Aptos (APT) sustainability report

Aptos is present on the following networks: Aptos Coin.

Aptos combines proof of stake with a Byzantine fault tolerant agreement protocol in the lineage of pipelined, leader-based designs. Validators hold voting power in proportion to the stake bonded to them, and a committee is fixed for the duration of an epoch, which lasts two hours; changes to the set take effect only at those boundaries. Within an epoch, a leader is chosen for each round to propose the next block, and the choice is weighted not only by stake but by recent performance, so an operator that repeatedly fails to get its proposals committed is proposed less often. Agreement is deterministic: once a quorum of voting power certifies a block, it is final and cannot be reorganized, in contrast to systems where confidence accumulates probabilistically.

Transaction data is disseminated separately from ordering. Validators batch incoming transactions and circulate them ahead of time, collecting proofs of availability, so a proposal need only reference batches that the rest of the committee already holds rather than carry the payload itself. This separation is what allows proposals to stay small and the agreement path to stay short. Successive protocol revisions have compressed that path further, cutting the number of network round trips needed to commit and allowing a leader to propose once per network delay instead of waiting out two, which has brought block intervals down to a few tens of milliseconds and finality comfortably below a second.

Execution is parallel and optimistic. Rather than requiring developers to declare in advance which state a transaction will touch, the engine speculatively executes transactions concurrently across processor cores, detects at runtime where one transaction read state another wrote, and re-executes only the affected transactions. Because the serial order fixed by agreement is used as the reference, the outcome is identical to sequential execution and every validator derives the same result. The network tolerates faulty or malicious behavior by up to one third of voting power.

Aptos is present on the following networks: Aptos Coin.

Validators are paid from newly issued units of the native asset at a rate set by on-chain governance, and the payment is conditioned on output rather than mere presence. The reward for an epoch scales with the stake bonded to a validator multiplied by the proportion of its block proposals that were actually committed, so an operator that is offline, slow, or frequently skipped earns proportionally less. Rewards are added to the staked balance at each epoch boundary and compound from there. Holders who do not run hardware delegate into a validator's staking pool and receive their share after the operator's commission; stake locks and unlocks on epoch boundaries rather than on demand.

Penalties are exclusionary rather than confiscatory. The protocol does not currently implement slashing: there is no mechanism that destroys a validator's bonded stake for double-signing or for prolonged inactivity. What an underperforming operator loses is reward, through the proposal-success term, and eventually its place, since a validator whose stake falls below the minimum required to remain in the active set is moved to inactive status at an epoch transition, and governance can remove a validator directly. Delegators exert the remaining pressure by moving stake elsewhere.

Users pay for transactions in gas units priced in a small subunit of the native asset. The charge separates two things. Execution and input-output gas covers computation and propagation, and its unit price floats with congestion, which is also how transactions are prioritized: validators select from the pending pool by offered gas unit price, banded into discrete tiers rather than compared continuously, so raising a bid within a tier changes nothing. Storage is billed separately and at a price fixed in the native asset rather than in gas units, so the cost of occupying state does not swing with congestion, and the charge is refunded when the storage is released, meaning a transaction that frees more state than it creates can settle as a net credit. Under the fee parameters governance currently sets, collected gas is burned in full rather than routed to the proposing validator.

Aptos is present on the following networks: Aptos Coin.

Consumption is built up from the population of machines that operate the network. The first input is a count of those machines, split into the validators that take part in agreement, the fullnodes that operators run beside them to distribute state, and the public fullnodes that serve reads to applications. One part of this population is unusually well observed: the active validator set is recorded on chain and reconstituted at every epoch transition, so its size is known rather than estimated. The wider fullnode population is not, and is inferred from network crawlers and publicly reachable peer information.

Each class of machine is matched to a hardware profile taken from the resources the node software is documented to require. Because the execution engine is built to occupy many processor cores at once, validator profiles assume multi-core server equipment with substantial memory and fast persistent storage rather than a modest configuration. Power draw for a profile comes from bench measurement of equivalent hardware, recorded both under load and at rest, with idle draw included because a node is expected to stay available continuously. Aggregating draw across the estimated population over the hours in the reporting period yields consumption for the network. Where a portion of that total is attributed to an individual asset issued on the network, the share is based on that asset's observed proportion of on-chain transfer activity.

Several qualifications apply. The fullnode population is inferred rather than counted; hardware is deduced from documented requirements rather than reported by the operators themselves; and facility overheads for cooling and power conversion are approximated using typical data-center factors instead of being metered at each site. Where the available evidence leaves a question open, the assumption taken is the one that tends to produce the higher consumption figure, so the disclosure is not made flattering by accident. Estimates are revised as observation improves and after protocol changes that alter the work a node has to do.

Aptos is present on the following networks: Aptos Coin.

Establishing where the network's electricity comes from begins with establishing where its machines are. Node addresses observable through crawlers and public peer information are resolved to a country or region, giving partial geographic coverage; it is only partial, because operators frequently sit behind hosting providers whose announced location need not match the facility actually running the hardware. Where the spread cannot be pinned down directly, the observed distribution of a structurally similar network is used in its place, selected because its staking economics and agreement protocol place comparable demands on operators and so tend to draw them toward comparable hosting markets.

Each located node is then assigned the generation mix of the grid supplying it. Those regional mixes are taken 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 proportion by the consumption estimated to sit there produces a figure for the network overall. The figure describes the grids the infrastructure physically draws on rather than any contractual arrangement: renewable purchase agreements held by individual operators are not reflected, since they cannot be observed from outside the network.

Energy intensity is a separate quantity and is narrower than dividing total consumption by transaction count. It expresses the additional energy associated with processing one further transaction. The distinction carries weight for a network of this design, where validators run continuously at a broadly constant power level and parallel execution absorbs additional transactions using capacity that is already switched on, so the marginal figure is small while the standing consumption of the validator set is not. Both numbers move with two independent inputs: the size, role mix and location of the node population, and the grid statistics for the years covered, which are themselves restated as national energy reporting is revised.

Aptos is present on the following networks: Aptos Coin.

The emissions calculation reuses the geographic picture built for energy sources and applies a different body of grid statistics to it. Node addresses visible through crawlers and public peer information are resolved to regions; where that resolution is incomplete, the distribution of a structurally comparable network is substituted, chosen because its incentive design and agreement protocol impose similar operating requirements and therefore a similar hosting footprint.

Each region is paired with the carbon intensity of its electricity, drawn 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 published under the CC BY 4.0 licence. Multiplying the electricity estimated to be drawn in each region by that region's intensity, then summing across regions, gives the emissions attributable to operating the network over the period.

Two scopes are distinguished. Scope 1 captures emissions from sources under the direct control of those running the infrastructure, such as fuel combusted on site. For a network whose nodes are servers in third-party facilities, this is normally nil or immaterial, and a zero entry records the absence of such sources rather than a gap in the data. Scope 2 captures the indirect emissions embodied in the electricity those machines buy, and accounts for effectively the entire footprint reported here.

Greenhouse gas intensity mirrors its energy equivalent in being marginal rather than average: it is the emission associated with one additional transaction, not an annual total divided by throughput. Because validators consume power at a fairly steady rate irrespective of how full their blocks are, that marginal figure stays small and should not be read as a per-transaction apportionment of the network's whole footprint. Both the absolute emissions and the intensity are sensitive to the underlying grid data, which is updated as national energy statistics are revised, and to any protocol change that alters the hardware a validator needs.