Starknet (STRK) sustainability report

Starknet is present on the following networks: Ethereum, Solana, Starknet.

Ethereum reaches agreement through proof of stake, adopted in September 2022 when the original mining-based chain was retired in favor of a validator-driven consensus layer. The protocol family is usually referred to as Gasper. A fork-choice rule named LMD-GHOST selects the head of the chain by following the branch carrying the greatest accumulated weight of validator votes, while a separate finality gadget, Casper FFG, periodically justifies and then finalizes checkpoints, so that reversing them would require destroying an enormous quantity of bonded value.

Time is divided into slots of twelve seconds, and thirty-two slots form an epoch. For each slot the protocol pseudo-randomly designates one active validator to assemble and publish a block, and assigns the rest to committees that vote on what they believe is the correct head and the correct checkpoints. Under healthy conditions a checkpoint becomes final two epochs after it is proposed, a little under thirteen minutes, after which everything beneath it is treated as settled.

Joining the validator set requires a deposit of no fewer than 32 units of the native asset. Since the protocol upgrade of May 2025 a single validator may hold a far larger balance, up to 2,048 units, and earn on the whole of it, which lets an operator running many minimum-sized validators consolidate them into fewer; the activation floor itself did not change. Entry and exit are rate-limited by a queue measured in staked weight rather than in validator headcount, which bounds how fast the composition of the set can turn over.

Security rests on voting power being bonded. A validator that signs contradictory messages can be proved to have done so and is penalized, and the size of that penalty scales with how much other stake was penalized at the same time, so a coordinated attack is punished far more severely than an isolated fault. Should the chain stop finalizing altogether, a separate mechanism gradually erodes the balances of validators that are not participating until the remainder again represents a large enough majority to finalize. Upgrades during 2024 and 2025 changed how large data payloads are distributed and sampled between nodes, without altering this underlying agreement process.

Solana runs a proof-of-stake network in which the right to produce a block is allocated in proportion to the quantity of the native asset staked to each validator. What sets the design apart is that ordering is established before agreement is sought. A designated leader runs a sequential hash chain, each output feeding the next input, so the chain cannot be computed faster than a fixed number of steps and a transaction's position within it is evidence of when that transaction was received. This construction, called proof of history, spares validators from negotiating timestamps with one another and lets the rest of the protocol treat the order of events as already settled.

Leadership is not auctioned block by block. At the start of each epoch, which runs for roughly two days, a schedule is derived deterministically from the active stake distribution and assigns every short slot in the epoch to a named validator. Slots follow one another a few hundred milliseconds apart. The scheduled leader gathers transactions, executes them and streams the resulting block to the rest of the set in small fragments relayed through a tree structure rather than pushed to every peer at once. Receiving validators replay the block independently and publish votes for the fork they consider canonical.

Fork choice weights those votes by stake, and each vote commits a validator to its chosen fork for a period that doubles with every further confirmation, so abandoning a block grows steadily more costly. A block that a supermajority of stake has voted on is treated as confirmed within about a second, and it is locked in permanently once enough additional confirmations accumulate, which takes on the order of ten seconds. Safety rests on the assumption that participants acting dishonestly control less than a third of staked value. A revision of the voting layer, approved in a stake-weighted validator vote, is being activated on the main network in stages; it retains stake-weighted validation and the existing block distribution scheme while replacing the incremental lockout rule with direct voting that settles in one or two rounds.

Starknet is a validity rollup that settles to Ethereum. Transactions are executed away from the settlement layer, and their correctness is established by a cryptographic proof rather than by a challenge period: a prover produces a succinct argument that a batch of transactions was executed according to the rules, a contract on Ethereum verifies that argument, and once verification succeeds the resulting state is final. There is no window during which a submitted state can be disputed, because an invalid state transition cannot be proven in the first place. The proof system relies on collision-resistant hash functions rather than elliptic-curve assumptions.

Ordering and execution are the responsibility of sequencers. A 2025 upgrade replaced the single sequencer with several running a Byzantine fault tolerant agreement among themselves at a majority threshold, which cut block times substantially and introduced pre-confirmations that give users a response well before a block closes. Those sequencer nodes remain operated by one organization, so the arrangement removes a single point of failure in the software sense without yet making participation open. Proving likewise remains centralized.

A staking system is being introduced in stages and is the route by which that changes. The first stage opened staking and delegation to anyone willing to run a full node; the second added attestation, which makes a validator's liveness and reliability publicly measurable; the stage now being rolled out has validators validate and vote on sequenced blocks, with a block finalized only once more than two-thirds of stake has voted for it; a final stage would hand validators responsibility for operating the network, with proof generation expected to remain the last centralized component. Published schedules for the later stages have moved, and the network should be described as partially decentralized: state validity is enforced by proofs verified on Ethereum and does not depend on trusting any operator, while liveness and transaction ordering still do. Data is published to the settlement layer, so the information needed to reconstruct state is available independently of the operators.

Starknet is present on the following networks: Ethereum, Solana, Starknet.

Payment inside the protocol flows to validators, the only participants the consensus layer compensates directly. A validator earns newly issued units of the network's native asset for voting promptly and correctly on the head of the chain and on the checkpoints being justified, for serving its turn in the committee that signs headers for light clients, and, when selected to propose, for the block itself. The proposer additionally keeps the priority portion of the fees in that block, together with whatever it receives from the separate market through which many proposers outsource block assembly. There is no delegation inside the consensus rules: stake is either operated directly or entrusted to an operator through arrangements that sit outside the protocol.

Users pay for execution in gas, metered per operation, with writes to persistent state priced far above arithmetic. Every transaction carries a base fee per unit of gas that the protocol sets algorithmically from how full recent blocks have been, and that amount is destroyed rather than paid to anyone, so sustained demand withdraws native asset from circulation. On top of it a user adds a voluntary tip, which goes to the proposer and governs how quickly the transaction is picked up. Data posted on behalf of Layer 2 networks is priced in a second, independent market whose fee is likewise destroyed; a December 2025 upgrade tied the floor of that market to ordinary execution costs so it cannot collapse to a negligible level, and capped the gas any one transaction may consume.

Penalties mirror the rewards. Failing to vote, or voting late or incorrectly, costs a validator roughly what correct behavior would have earned it. Provable equivocation is treated far more harshly: the offender is scheduled for ejection, forfeits part of its balance immediately, and later incurs an additional correlated penalty computed from how much other stake was penalized nearby in time. Prolonged absence while the chain is failing to finalize drains balances until finality can resume. Stakers may take out accumulated rewards without leaving the set, and since 2025 may also trigger a full exit from the execution layer rather than only from the consensus client.

Two streams of payment reach validators. The protocol issues new units of the native asset on a defined schedule and distributes them at the close of every epoch to validators and to the stake delegated to them, in proportion both to that stake and to the voting credits the validator actually accrued over the epoch; an operator that missed its slots or stopped voting accrues fewer credits and receives a correspondingly smaller share. Holders who do not wish to run hardware delegate through a stake account to an operator of their choice, retain control of that account, and receive the reward net of whatever commission the operator has set. Delegated stake becomes active and inactive only at epoch boundaries, so capital committed to securing the network cannot be pulled out on demand.

Users pay a fixed base fee for every signature a transaction carries. Half of that amount is destroyed and half is paid to the validator that produced the block. A transaction may attach an optional priority fee, quoted per unit of requested compute, which under a protocol change adopted in 2025 goes in full to the block producer; this is the mechanism that rations capacity when demand exceeds what a slot can hold. Program execution is metered in compute units against a per-transaction ceiling, so the cost of a contract call tracks the work it requests rather than a flat tariff.

Storage is charged once, not continuously. An account has to hold a minimum balance scaled to the number of bytes it occupies in order to be exempt from rent, and that balance is a refundable deposit rather than a fee: closing the account returns it. Recurring rent collection has been switched off at the protocol level and rent-paying accounts can no longer be created, so ongoing storage charges do not form part of the fee model as it now stands.

Staked assets are not confiscated by the protocol. No implemented mechanism automatically destroys a validator's stake for equivocation or for being offline; the cost of downtime is forgone reward set against operating expense, including the fees an operator pays to submit its own votes. A scheme to record provable duplicate-block violations on chain, as groundwork for any future economic penalty, is still at proposal stage and would not itself remove stake.

Users pay a single fee that covers three distinct costs, and understanding the split explains why this network's economics differ from those of a standalone chain. The first is execution: contracts are metered in computational steps and in calls to specialized built-in operations, and the submitter pays for the resources their transaction consumes. The second is the cost of publishing data to the settlement layer so that anyone can reconstruct the rollup's state, which is passed through at whatever the settlement layer charges for the dedicated data space it makes available, and which therefore moves with conditions on a network this one does not control. The third is the cost of producing and verifying the validity proof.

That third component behaves in a way with no equivalent on a monolithic chain. A proof covers a whole batch, and the cost of generating it and having it verified is amortized across every transaction inside that batch, so the per-transaction share falls as the batch fills. Periods of heavy use are therefore cheaper per transaction than quiet ones, which inverts the usual relationship between congestion and cost for this part of the fee. Fees may be paid in the network's native asset or in ether.

On the incentive side, validators lock the native asset and delegators may assign their holdings to a validator without running infrastructure, sharing in rewards after the validator's commission. Rewards are funded by protocol issuance. Eligibility is conditioned on performance: attestation makes it observable whether a validator is doing the work, and a validator that fails to attest does not earn for that period. The published design penalizes absence by withholding rewards, and this text does not assert any confiscatory penalty on locked principal, as the parameters governing that are still being settled alongside the remaining decentralization stages. Sequencer and prover operation is currently funded by the operating organization rather than by an open market in those roles.

Starknet is present on the following networks: Ethereum, Solana, Starknet.

The figure reported for this network is assembled machine by machine, treating the computers that run the protocol as the thing that draws electricity. The starting point is an estimate of how many independent nodes are operating, built from crawlers that walk the peer-to-peer layer and record every peer they can reach, supplemented by public listings of infrastructure and staking providers and by the protocol's own visible record of how much stake is active and how it is spread across operators.

A representative hardware profile is then inferred for those machines. The client software publishes what it requires in processor, memory and disk terms, and operators have little reason to provision far beyond that, so the profile is derived from those stated requirements rather than from a survey of individual operators. Power draw for the resulting device classes comes from measurement on representative equipment under controlled laboratory conditions, capturing both the load validating places on a machine and the draw of a machine that is powered on but momentarily idle, which for a network of this kind accounts for a large share of the total. Multiplying measured per-device draw across the estimated population over the reporting period yields the network figure. Where a disclosure concerns one of the many assets issued on this network rather than the network itself, a portion of the network total is assigned to it in proportion to observed on-chain transfer volumes.

The limits deserve stating plainly. The node count records what is reachable, not a census, and machines behind restrictive network configurations are missed. The hardware profile is a reasoned inference from published software requirements, not a record of what any particular operator bought. Nothing here is metered at the wall. Where the evidence runs out, the assumptions chosen are those that push the estimate upward rather than downward, so the result is more likely to overstate consumption than to understate it, and it is revised as observation improves. The network's own account of its energy profile is published at Ethereum energy consumption.

The figure reported for this network is assembled from the machines that run it rather than inferred from any single aggregate quantity. The starting point is a count of active nodes, put together from network crawlers, publicly reachable cluster and gossip information, and data operators choose to publish. That population is then divided by role, because a validator taking part in voting, a machine that only replays the ledger, and the infrastructure that answers application requests do not draw comparable amounts of power.

Each role is matched to a representative hardware profile derived from the resources the client software is documented to need. Requirements here are heavy by the standards of proof-of-stake systems, running to many processor cores, large memory and fast solid-state storage, and the profiles reflect that rather than assuming commodity equipment. Electrical draw per profile is taken from controlled bench measurement of equivalent devices, capturing both the load imposed by processing and the draw of a machine that is powered up but idle, since a node consumes electricity continuously whether or not it is producing a block. Multiplying profiles by the estimated population across the hours of the reporting period gives consumption for the network as a whole. Where a figure is attributed to one asset issued on the network rather than to the network itself, the share is taken from observed on-chain transfer activity for that asset.

The output is an estimate and should be read as one. The node count rests on what is visible from outside, and operators are under no obligation to be visible; the hardware mix is inferred from stated requirements rather than surveyed; and facility overheads such as cooling and power conversion are approximated rather than metered. Where the evidence does not settle a question, the assumption adopted is the one more likely to overstate consumption than understate it, and figures are restated as observation improves or as protocol changes alter the work a node must perform. The network's own climate reporting is published at Solana Climate Dashboard.

A rollup's energy accounting has two parts, and reporting only one of them would misstate the result. The first part is the infrastructure this network runs itself: the sequencer nodes that order and execute transactions, the full nodes that follow and serve the chain, and the proving infrastructure. Proving deserves separate treatment because it is unlike anything in a conventional validator set. Generating a succinct proof of a batch is a heavy, sustained computation run on specialized hardware in a small number of facilities, and it recurs for every batch, so it is a continuous load rather than an occasional one. Because that work is concentrated in few locations operated by one organization rather than spread across an anonymous population, the count of machines involved is a far smaller and better-characterized number than a node crawl would produce, though the specification of that hardware is not publicly detailed and has to be approximated from the class of equipment such workloads require.

The second part is the share of the settlement layer attributable to this network. Ethereum's own consumption is estimated from its validator population, and a portion is assigned here in proportion to the settlement resources this network consumes — the data space it occupies and the verification it triggers — measured against the total those resources represent. The settlement layer publishes its own account of its energy profile at ethereum.org.

Where data availability sits matters to the boundary and is stated rather than assumed: this network publishes its data to the settlement layer, so the storage and bandwidth burden of keeping that data retrievable falls inside the settlement share rather than on a separate network. The usual caveats apply and are sharper here. The operator-run portion is not independently observable, so its estimate rests on the class of hardware such work requires rather than on a measured inventory. Where evidence is thin, conservative assumptions are used that are more likely to overstate than understate, and figures are revised as the network's operation opens up and more of it becomes externally measurable.

Starknet is present on the following networks: Ethereum, Solana, Starknet.

The renewable share reported here is a weighted average of grid mixes rather than a record of what any operator actually buys. It is produced in two steps: establish where the infrastructure sits, then attach regional electricity statistics to those places.

Location is inferred from what the network exposes publicly. Nodes advertise network addresses in order to be reachable by peers, and those addresses resolve to a country accurately enough to describe an aggregate distribution, even though any single resolution may be wrong. Crawlers of the peer-to-peer layer and public directories of hosting and staking infrastructure supply the input. Where the observable sample is too thin or too skewed to stand for the whole population, the geographic spread of a structurally similar network is substituted, chosen because its participants face comparable hardware costs and comparable pressures over where to site machines, on the reasoning that operators respond to the same commercial forces even where the software differs.

Each location is then matched to published statistics on how electricity in that country or region is generated. The renewable proportion for the network is the consumption-weighted share falling in regions where generation is renewable. Grid averages are used because the alternative, knowing each operator's actual supply contract, is not observable; an operator on a dedicated renewable supply and one drawing ordinary grid power in the same country are treated alike.

Energy intensity is reported on a different basis from total consumption. It is a marginal quantity: the additional electricity attributable to processing one further transaction on the network as it currently runs. For a network whose consumption is driven by a validator set that operates continuously regardless of how busy the chain is, that marginal figure is small, and it is not the total divided by the transaction count. The generation statistics are drawn from Share of electricity generated by renewables, compiled by Our World in Data from Ember's electricity datasets and the Energy Institute's Statistical Review of World Energy.

The renewable share is derived geographically. Node locations are inferred from what the network exposes about itself: addresses observable through crawlers and public cluster information, resolved to a country or region. Coverage is never complete, because operators may sit behind hosting providers or relays that obscure where the hardware physically sits. Where the geographic spread of this network cannot be observed directly, the distribution of a structurally similar network stands in as a proxy, chosen because its validator economics and agreement protocol place comparable demands on operators and therefore tend to attract them to comparable locations.

Each located node is then assigned the generation mix of the grid that serves it. Those regional mixes come from Share of electricity generated by renewables, compiled by Our World in Data from Ember's yearly electricity data and the Energy Institute's Statistical Review of World Energy. Weighting each region's renewable share by the estimated consumption sitting in that region produces a network-wide proportion. The result describes the grids the infrastructure draws from, not contractual purchases: an operator buying renewable certificates is not treated differently from a neighbor on the same grid, because that distinction cannot be observed from outside.

Energy intensity is reported separately and means something narrower than total consumption divided by transaction count. It is the marginal quantity of energy associated with processing one further transaction. That distinction matters for a network of this type, where validators run continuously at close to constant power regardless of how full the blocks are, so the incremental energy attached to an additional transaction is small while the standing consumption of the validator set is not. Both the renewable share and the intensity figure therefore move with two separate things: the composition and location of the node population, and the grid statistics for the years covered, which are themselves restated as national reporting is revised.

Two populations have to be located separately, and they are known with very different confidence. The proving and sequencing infrastructure is operated by a single organization from a small number of facilities, so rather than inferring a geographic distribution from network observation, the estimate rests on the far smaller question of which jurisdictions that organization operates in. That is a narrower uncertainty than a distributed validator set presents, and it means a single siting decision moves this network's renewable share in a way that no individual operator could move a chain with thousands of independent nodes.

The settlement layer share carries the geographic distribution of the settlement layer's own validator population, which is estimated from network observation and is far more dispersed. The renewable proportion reported here is therefore a blend: the operator-run infrastructure weighted by its own consumption and located where that organization runs it, combined with a share of the settlement layer weighted by the settlement resources this network consumes and located according to that network's validator distribution. Each location is matched to published statistics for its regional grid.

The limitations are specific. Hosting arrangements can place equipment in a jurisdiction other than the operator's own, and commercial facilities do not generally publish their supply mix, so the grid average is used in place of the actual supply of a particular building. Grid statistics are annual, which smooths over seasonal and daily variation. Contractual renewable purchases are not counted, since this describes the physical grid mix rather than a procurement position.

Energy intensity per transaction is period consumption divided by transactions settled, and for this network the quotient falls meaningfully as usage rises, more so than on a monolithic chain. Proving and settlement costs are largely per-batch rather than per-transaction, so filling batches spreads a near-fixed cost across more transactions. The figure describes an average across the period rather than the energy one additional transaction causes. Source data is processed by Our World in Data from Ember and the Energy Institute's Statistical Review of World Energy: Share of electricity generated by renewables.

Starknet is present on the following networks: Ethereum, Solana, Starknet.

Emissions are derived from the consumption estimate rather than measured, by attaching a carbon intensity to each unit of electricity the network is estimated to draw and summing across the network.

The geographic step repeats the one used for the renewable share. Node locations are inferred from publicly observable network data, principally the addresses peers advertise so that others can connect to them, gathered by crawlers and supplemented by public information about where staking and hosting infrastructure is operated. Where that observation is too sparse to characterize the whole population, the distribution of a comparable network stands in for it, selected because its participants face similar operating economics rather than because its software resembles this one. Each region is assigned a carbon intensity, meaning the average greenhouse gas released per unit of electricity generated on that grid, expressed in carbon dioxide equivalent so that methane and the other gases are counted on a common basis. Estimated consumption in a region multiplied by that region's intensity, summed across regions, gives the network total.

The reporting separates two scopes. Scope 1 covers emissions from sources the operators of the infrastructure control directly, such as fuel burned on site in a generator. For a network of this kind, whose participants overwhelmingly run ordinary servers connected to a public grid, there is generally nothing in that category, and it is reported as such rather than left out. Scope 2 covers the indirect emissions embodied in the electricity purchased to run that infrastructure, and that is where essentially the whole footprint sits. Emissions further up the supply chain, such as those from manufacturing and shipping the hardware, fall outside this boundary.

Greenhouse gas intensity follows the same marginal logic as energy intensity: it expresses the additional emissions attributable to one further transaction rather than an average spread across all of them. Because it inherits both the consumption estimate and the grid averages, its uncertainty combines theirs. Carbon intensities are taken from Carbon intensity of electricity generation, compiled by Our World in Data from Ember's electricity datasets and the Energy Institute's Statistical Review of World Energy, and made available under the CC BY 4.0 license.

Emissions are derived from the same geographic picture used for energy sources, applied to a different set of grid statistics. Node locations are inferred from addresses observable through crawlers and public cluster information and resolved to a region; where direct observation falls short, the geographic distribution of a structurally comparable network is substituted, selected on the basis that its incentive design and agreement protocol impose similar operating demands.

Each region is then paired with a carbon intensity for its electricity, taken from Carbon intensity of electricity generation, compiled by Our World in Data from Ember's yearly electricity data and the Energy Institute's Statistical Review of World Energy and made available under the CC BY 4.0 licence. Multiplying the electricity estimated to be consumed in a region by that region's carbon intensity, and summing across regions, gives the emissions attributable to running the network.

The disclosure separates two scopes. Scope 1 covers emissions from sources the operators of the infrastructure control directly, such as fuel burned on site; for a network of this kind, whose nodes are ordinary servers in rented facilities, this is normally nil or immaterial, and a zero figure reflects the absence of such sources rather than an omission. Scope 2 covers the indirect emissions embodied in the electricity those machines purchase, and is where essentially the whole footprint of this network falls.

Greenhouse gas intensity follows the same marginal logic as its energy counterpart: it expresses the emissions associated with one additional transaction rather than an average obtained by dividing an annual total by throughput. Because validators consume electricity at a fairly steady rate whether or not blocks are full, the marginal figure is small and is not a proxy for the footprint of the network as a whole. Both the absolute emissions and the intensity figure are sensitive to the grid statistics underlying them, which are revised as national energy reporting is updated.

Emissions follow from the consumption estimate by applying a carbon intensity to the electricity drawn at each location, and the two-part structure of the consumption estimate carries straight through. The operator-run sequencing and proving infrastructure is assigned the carbon intensity of the grids serving the facilities it runs in. The attributed share of the settlement layer is assigned the intensity implied by that layer's own validator distribution. Summing the two gives the total, and the boundary is operational electricity: neither the manufacture of proving hardware nor the construction of the facilities housing it is included.

Scope 1 covers emissions from sources the operators directly control, which for infrastructure hosted in commercial facilities means occasional backup generation and little else. It is a negligible contributor here and is reported as such rather than modeled in detail. Scope 2 covers emissions embodied in purchased electricity and accounts for effectively the entire footprint, on both the operator-run side and the attributed settlement share.

Greenhouse gas intensity per transaction is period emissions divided by transactions settled in the period. It inherits the batching property described for energy intensity, falling as batches fill, and it also inherits the uncertainty of every step behind it: an error in the assumed proving hardware propagates into consumption and from there into emissions, while the settlement share depends on both that layer's own estimate and the attribution rule used to divide it. Grid carbon intensities are annual averages that conceal substantial variation across a day and a year, and a network whose infrastructure sits in few locations is more exposed to that variation than one spread across many grids, because there is less averaging to smooth it out. Figures are restated each period as the network's operation becomes more openly observable and as grid data is updated. Carbon intensity data is processed by Our World in Data from Ember and the Energy Institute's Statistical Review of World Energy, and is made available under a Creative Commons BY 4.0 license: Carbon intensity of electricity generation.