Cambridge Bitcoin Electricity Consumption Index (CBECI) — The Definitive Reference

Executive Summary

CBECI was launched in 2019 to replace a public debate that had been running on bad numbers. Estimates of Bitcoin’s electricity use at the time ranged across an order of magnitude depending on who was publishing and what they were trying to argue. CCAF’s contribution was a transparent, version-controlled model with documented assumptions, an open hardware list, and an honest treatment of uncertainty — a lower bound, an upper bound, and a best-guess rather than a single headline figure engineered to win an argument. That methodological discipline is why the index became the de facto reference: the IEA and Ember both substitute CBECI data into their global electricity models for recent years, and most corporate, financial, and regulatory references to Bitcoin’s energy footprint resolve, directly or indirectly, to Cambridge.

For a corporate practitioner, three things make CBECI worth understanding rather than merely citing. First, it is a measurement index, not an accounting standard — it tells you how much electricity the network draws, not how to allocate emissions to your balance sheet. The conversion from terawatt-hours to tonnes of CO₂e is a separate step governed by the CBNSI greenhouse-gas methodology and an assumed power mix, and conflating the two is the most common error in crypto-exposure disclosure. Second, it produces a range, not a point. The best-guess estimate is the headline, but the lower and upper bounds are part of the result, and a disclosure that quotes the best-guess as if it were a precise measurement misrepresents the index. Third, its figures carry a vintage: the model has been revised, historical numbers have moved, and the geographic distribution data in particular has a known staleness problem. A defensible citation states the figure, the date, and the model version.

Where CBECI Sits in the Carbon-Accounting Stack

Before the methodology is useful, the practitioner needs to see where a CBECI number travels. Like a national emission-factor dataset, CBECI is a Layer 3 factor source: it is not the corporate accounting methodology and it is not the disclosure regime, but it is the dataset those layers reach for when crypto-asset electricity is in scope. The chain below is the answer to “where did that Bitcoin energy figure come from?” — and it is the chain an auditor traces when a disclosure references digital-asset electricity or emissions.

Two consequences follow. First, when a corporate disclosure cites “Bitcoin uses around 175 TWh,” that figure is a Layer 2 model output, not a meter reading — it carries the model’s assumptions and uncertainty, and a careful citation says so. Second, the jump from Layer 2 (electricity) to Layer 4 (emissions or financed-emissions exposure) is not automatic: it requires a power-mix factor, and the choice of that factor — global average grid, mining-specific survey mix, or a marginal grid factor — can move the resulting CO₂e figure by a factor of two or more. That conversion step is where the electricity emission factor and the GHG Protocol Scope 2 Guidance location-based-versus-market-based logic re-enter the picture. CBECI gives you the kilowatt-hours; the emissions are your accounting decision on top of them.

What CBECI Is — and the CBNSI Family

CBECI is one component of a broader effort, and a recurring source of confusion is that three distinct Cambridge artefacts are routinely cited as if they were the same thing. Disambiguating them is the first discipline of citing Cambridge correctly.

All three are produced by the Cambridge Centre for Alternative Finance (CCAF), an independent research institute at Cambridge Judge Business School, through its Cambridge Digital Assets Programme (CDAP). The index and its data are published under a Creative Commons Attribution-NonCommercial-ShareAlike licence, which is what makes Cambridge data freely reusable in corporate reporting (with attribution) while restricting commercial redistribution. The underlying hardware list is public and open to comment, and every methodological change is recorded in a published change log — the transparency that distinguishes CBECI from proprietary “live counter” estimates.

The Two Headline Figures — Power and Consumption

The CBECI landing page displays two numbers for each estimate, and they answer different questions. Confusing them is a frequent reporting error.

The first figure, power demand in gigawatts, is the rate at which miners currently draw electricity — the load. It answers “how much power is the network pulling right now?” The second figure, annualised consumption in terawatt-hours, takes that power demand and assumes it holds constant for a full year, answering “if the network ran at this load for twelve months, how much energy would it use?” Because hashrate is volatile, CBECI applies a seven-day moving average to the consumption figure to make it more stable and more suitable for comparison against the annual consumption of countries or industries.

The practical consequence: the annualised TWh figure is a projection, not a record of energy actually consumed over the past year. Bitcoin’s load follows market cycles — when price rises, miners switch on more rigs and the load climbs; when price falls, older machines go dark. A TWh figure captured on a high-load day annualises a peak; the same figure captured during a downturn annualises a trough. For year-on-year reporting, the appropriate figure is CBECI’s cumulative annual total (the sum of daily consumption across the year), not a single annualised snapshot.

The Hybrid Top-Down Methodology

CBECI’s model, originally developed by Marc Bevand and substantially extended by CCAF, rests on a single behavioural assumption: miners are rational economic agents who run a piece of hardware only while it is profitable. From that assumption and the network’s measured hashrate, the model works backwards to the power that must be consumed to produce that hashrate. It is “top-down” because it starts from the network total rather than summing individual facilities, and “hybrid” because it blends that economic logic with a real-world hardware basket.

Model parameters

The profitability threshold

The model’s engine is a daily profitability test. For each hardware model in the basket, mining revenue (driven by block subsidy, fees, and the device’s share of network hashrate) is compared against operating cost (the device’s energy efficiency multiplied by the assumed electricity price). A device is “profitable” on a given day if revenue meets or exceeds electricity cost. This yields a daily profitability threshold — the minimum efficiency a machine must have to be worth running — and the set of all devices that clear it. As the network has grown and hardware has improved, that threshold has tightened relentlessly: miners must continually upgrade to more efficient ASICs to stay economical, which is why old hardware drops out of the profitable set over time.

To avoid the model reacting to short-term price and hashrate noise, a 14-day moving average is applied to the profitability threshold. The power the network must be consuming is then inferred from the hashrate and the energy efficiency of the profitable hardware set, with a PUE multiplier added for facility overhead. The result is a power-demand figure that, multiplied out over a year, gives the annualised consumption.

Lower Bound, Upper Bound, Best-Guess

Because the actual hardware mix and facility efficiency cannot be known, CBECI brackets the truth with two hypothetical extremes and then offers a realistic estimate between them. The three estimates differ in two parameters: which hardware the network is assumed to run, and what facility overhead (PUE) is applied.

The PUE assumptions matter and are deliberately conservative relative to general data-centre norms. A PUE of 1.0 would mean zero facility overhead — every watt goes to the ASIC. The lower bound’s 1.01 reflects near-perfect mining operations; the best-guess 1.10 matches what miners report in practice and is more efficient than a typical enterprise data centre (often 1.8 or higher); the upper bound’s 1.20 sits at the high end of what miners report. For the meaning and benchmarking of PUE itself, see the Uptime Institute PUE reference.

Key Assumptions and Why They Matter

Four assumptions drive most of the model’s output, and each is a point an informed reader — or an auditor — may probe.

Two further constraints are worth noting: only the three dominant ASIC manufacturers (Bitmain, MicroBT, Canaan, a combined share estimated at 85%+) are included for hardware released after mid-2014, removing “exotic” low-sales devices; and during any period with no profitable hardware, the model carries forward the last known profitable set rather than dropping to zero. These are the kinds of detail that make CBECI defensible — and the kinds of detail a thorough disclosure footnote can point to.

The Methodology Revisions — v1.1.0 to v1.4.0

CBECI is version-controlled, and the versions matter because they have changed historical figures. A 2021 number quoted from the pre-revision model is not the same as the same year quoted from the current model. The most consequential revision, in August 2023, lowered historical estimates after analysis showed the equally weighted hardware basket had overstated the influence of old, inefficient machines — particularly in 2021, when mining was exceptionally profitable and the old approach kept too much legacy hardware “on.”

The magnitude of the revision is the reason vintage discipline matters. The chart below shows the previous model’s annual estimates against the revised model for the years most affected — a gap large enough to be material in any disclosure that reused the older figures.

The 2021 figure alone moved from 104.0 TWh to 89.0 TWh — a 15 TWh reduction, roughly the annual electricity use of well over a million homes. The lesson for citation is unambiguous: a CBECI figure is only meaningful alongside the date it was retrieved and the model version that produced it.

From Electricity to Emissions — The GHG Index

This is the step practitioners most often collapse, and collapsing it is the largest single source of error in crypto-emissions figures. CBECI measures electricity. It does not measure emissions. Turning terawatt-hours into tonnes of CO₂e requires multiplying consumption by an emission intensity — and that intensity depends entirely on the electricity mix the network is assumed to run on. The same TWh figure produces wildly different CO₂e totals depending on whether you assume coal, hydro, or a realistic blend.

The separate CBNSI Bitcoin greenhouse-gas methodology brackets this the same way the electricity index brackets power demand: it presents a coal-only upper bound and a hydro-only lower bound as illustrative extremes, with a best-guess based on a realistic power mix in between. The realistic estimate requires knowing the actual energy mix of mining — and that is precisely what the daily model cannot observe, which is why the survey-based Digital Mining Industry Report became necessary.

For corporate reporters, this maps directly onto familiar territory. Choosing the power-mix factor for a crypto-asset exposure is the same decision as choosing a grid factor for purchased electricity — the Scope 2 location-based-versus-market-based choice — except applied to an activity you may not operationally control. If the exposure sits in a value chain or an investment portfolio rather than your own operations, it belongs in Scope 3, and the conversion from a CBECI-derived TWh figure to a reported tonnage should be documented with the same rigour as any other emission factor. See the grid emission factors reference for the intensity values that would drive that conversion.

The 2025 Digital Mining Industry Report

The April 2025 Cambridge Digital Mining Industry Report is the most significant addition to the Cambridge data since the index launched, because it replaces modelling with reported data for the variables the daily model cannot observe — energy mix and emissions. CCAF surveyed 49 mining firms representing roughly 48% of the implied network hashrate, with 41% of respondents publicly listed, headquartered across 16 jurisdictions and operating in 23 countries. For the energy-mix and emissions questions, this is primary operational data rather than inference.

The headline findings reset several long-standing assumptions. Sustainable sources — renewables at 42.6% plus nuclear at 9.8% — collectively supplied 52.4% of surveyed mining electricity, a marked rise from the 37.6% figure carried in earlier estimates. The fuel narrative also inverted: natural gas (38.2%) replaced coal (8.9%) as the largest single source, where coal had been 36.6% in 2022. On that reported activity, the report estimates annual electricity consumption of about 138 TWh and network-wide emissions of 39.8 MtCO₂e — roughly 0.08% of global greenhouse-gas emissions, comparable to a small European economy. The survey-based emissions figure sits well below the 69.6 MtCO₂e an IP-address-based model produced, and could fall to the 32.9–37.6 MtCO₂e range once the mitigating effect of using otherwise-flared gas is accounted for.

Surveyed mining electricity mix, 2024 reported data. Renewables + nuclear = 52.4% sustainable. Source: Cambridge Digital Mining Industry Report, April 2025. Shares are for the surveyed population (~48% of network hashrate), not the full network.

The report also documented operational flexibility that reframes mining’s grid relationship: surveyed miners curtailed 888 GWh of load during 2023, and 70.8% reported active climate-mitigation measures — evidence that large mining loads can act as flexible, interruptible demand that provides ancillary services to grid operators rather than purely as a drain.

Geographic Distribution and the Mining Map

The geographic layer is the most useful and the most dangerous part of the Cambridge data — useful because location drives the emission intensity that converts electricity to CO₂e, dangerous because the published distribution data has a known staleness problem. The CBECI mining map historically derived country shares from mining-pool IP data, and the most recent published mining-map snapshots predate the survey. The April 2025 survey reported the United States at 75.4% of reported mining activity, followed by Canada at 7.1% — a concentration far higher than older mining-map figures, reflecting both the post-2021 migration out of China and the survey’s North American respondent skew.

Why this matters for emissions: a network assumed to mine predominantly in a coal-heavy grid produces a very different CO₂e figure than one mining on the US or Canadian grid mix. Geographic distribution is therefore not a side fact — it is an input to the emissions calculation. The reporting discipline is to state which geographic dataset and vintage underlies any emission figure, because a CO₂e number built on a stale or unrepresentative geographic split inherits that error directly.

CBECI vs Other Estimates

CBECI is not the only Bitcoin energy estimate, and the spread between estimates is itself informative — it tells you how much the answer depends on methodology. The most-cited alternatives use different approaches, and the differences are not noise: they reflect genuine methodological choices about hardware mix, electricity price, and temporal averaging.

A practical note on the Digiconomist comparison: its per-transaction energy figure is widely quoted and widely misunderstood. Bitcoin’s energy use is driven by hashrate and miner economics, not by transaction count — adding transactions to a block does not meaningfully change the network’s power draw. Dividing total energy by transaction count produces a large, attention-grabbing “kWh per transaction” number that does not represent a causal cost of a transaction. CBECI deliberately reports network totals rather than per-transaction figures for this reason, and any disclosure should avoid the per-transaction framing as a measure of marginal impact.

How to Cite and Use CBECI in a Corporate Inventory

For most reporting entities, Bitcoin electricity is not a Scope 1 or Scope 2 line — it is a value-chain or investment exposure. A custodian holding crypto assets, an asset manager with mining-company equity, a payment processor settling in bitcoin, or a fund disclosing portfolio emissions all face the same question: how do I put a defensible number against this, and how do I cite Cambridge correctly? The discipline below is what makes a CBECI-derived figure survive assurance.

The conversion arithmetic itself is ordinary once the inputs are fixed: applied electricity (from CBECI) multiplied by an emission intensity (your chosen power-mix factor) gives CO₂e. The judgement is entirely in the inputs — which estimate, which vintage, which mix. A team running this calculation for purchased electricity in their own operations would do it through a Scope 2 electricity calculator; for a crypto exposure the structure is identical, with CBECI supplying the activity data in place of a meter reading.

Compliance and Disclosure Implications

Crypto-asset electricity is moving from a reputational talking point to a disclosed datapoint, and CBECI is the dataset most disclosure regimes implicitly rely on. Four frameworks are where this surfaces.

Beyond these, the EU’s Markets in Crypto-Assets (MiCA) regime introduces sustainability-disclosure requirements for crypto-asset service providers and issuers, for which a network-level energy and emissions reference is a natural input. Across all of them, the assurance question is the same one that applies to any emission factor: which dataset, which version, which vintage, and why was it appropriate for this exposure? An entity that cites “Cambridge says Bitcoin uses X” without the version, the date, and the electricity-versus-emissions distinction has a documentation gap that a reasonable-assurance reviewer will surface — the same gap that surfaces when any factor is cited without provenance under IFRS S2 or ESRS reasonable assurance.

Limitations and Common Misinterpretations

Six recurring errors surface wherever CBECI figures are reused. Each is the kind of mistake an informed reviewer catches.

Future Evolution

Three trajectories will shape the Cambridge data over the coming years. First, the shift from modelling to primary data begun by the Digital Mining Industry Report is likely to continue — firm-level survey data closes the gap on the variables the daily model cannot observe (energy mix, geography, mitigation activity), and repeated survey cycles will improve the emissions estimate’s grounding. Second, the convergence of mining with AI and high-performance computing is blurring the boundary of what the index measures; some mining facilities are pivoting to AI training and inference, and the energy-attribution question for shared data-centre infrastructure is becoming harder. Third, mitigation accounting — flared-gas use, waste-heat recovery, demand-response curtailment, behind-the-meter generation — currently sits outside the standard emissions estimate, and incorporating it credibly is an open methodological frontier that will narrow the gap between gross and net emissions figures.

For the corporate practitioner, the practical implication is that CBECI figures will keep moving — not only with the network, but with methodology improvements and better primary data. That reinforces the central discipline of this page: cite the figure with its date and version, separate electricity from emissions, and treat the range as the result. A figure that was correct in 2023 may be superseded by 2026, and a defensible disclosure is one that records exactly which figure it used and when.

Frequently Asked Questions

Sources and References

Every figure and methodological statement on this page reconciles to the primary sources below, retrieved June 2026. Figures are hardcoded against the dated source per Editorial Standards §9b — Standards pages are historical records and do not pull live values.

Primary Cambridge sources

• Cambridge Centre for Alternative Finance, Cambridge Bitcoin Electricity Consumption Index (CBECI): Methodology, model v1.4.0. ccaf.io/cbnsi/cbeci/methodology.
• Cambridge Centre for Alternative Finance, CBECI: Bitcoin GHG Emissions — Methodology. ccaf.io/cbnsi/cbeci/ghg/methodology.
• Cambridge Centre for Alternative Finance, Cambridge Blockchain Network Sustainability Index (CBNSI). ccaf.io/cbnsi/cbeci.
• Neumueller, A., Pieters, G. C., Mohaddes, K., Rousseau, V., & Zhang, B. Z., Cambridge Digital Mining Industry Report: Global Operations, Sentiment, and Energy Use, Cambridge Judge Business School / CCAF, 28 April 2025.
• Cambridge Judge Business School, Bitcoin electricity consumption: an improved assessment, August 2023 (the v1.2.0 revision and the historical-figure adjustments).

Corroborating and downstream sources

• International Energy Agency, Global Energy Review — cryptocurrency-mining electricity demand.
• Ember, Global Electricity Review, supporting materials (CBECI substituted for 2023–2024 crypto-mining demand growth).
• US Energy Information Administration, Tracking electricity consumption from U.S. cryptocurrency mining operations, 2024.
• De Vries, A., Gallersdörfer, U., Klaaßen, L. & Stoll, C., Revisiting Bitcoin’s carbon footprint, Joule, 2022.
• Stoll, C., Klaaßen, L. & Gallersdörfer, U., The Carbon Footprint of Bitcoin, Joule, 2019.

Related GreenCalculus references

• ML CO2 Impact Calculator
• Sustainable Web Design Model
• IEA Global Energy Review 2026
• Ember Yearly Electricity Data
• GHG Protocol Scope 2 Guidance
• GHG Protocol Scope 3 Standard
• CSRD / ESRS E1
• IFRS S2 Climate Disclosures
• SEC Climate Disclosure Rules
• PCAF Financed Emissions Standard
• Uptime Institute PUE
• Grid emission factors 2026
• Electricity emission factor
• Scope 2 emissions
• Scope 3 emissions
• CO₂e (carbon dioxide equivalent)
• Scope 2 Electricity Calculator

What changed in this revision

Published 20 June 2026. Initial publication. Documents the CBECI hybrid top-down model (v1.4.0), the lower-bound / best-guess / upper-bound structure, the key assumptions (0.05 USD/kWh electricity price, five-year hardware lifetime, release-date weighting, two-month deployment lag), the v1.1.0–v1.4.0 revision history and the August 2023 downward revision of historical figures, the separate CBNSI greenhouse-gas methodology and the electricity-versus-emissions distinction, and the April 2025 Cambridge Digital Mining Industry Report (52.4% sustainable share, 39.8 MtCO₂e, US at 75.4% of reported activity). Positions CBECI as a Layer 3 factor source upstream of IEA and Ember and of corporate crypto-exposure disclosure under CSRD, IFRS S2, the SEC rules, PCAF, and EU MiCA.