Filecoin (FIL) sustainability report

Filecoin is present on the following networks: Filecoin.

Filecoin is secured by Expected Consensus, a leader-election protocol in which a participant's influence is set by the storage it has committed and keeps proving, not by computation performed or by a balance of the native asset. The participants are storage providers. Each registers a miner actor on chain and accrues storage power as the sectors it commits are accepted, and that power counts toward consensus only once a provider holds more than a minimum threshold of it. Capacity holding verified client data carries a multiplied weighting, so elections use quality-adjusted power rather than raw capacity.

Time advances in epochs of thirty seconds. In each epoch every eligible provider takes a value from an external randomness beacon and evaluates a verifiable random function over it with its own key; compared against that provider's share of total power, the output decides whether it has won and how many times. Parameters are tuned so that several blocks, not one, are expected per epoch. A winner must then produce, within the same epoch, a proof covering a randomly chosen sector it stores, so the right to propose is tied to data actually held at that moment. Blocks of the same epoch built on the same parents are merged into a tipset, and fork choice follows the heaviest chain, whose weight rises with the total power behind it and with the number of blocks in its tipsets. Provable misbehavior, such as signing two blocks for one epoch, is a consensus fault that carries a penalty and suspends the offender from election for a period.

Expected Consensus on its own settles transactions only probabilistically, and clients treat a tipset as irreversible after 900 epochs, roughly seven and a half hours. Since April 2025 a separate finality layer called F3 has run on the main network alongside it. Providers vote in a Byzantine fault tolerant protocol weighted by quality-adjusted power, and a tipset endorsed by at least two thirds of that power receives a certificate beyond which it cannot be reorganized, usually within tens of seconds. Should the finality layer stall, block production carries on and settlement falls back to the longer probabilistic threshold.

Filecoin is present on the following networks: Filecoin.

Storage providers are the only participants the protocol pays, and they earn in two ways. The first is the block reward: a provider elected in an epoch that supplies the required storage proof is credited with freshly minted units of the network's native asset. Issuance has two parts. One follows a fixed schedule that decays smoothly with time; the other is released only as total raw storage on the network keeps pace with a baseline target that itself rises over time, which ties part of issuance to measured capacity rather than to the calendar. Most of each reward is locked and released gradually over about six months, and the locked balance can be drawn on to cover penalties. The second stream comes from clients, who pay providers to store and retrieve data under deals priced between the parties rather than by the protocol.

Committing storage requires collateral. Every sector carries an initial pledge in the native asset, sized from the power it adds, and deals can require provider collateral of their own. Capacity holding data from clients approved under the Filecoin Plus program counts tenfold toward quality-adjusted power, scaling rewards, pledge and penalties alike. Since April 2025 each sector also owes a small daily fee, fixed at commitment and capped relative to its expected rewards, which is destroyed.

Penalties come out of the provider's own funds. A sector that misses its daily proof accrues a fault fee for each day it stays offline, ending a sector early costs a fixed share of its pledge, and a consensus fault brings a penalty scaled to the block reward. Forfeited amounts are burned, and the protocol offers no delegation.

Every message pays gas for the computation and state access it triggers, including calls to user-deployed contracts. A base fee per unit is set algorithmically and destroyed; the sender adds a premium that goes to the block producer, and gas reserved well beyond actual use is partly burned too. A May 2026 upgrade made the base fee respond to premium levels rather than block fullness, because several producers per epoch often include the same messages.

Filecoin is present on the following networks: Filecoin.

The figure for Filecoin is a hardware estimate built around what storage providers physically run, because on this network the machines that take part in consensus are the same ones that hold client data. Sizing the population does not rest on crawling alone. Each provider appears as a miner actor in the on-chain power table, and its committed raw capacity, the sectors it seals and the proofs it submits are all publicly recorded, so the size and growth of the storage base can be read directly rather than guessed.

The estimate has to account for the different kinds of work this infrastructure performs. Keeping data stored is continuous: disks, enclosures and the servers in front of them draw power continuously in proportion to the capacity committed. Next comes sealing, the one-off encoding every sector must pass before it counts as power, which runs for hours on many-core processors and is completed on graphics processors, so its contribution follows the pace at which new capacity is onboarded rather than the amount already held. Generating the recurring storage proofs, running chain nodes and producing blocks add a further, smaller load. A representative device profile for each kind of work is derived from the documented requirements of the provider software and proving pipeline, power draw for those devices is taken from measurement, and an allowance for cooling and power conversion is added, since most capacity sits in data centers. Where a disclosure covers an asset that also circulates on other chains, those chains are estimated separately and a share is assigned from observed on-chain activity; the network's own total needs no such split.

The limitations follow from the method. On-chain records show how much is stored, not on what equipment, and drive generation, utilization and enclosure density vary widely between operators. Sealing can be outsourced to specialist facilities whose hardware is invisible from the chain. Nothing is metered at the wall. Where evidence is thin, the higher of the plausible values is used, so the result leans toward overstatement, and it is revised as observation improves. Independent research on this network has arrived at a similar split between storing and sealing, while noting how few operators supplied the measured constants.

Filecoin is present on the following networks: Filecoin.

The renewable share is derived from where storage capacity is located and how the grids serving those places generate electricity, not from the supply contracts operators hold. Locating infrastructure is more tractable here than on many networks. Each provider's miner actor records on chain the network addresses at which it can be reached for deals and retrieval, and those addresses, combined with peer-to-peer observation and public information about the facilities providers occupy, resolve to a country or region. The mapping is imperfect: a provider may expose an endpoint in one place while its disks sit in another, and a single operator can run many miner actors across several sites. Where part of the capacity cannot be placed, the spread observed on a network with a similar participant profile and reward design stands in for that remainder only.

Weighting counts for more than headcount. Consumption is dominated by stored capacity and sealing throughput, both concentrated among large operators, so each location is weighted by the energy attributed to the capacity found there rather than by the number of providers. A small number of large facilities can therefore move the result materially from one period to the next.

Each weighted location is matched to the annual generation mix of its grid, and the network-level share is those mixes averaged by the consumption assigned to each location. The method is location-based. The ecosystem has run programs through which providers bought renewable energy certificates or disclosed their own supply arrangements, but contractual instruments of that kind are not reflected, because they cannot be verified on a uniform basis from outside.

Energy intensity expresses the marginal energy linked to one additional transaction. On this network the gap between that marginal figure and a simple average is especially wide: nearly all electricity goes to keeping data stored and sealed, work that continues whatever the message volume, so one more transaction adds very little. Grid data comes from Share of electricity generated by renewables, processed by Our World in Data from Ember and the Energy Institute's Statistical Review of World Energy.

Filecoin is present on the following networks: Filecoin.

Emissions are computed from the consumption estimate together with the geographic distribution of capacity used for the renewable share. Each region's share of the electricity is priced at the average carbon intensity of its grid, stated in carbon dioxide equivalent to put every greenhouse gas on one scale, and the regional results are summed. Because consumption tracks stored capacity and sealing throughput, and both are concentrated in comparatively few large facilities, the total is sensitive to where those facilities are and which grids supply them; one large operator relocating or expanding can shift it more than the arrival of many small ones.

Scope 1 covers emissions from sources operators control directly, such as fuel burned in on-site generators. Storage facilities of this kind draw from the grid and run backup generation only rarely, so the category is typically immaterial and is recorded as nil instead of being omitted. Scope 2 captures the indirect emissions of bought-in electricity used to hold data, seal sectors and compute proofs, and it carries essentially the entire figure. Accounting is location-based, so certificates or supply agreements held by an individual provider do not lower the emissions attributed to it. Upstream emissions from manufacturing hard drives, servers and graphics processors fall outside the boundary; for a network whose footprint rests on very large volumes of storage media that are replaced as they age, that exclusion is worth bearing in mind.

Greenhouse gas intensity is the marginal emission linked to one further transaction, derived on the same basis as energy intensity, and it is small for the same reason: storing and proving data dominates consumption and does not scale with message traffic. It inherits and compounds the error in each upstream input: the mapping from capacity to hardware, the facility overhead allowance, the inferred locations and the grid averages. Carbon intensities come from Carbon intensity of electricity generation, processed by Our World in Data from Ember and the Energy Institute's Statistical Review of World Energy and made available under the CC BY 4.0 license.