares are then combined with published statistics on how each region generates its electricity, giving a consumption-weighted renewable share for the network as a whole. The generation data used is Share of electricity generated by renewables, compiled and processed by Our World in Data from Ember and the Energy Institute's Statistical Review of World Energy. Energy intensity is expressed as a marginal quantity: the additional energy attributable to one further transaction over the period.

Two limits deserve stating. The grid figures are regional averages, so electricity bought under a direct supply contract with a generator, produced on site, or drawn from power that would otherwise have been curtailed or flared is represented only as whatever the surrounding region averages. And because blocks are produced on a schedule that does not respond to how many transactions are waiting, the per-transaction figure is an allocation of a largely fixed total rather than a measure of what one transaction causes. Both the geographic inference and the underlying statistics are revised as better data becomes available.

Bitcoin is present on the following networks: Bitcoin.

Emissions are derived from the same geographic breakdown that underpins the energy assessment. Once estimated consumption has been distributed across countries and regions, each share is multiplied by a factor describing how much greenhouse gas that region's electricity carries per unit of energy delivered, and the products are summed to give the network total. The factors come from Carbon intensity of electricity generation, compiled and processed by Our World in Data from Ember and the Energy Institute's Statistical Review of World Energy; that dataset is made available under the CC BY 4.0 license.

The two reported scopes mean different things. Scope 1 covers emissions released by sources the operators of the infrastructure own or control directly — fuel burned on their own premises, for instance in generators or heating plant. Running hashing hardware involves no such combustion in the ordinary case, and nothing in the protocol requires it, so no scope 1 emissions are attributed to the network and the figure is reported as zero. Scope 2 covers the emissions produced elsewhere, at the power stations that supply the electricity the infrastructure consumes. For this network scope 2 accounts for essentially the whole footprint, and it inherits every uncertainty already present in the consumption estimate and the location estimate, since it is the product of the two.

Greenhouse gas intensity is stated on the same marginal basis as energy intensity: the emissions attributable to one additional transaction over the reporting period.

Several boundaries are worth being explicit about. The method is location-based, attributing to each unit of electricity the average emissions of the grid it was drawn from; it does not reflect contractual arrangements such as renewable supply agreements or certificate purchases, which a market-based calculation would credit. Emissions embodied in manufacturing, shipping and eventually disposing of the hardware fall outside operational scope 1 and scope 2 reporting and are not included. The published carbon intensities are annual regional averages and lag the period they describe, and the generation that actually responds to an additional unit of demand may be dirtier or cleaner than that average.