Sui (SUI) sustainability report

Sui is present on the following networks: Sui.

Sui runs a delegated proof-of-stake network whose validator committee is fixed for an epoch of roughly one day, with voting power proportional to the stake bonded to each member. Agreement uses a directed acyclic graph rather than a single chain of proposals: validators produce blocks each round that reference blocks from the previous round, and the resulting structure is read directly by every validator to work out which blocks are committed and in what order. Because the commit rule is evaluated over the graph itself, no separate round of explicit certification is needed, which removes network round trips from the critical path and brings commit latency down to a few hundred milliseconds. A more recent revision folded transaction validation into the same process and routes each submitted transaction through a single coordinating validator, cutting duplicated signature work.

The distinctive part of the design is that not every transaction has to pass through that machinery. State is modeled as discrete objects, each either owned by a single address, shared, or immutable, and a transaction declares the objects it will read and write before it executes. A transaction touching only objects owned by its sender cannot conflict with anything another party might submit, so it does not need a global ordering: a quorum of stake signing it by reliable broadcast is enough to settle it, and it finalizes on this fast path in roughly the time of two network round trips. Transactions that touch shared objects, which is what most decentralized finance activity involves, do require the graph-based protocol to sequence competing accesses, and carry slightly higher latency and cost as a result, rising further when many transactions contend for the same popular object.

Execution takes advantage of the same declarations: transactions whose object sets do not overlap run concurrently. The protocol remains safe provided faulty or malicious validators hold less than a third of voting power.

Sui is present on the following networks: Sui.

Stake is bonded by wrapping the native asset in a stake object delegated to a validator's pool; the holder keeps custody of that object and the pool's exchange rate appreciates as rewards accrue, so redeeming it later returns principal plus accumulated reward net of the validator's commission. Rewards are settled at epoch boundaries and come from the computation fees collected during the epoch together with issuance intended to support the validator set in the network's early years. Stake counts toward rewards only for epochs it was active throughout.

Penalties bite on reward, not on principal. Validators grade each other over the epoch, and if holders of more than two thirds of voting power report a given validator for poor operation, that validator's rewards for the epoch are reduced or removed entirely, and the holders who delegated to it forfeit their reward for that epoch as well. The bonded principal itself is not destroyed. A validator whose stake falls below the threshold for committee membership simply leaves the committee. The same reward multiplier enforces the fee market: at the start of each epoch validators submit the lowest price at which they will process transactions, the reference price for the epoch is the stake-weighted two-thirds percentile of those quotes, and validators that quote a low price and then honor it are rewarded relative to those that do not.

Users pay two components. Computation is metered in units placed into coarse buckets, so similar transactions cost the same and developers are not pushed into micro-optimization, and the bucket is priced at the epoch's reference rate. Storage is charged per byte at a price fixed by governance rather than set by congestion, and the amount paid is routed into a storage fund instead of to the validators of the day. Deleting data returns almost all of the storage charge as a rebate, so a transaction that frees more state than it occupies can settle at a net credit. The storage fund is the unusual element: it is counted alongside user stake when rewards are computed, most of the return it earns is paid to whichever validators are currently storing historical data, the remainder is reinvested, and its principal is never paid out, so it cannot be drained by the validators it compensates.

Sui is present on the following networks: Sui.

The estimate starts from the machines that operate the network and works upward. Its first input is the population of nodes, separated into the validators that form the committee, which is recorded on chain and therefore countable rather than guessed, and the far larger and less visible set of full nodes that replicate state, index it and serve application traffic. The size of that second group is inferred from network crawlers and publicly reachable peer information, and is the main source of uncertainty in the count.

Each group is matched to a hardware profile derived from the resources the node software is documented to need. Validators here are expected to execute transactions in parallel across cores and to maintain the object store and its indexes, so their profiles assume multi-core server hardware with generous memory and fast solid-state storage rather than modest equipment. Power draw for each profile is taken from bench measurement of comparable devices under load and at rest, and idle draw is counted, because a node is expected to remain available continuously whether or not transactions are arriving. Multiplying draw by the estimated population across the hours of the period gives consumption for the network. Where a share of that total is attributed to an individual asset issued on the network, that share is based on the asset's observed proportion of on-chain transfer activity.

The limits of the method should be read alongside the result. Node counts outside the committee rest on what is observable from the public network, and operators need not be observable; hardware is inferred from documented requirements rather than collected from operators; and the overhead of cooling and power conversion in hosting facilities is approximated from typical factors rather than metered. Where the evidence does not resolve a question, the assumption chosen is the one that tends to raise rather than lower the reported figure, and estimates are revised as observation improves and as protocol changes alter what a node must do.

Sui is present on the following networks: Sui.

Working out the renewable share means first working out where the machines sit. Addresses visible through network crawlers and public peer information are resolved to a country or region, which gives incomplete coverage: many nodes sit behind hosting providers or relays, and the announced location of an address need not match the facility housing the hardware. Where the geographic spread cannot be observed directly, the distribution of a structurally similar network is substituted, chosen because its staking economics and agreement protocol impose comparable operating demands and so tend to concentrate operators in comparable hosting markets.

Located nodes are then assigned the generation mix of the grid that supplies them, using 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 in that region produces a share for the network as a whole. That share describes the physical grids the infrastructure draws from, not contractual sourcing: an operator holding renewable supply agreements is treated the same as any other operator on the same grid, because such arrangements are not visible from outside the network.

Energy intensity is reported as a separate quantity and means the additional energy associated with handling one further transaction, not the annual total divided by the number of transactions. The distinction is material for this network, where the committee runs continuously at a broadly constant power level and much of the traffic settles over a path that adds little incremental work, so the marginal figure is small while the standing consumption of the node population is not. Both the renewable share and the intensity respond to two separate inputs: the size, mix and placement of the node population, and the grid statistics for the years covered, which are themselves restated as national energy reporting is revised.

Sui is present on the following networks: Sui.

Emissions are derived by taking the geographic picture assembled for energy sourcing and applying carbon statistics to it instead of generation-mix statistics. Node addresses observable through crawlers and public peer information are resolved to regions, and where that resolution is incomplete the distribution of a structurally comparable network stands in, chosen because its incentive structure and agreement protocol create similar operating requirements and therefore a similar hosting pattern.

Every region is then paired with the carbon intensity of the electricity generated there, 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 published under the CC BY 4.0 licence. The electricity estimated to be consumed in each region is multiplied by that region's intensity and the products are summed to give the emissions attributable to operating the network over the reporting period.

The disclosure keeps two scopes apart. Scope 1 covers emissions from sources directly controlled by those running the infrastructure, for instance fuel burned on site. For a network whose nodes are servers in third-party data centers this is normally nil or immaterial, and a zero figure records the absence of such sources rather than missing data. Scope 2 covers the indirect emissions carried in the electricity those machines purchase, and accounts for substantially the whole footprint reported for this network.

Greenhouse gas intensity is marginal in the same sense as its energy counterpart: it is the emission associated with one additional transaction rather than an average produced by dividing an annual total by throughput. Because the committee draws power at a fairly steady rate regardless of how busy the network is, that marginal figure stays small and should not be treated as each transaction's share of the network's overall footprint. Both the absolute emissions and the intensity shift with the grid data underneath them, which is revised as national energy statistics are updated, and with any protocol change that alters the hardware a node requires.