Kusama (KSM) sustainability report

Kusama is present on the following networks: Kusama.

Kusama is a nominated proof-of-stake network that serves as the proving ground for the protocol family it belongs to, receiving changes before its more conservative sibling does. It carries real economic value and real users, so it is a live network rather than a test environment, but its parameters are deliberately faster and its tolerance for disruption higher.

Two separate mechanisms divide the work of agreeing on a chain. Block production runs as a slot lottery in which each validator privately evaluates a verifiable random function against the current randomness and their entitlement, and a winner can prove the right to author without having announced anything in advance. Finality is a distinct process running alongside it: validators vote on chains rather than on individual blocks, and a round concludes by finalizing everything up to an agreed point at once, which means many blocks can be finalized together and finality is explicit rather than probabilistic.

The active validator set is chosen by an election that does not simply rank candidates by stake. Holders nominate several candidates, and an optimization based on published research distributes nominated stake across the elected set so as to even out the backing behind each validator, which prevents a popular candidate from accumulating disproportionate weight and raises the cost of capturing the set. Rather than fixed blockspace allocations, the network sells access to its validation capacity as a tradable commodity that chains buy on demand or in bulk, replacing the earlier model of long auctions with large locked deposits.

Unlike several other proof-of-stake networks, this one genuinely confiscates stake. Signing two conflicting blocks, or backing an invalid state transition as a parachain validator, results in a proportion of the validator's bond and of the stake nominated to it being taken, with severity scaled to how many validators failed at once, so isolated faults are treated far more leniently than correlated ones that suggest coordination. Slashes are queued before application and governance can reverse them, a power that has been exercised where the cause was found to be operational rather than malicious. Unbonding takes seven days.

Kusama is present on the following networks: Kusama.

Staking rewards are funded by issuance targeting a desired proportion of total supply being staked: when less is staked than the target, rewards rise to attract more, and when more is staked they fall. Rewards accrue to validators by era, and the important detail is that they are divided roughly equally among active validators rather than in proportion to the stake behind them. A validator backed by a great deal of stake therefore returns less per unit staked than one backed by little, which gives nominators a continuous incentive to move toward smaller validators and works against concentration. Each validator takes a declared commission before the remainder is shared among its nominators in proportion to their contributions.

Nominators carry real risk, not merely opportunity cost. Because slashing applies to nominated stake as well as to the validator's own bond, choosing a validator is an underwriting decision, and a nominator backing an operator who equivocates loses a proportion of their funds. This is the sharpest difference between this network's staking economics and those of chains where delegation is risk-free.

The treasury receives a share of issuance, a portion of transaction fees, and funds recovered through slashing that are not paid to whoever reported the offense; a reporter's share exists specifically to make it worthwhile to detect and prove misbehavior. Treasury funds are disbursed by on-chain governance.

Users face fees computed from a transaction's weight, an explicit measure of the execution resources it consumes, plus a charge on its length. A multiplier adjusts fees upward as blocks fill and downward when they empty, damping congestion without a bidding war. Storing data on-chain requires a deposit proportional to the space occupied, returned when the data is removed, which prices persistent state as a rental rather than a one-time purchase. Chains obtain execution capacity by purchasing coretime, and their own users may pay in whatever asset that chain chooses.

Kusama is present on the following networks: Kusama.

This network's footprint has to be split across two populations, because the shared-security design means a chain running here does not maintain its own validator set. The relay chain operates one validator set that provides security to every chain occupying a core, and each of those chains additionally runs collators, which assemble candidate blocks and hand them to relay chain validators for checking but do not themselves participate in consensus. An estimate that counted only one of these two populations would misstate the total.

The relay chain validator population is unusually observable. The active set is capped and its size is a published on-chain parameter, so the count is known exactly rather than estimated from crawling, and validators commonly register on-chain identity information because nominators select them on reputation. That turns the usual hardest question — how many machines are running — into a known quantity, and shifts the uncertainty onto what each operator runs and how many spare or backup machines sit behind each registered identity. Beyond the active set, a larger pool of candidates waits to be elected and runs infrastructure in readiness, which must be counted too since those machines consume power whether or not their operator is currently elected.

Collator populations for the chains occupying cores are smaller, less formally registered and estimated from public listings and network observation, with rather less confidence.

Representative hardware follows from the published requirements for running validator software, which state processor, memory, storage and bandwidth expectations that are firmer than those of lighter chains because parachain validation is demanding. Power draw for machines of that description, measured under load and at idle, is applied across both populations, weighted toward idle because validation and gossip occupy a node far less than continuously.

Limitations are stated plainly: backup and redundant machines are largely invisible from outside, hosting arrangements obscure how many physical devices sit behind a registered identity, and where evidence is thin the conservative assumption is used, more likely to overstate than understate. Figures are revised as observation improves.

Kusama is present on the following networks: Kusama.

Locating this network's machines is easier than for most, because validators want to be found. Nominators choose operators on reputation and reliability, so validators commonly publish identity information on-chain and many operate publicly known infrastructure businesses with stated hosting arrangements. Telemetry services to which nodes voluntarily report add a further view of where they sit. Together these give a distribution across countries and regions derived from disclosure rather than inference, which is a materially stronger basis than crawling anonymous peers, and it covers the relay chain validator set well.

Collators for chains occupying cores are less well documented, and their locations are estimated with the weaker methods used elsewhere: network observation, hosting provider address ranges, and where nothing better exists, the assumption that they follow a distribution similar to that of the validator set they serve.

Each located machine is matched to published statistics for the grid supplying its region, and the renewable share reported is the average across those grids, weighted by consumption attributed to each location rather than by machine count. Two limitations bear on this. Voluntary disclosure is self-selecting, so the operators who publish least may differ systematically from those who publish most. And a large share of validators run in commercial hosting facilities whose actual supply arrangements are not public, so a regional grid average stands in for the real supply of a specific building. Contractual renewable purchases are not counted, since this describes the physical grid mix.

Energy intensity per transaction divides period consumption by transactions settled in the period. The caveat is the one common to stake-based networks and is worth being explicit about here because the validator set is capped: the number of machines does not change when usage rises, so consumption is close to fixed with respect to throughput and an additional transaction causes almost no additional energy. Intensity falls as the network is used more, and the figure is an average rather than a marginal cost. 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.

Kusama is present on the following networks: Kusama.

Emissions are derived from the consumption estimate by applying a grid carbon intensity to the electricity drawn at each location. The distribution of validators and collators established for the renewable calculation is reused, each location matched to a published intensity for its grid, and the products summed across both populations. The boundary is operational electricity; manufacture of the servers and the construction of facilities housing them are outside it, and no factor is applied for either.

Scope 1 covers emissions from sources directly controlled by the operators. For a network of ordinary servers in commercial hosting facilities this amounts to occasional backup generation during outages, is negligible in the total, and is reported as such rather than modeled in detail. Scope 2 covers emissions embodied in purchased electricity and represents effectively the whole footprint.

One structural point distinguishes this network's emissions profile from that of a chain with an open-ended validator population. Because the active set is capped by protocol parameter, total consumption and therefore total emissions are bounded in a way that does not depend on how heavily the network is used or on how valuable its rewards become. Growth in usage does not add machines; it adds work to machines already running. Emissions move when the cap changes, when the waiting pool of candidates grows, or when the geographic distribution shifts, rather than with demand.

Greenhouse gas intensity per transaction is period emissions divided by transactions settled, and carries the same caveat as energy intensity: it describes an average across throughput rather than the emissions caused by one further transaction. Uncertainty compounds through the calculation, since an error in the machine population propagates into consumption and then into emissions, and grid intensities are annual averages concealing considerable daily and seasonal variation. Figures are restated each period as disclosure and grid data improve. 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.