Kaspa (KAS) sustainability report

Kaspa is present on the following networks: Kaspa.

Kaspa is secured by proof of work, but it does not organize blocks into a single chain. Blocks form a directed acyclic graph in which a new block references every tip it can see, and the ordering protocol takes that graph and produces one agreed sequence from it. The rule identifies a well-connected cluster of blocks whose members all reference one another, treats that cluster as the honest backbone, and orders the remaining blocks around it. Blocks that in a conventional design would be discarded as orphans because a competitor was found at the same moment are instead absorbed into the ordering and their transactions retained. A parameter bounds how much simultaneity the protocol tolerates, and it is set from the block rate and expected propagation delay.

That structure is what allows an unusually high block rate. A 2025 upgrade raised production to ten blocks per second, having first required a complete rewrite of the node software into a faster implementation; the consensus parameters governing tolerated concurrency, the depth at which ordering is treated as settled, and the horizon beyond which old data is pruned were recalibrated together to preserve security at the higher rate. With roughly a tenth of a second between blocks, many blocks are in flight across the network at once and parallel creation is the normal case rather than an anomaly.

Mining uses a memory-oriented hash algorithm and is performed by purpose-built hardware. There is no staking, no bonded capital, no delegation and no mechanism for confiscating a participant's funds; a miner who finds blocks is paid and a miner who does not is not, which is the entirety of the incentive structure. Settlement strengthens with depth as in other proof-of-work systems, with the significant difference that work spent on blocks that lose a race is not wasted, because those blocks still enter the ordering. Later upgrades have continued to adjust the protocol, and further increases in block rate are described as requiring additional changes to how miners reference the graph.

Kaspa is present on the following networks: Kaspa.

Miners are paid a block subsidy plus the fees attached to the transactions they include. Issuance is expressed as a quantity per second rather than per block, so raising the block rate divided the same emission into more frequent, smaller payments without altering the total released over time. The subsidy declines on a smooth schedule of frequent small reductions rather than in occasional large steps, which avoids the abrupt revenue cliffs that a halving imposes on mining operations elsewhere and makes the transition gentler for participants whose margins are thin.

Fees are a bid for inclusion. A user attaches a fee and miners prefer transactions offering more per unit of the space they occupy, so the price of prompt inclusion rises with demand and falls when the network is quiet. The high block rate changes the practical experience of this market: capacity arrives ten times a second, so the queue clears quickly and the fee required for timely inclusion stays low outside sustained congestion.

The absorption of parallel blocks into the ordering has a direct economic consequence worth stating plainly. In a single-chain design, a miner who finds a block at nearly the same moment as someone else loses everything spent on it, and that risk grows as block times shorten, which is why conventional chains cannot safely raise their block rate very far. Here a block found simultaneously with another is still ordered and its finder still compensated, so the penalty for losing a race largely disappears. That removes the structural advantage that large, well-connected mining operations enjoy elsewhere, where superior propagation translates directly into fewer wasted blocks.

There is no staking, so there are no staking rewards, no commission, no unbonding period and no penalties applied to bonded capital. There are no storage rents or recurring state charges levied on users; the cost of a transaction is the fee paid at submission.

Kaspa is present on the following networks: Kaspa.

A proof-of-work network is estimated from the top down, because its consumption is governed by mining economics rather than by a count of participants. Nobody can enumerate the machines securing the network, but the aggregate difficulty is public and reveals how much computational work is being performed. The question the model answers is what hardware is plausibly performing that work and what it draws.

The hash algorithm determines the answer to the first part. It is memory-oriented and has purpose-built hardware designed for it, so the machines in service are drawn from a known and fairly short list of commercially available devices, each with published throughput and power figures. A profitability threshold then narrows that list: mining revenue over the period is compared against the operating cost of each device at representative electricity prices, and only hardware that could plausibly cover its running costs is assumed to be in service. Older, less efficient machines drop out of the estimate as revenue falls and return when it rises, so the modeled fleet shifts with market conditions rather than being fixed. The remaining devices are combined into a weighted mix that accounts for the observed aggregate work, and their power draw is summed, with an allowance for the cooling and power-delivery overhead of the facilities housing them.

This chain has no merge-mining relationship with another network, so its work is not shared with or double-counted against any other chain, and the whole of the modeled consumption is attributable here.

The limitations are inherent to the approach. The hardware mix is inferred rather than observed, and the profitability threshold depends on an assumed electricity price that varies enormously between operations. Facility overhead is a broad assumption applied uniformly. Where the evidence is thin the conservative choice is taken, which is more likely to overstate than understate, and figures are revised as the hardware population and market conditions change. Raising the block rate does not by itself change consumption, since the same aggregate work is simply distributed across more blocks.

Kaspa is present on the following networks: Kaspa.

Locating proof-of-work hardware is harder than locating the nodes of a stake-based network, because a miner has no reason to identify itself and every reason not to. The distribution used here is inferred from what can be observed at the aggregation points: mining pools, through which most participants direct their work, expose information about where the work reaching them originates, and the network topology of the nodes that relay blocks provides a second, partial view. Those observations produce a distribution across countries and regions that is treated as a sample of the whole rather than a census.

Each region is then matched to published statistics on how its electricity is generated, and the renewable proportion is the average of those figures weighted by the consumption attributed to each region. Mining is unusually mobile compared with other computing loads, because the equipment is fungible and the economics are dominated by the price of power, so the fleet migrates toward cheap electricity and the geographic distribution can shift materially within a single reporting period. It also concentrates where power is cheap for reasons that correlate with generation type, in both directions: hydroelectric capacity and stranded gas both produce low prices and pull very different renewable shares. The distribution should therefore be read as a snapshot, not a stable characteristic.

Grid statistics are annual and regional, so seasonal variation is invisible. That matters more here than for most workloads, because some mining capacity is deliberately sited to follow seasonal hydroelectric surpluses and relocates when they recede.

Energy intensity per transaction divides period consumption by the transactions settled. For a proof-of-work chain this quotient requires an explicit caveat: consumption is determined by mining revenue and difficulty, not by how many transactions are processed, so an additional transaction causes essentially no additional energy. The figure is an accounting average, and it falls as the network is used more without any change in the underlying energy use. 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.

Kaspa is present on the following networks: Kaspa.

Emissions are calculated by applying a carbon intensity to the electricity attributed to each mining region, using the same regional distribution established for the renewable share and the same consumption estimate derived from difficulty and hardware economics. The products are summed across regions. The boundary is operational electricity only; the manufacture of mining hardware is not included, which is a material omission for proof of work specifically, because purpose-built machines have short economic lives and are replaced frequently, so their embodied emissions are proportionally larger than they would be for general-purpose servers. That is a limitation of the boundary rather than an assertion that those emissions are small.

Scope 1 covers emissions from sources under the direct control of the operators. For most mining this means little, as facilities draw from the grid, but it is not uniformly negligible in this sector: operations sited at gas production facilities to consume gas that would otherwise be flared involve combustion under the operator's own control, and where such capacity forms part of the modeled fleet it belongs in scope 1 rather than scope 2. The estimate treats it as a minor component and does not attempt to model it in detail, which is a stated limitation.

Scope 2 covers emissions embodied in purchased electricity and accounts for the large majority of the total.

Greenhouse gas intensity per transaction is period emissions divided by transactions settled, and carries the same caveat as energy intensity: the numerator is driven by mining economics rather than by usage, so the quotient is an accounting average and not the emissions caused by one more transfer. Uncertainty accumulates through every stage, since errors in the hardware mix propagate into consumption and then into emissions, and the geographic distribution on which intensity depends is the least certain input of all. Figures are restated each period. 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.