Methodology & data sources
Every number on this site can be traced to a public source. This page says which, under what licence, and where the numbers are soft.
Carbon & energy data — CCRI
Electricity, emissions and power figures come from the Crypto Carbon Ratings Institute (CCRI) public indices (indices.carbon-ratings.com), retrieved 2026-11-17. CCRI models each network from observable network data — hashrate and hardware efficiency for proof-of-work, validator counts and hardware profiles for proof-of-stake — combined with regional electricity emission factors. Their series are the same indicators used for MiCA sustainability disclosures in the EU.
Licence: CCRI publishes the indices data for free non-commercial use under CC BY-NC-SA 4.0. chain.eco is non-commercial — no ads, no sponsored listings, no paid placement — and attributes CCRI on every page that uses its data. Commercial reuse would require a CCRI licence; this site does not do that.
Directory & TVL data — DefiLlama
Chain names, chain IDs, native tokens, total value locked and protocol counts come from DefiLlama's free public API (api.llama.fi), snapshot 2026-11-17. DefiLlama tracks 424 chains in the snapshot used here.
Units
Source data arrives in per-network units (Bitcoin in TWh and Mt; smaller networks in kWh and kg). Everything is normalised before publication: electricity to kWh internally and displayed as TWh/GWh/MWh as magnitude requires; emissions likewise (Mt/kt/t/kg). "CO₂ per kWh" is the ratio of the two modelled series — it reflects the grid mix CCRI assumes for each network's nodes or miners, which is why two networks with similar electricity can have different emissions.
Caveats — read these before quoting a number
- These are models, not meters. No one meters a decentralised network. For Bitcoin, published annual estimates range from ~138 TWh (Cambridge CBECI / CCRI territory) to ~204 TWh (Digiconomist), driven by hardware-efficiency and electricity-price assumptions. We publish one consistent series (CCRI) so networks can be compared with each other, and we flag the range where it matters most.
- Annualised ≠ last calendar year. Figures are annualised from current average power draw; they move with hashrate and validator counts.
- No per-transaction league table. Dividing a network's total energy by its transaction count rewards spammy throughput and punishes batching and layer-2 design; the metric says more about the divisor than the network. We show totals and intensity instead.
- TVL is market weather. Directory TVL is a point-in-time snapshot (2026-11-17) and swings with prices.
- Coverage is partial by design. 37 networks are measured; 387 in the directory are not. An unmeasured network is not a zero-footprint network.
Corrections
Found an error? The dataset behind this site is published at /data.json so anyone can check our arithmetic against the sources. An LLM-readable summary is at /llms.txt, with the full text at /llms-full.txt.