Kaspa (KAS)

What Kaspa is

Kaspa is a proof-of-work network that replaces the single chain of blocks with a directed acyclic graph, so blocks produced at the same moment are all kept rather than discarded. Its own site describes the project as a "Proof-of-Work blockDAG for Real-Time Decentralization" offering "bitcoin's proof-of-work without the wait" (source: Kaspa, read September 2026).

The academic basis is published and citable. The ordering rule is GHOSTDAG, introduced in "PHANTOM and GHOSTDAG: A Scalable Generalization of Nakamoto Consensus" by Yonatan Sompolinsky, Shai Wyborski and Aviv Zohar. The paper states that "PHANTOM generalizes Nakamoto's blockchain to a direct acyclic graph of blocks (blockDAG)," introduces a parameter k that "controls the level of tolerance of the protocol to blocks that were created concurrently," and shows that "GHOSTDAG's ordering of blocks is irreversible up to an exponentially negligible factor" (source: PHANTOM and GHOSTDAG paper, IACR ePrint 2018/104).

For where KAS trades and what buying it costs from your country or state, see where to buy Kaspa and the venues at Exchanges.

How it works

The problem GHOSTDAG solves is orphan waste. In Bitcoin, if two miners find a block at nearly the same time, one of them is discarded and the work behind it is thrown away. That is why block intervals are long: a short interval means many simultaneous blocks and many orphans, and orphans weaken security by wasting honest hash rate.

A blockDAG keeps both blocks. Every block references multiple predecessors, so nothing is orphaned, and the protocol's job becomes ordering rather than selecting. GHOSTDAG does this by identifying a well-connected cluster of blocks that honest miners would have seen, ordering those first and the rest afterwards, with the parameter k setting how much concurrency the protocol tolerates. Because no honest work is discarded, the block rate can rise without the security penalty that would hit a linear chain.

Kaspa has exercised that headroom. The rusty-kaspa client's own documentation records that the Crescendo hard fork on May 5, 2025 "has transitioned the network from a blockrate of 1 BPS to that of 10 BPS" (source: rusty-kaspa repository README, read September 2026). Ten blocks per second is roughly six thousand times Bitcoin's rate, and the confirmation experience follows accordingly.

Everything else is deliberately conservative. Kaspa is proof of work with a utxo model, no smart contract platform in the Ethereum sense, and no premine. Its site describes verification of the genesis UTXO commitment against an empty set, with Bitcoin block hashes embedded as timestamps to prove no hidden premining occurred, and publishes a notebook so anyone can check that claim themselves (source: Kaspa, September 2026).

That verifiable fair launch is unusual and worth noting. Most projects assert a fair launch. Kaspa published a reproducible proof that the genesis block committed to an empty UTXO set at a time provably after a known Bitcoin block.

Supply and tokenomics

Kaspa's emission is defined in the client source rather than in a marketing page, which is the more reliable place to read it. The consensus code defines a pre-deflationary phase base subsidy of 50,000,000,000 sompi per block, and a deflationary phase initial subsidy of 44,000,000,000 sompi per second at one block per second. Sompi is Kaspa's smallest unit, and the subsidy constants are denominated in it (source: rusty-kaspa consensus source, coinbase module, read September 2026).

The halving is monthly-smoothed rather than a cliff. The source defines SECONDS_PER_MONTH as 2,629,800, commented as "365.25 / 12 = 30.4375 days", and SECONDS_PER_HALVING as twelve of those months, with a SUBSIDY_BY_MONTH_TABLE of 426 entries giving "the reward per second for each month" until the subsidy reaches zero after 36 halvings (source: rusty-kaspa consensus source, September 2026).

History

Kaspa's history is a research result becoming a network. The PHANTOM and GHOSTDAG paper was published as IACR ePrint 2018/104, and the protocol it describes is what the client implements: not a variant of Nakamoto consensus but, in the authors' framing, a generalization of it, with the linear chain as the special case where concurrency is not tolerated.

The launch followed the paper's ethos. No premine, no allocation, and a genesis commitment constructed so the absence of hidden coins could be checked rather than trusted, with Bitcoin block hashes embedded to prove the timing.

Crescendo, on May 5, 2025, is the milestone in the network's operational life. Raising the block rate tenfold, from one block per second to ten, is a change that only a DAG design can absorb, and it is the clearest demonstration that the orphan-free property is doing real work rather than being a theoretical nicety. A hard fork of that magnitude also requires near-universal client upgrade, which is a governance fact as much as a technical one.

Risks and what to watch

Mining concentration is the standing risk of any proof-of-work network. A high block rate does not change the fact that security is bought with hash rate, and where that hash rate is concentrated, in which pools and on what hardware, is the question that matters.

Client diversity is a specific concern here. The reference implementation is the one the community runs, and client diversity is a separate property from decentralization of mining. A single implementation means a single bug surface.

Ten blocks per second imposes real requirements on node operators, in bandwidth, storage and pruning. A chain that produces blocks faster than most consumer connections can comfortably relay them tends toward fewer independent full nodes over time, which is the trade-off any high-throughput design makes.

Finally, Kaspa is deliberately narrow. It is a payments and settlement network without a general smart contract platform, and that focus is a design choice rather than a missing feature. Judge it as money, not as an application platform.

Frequently asked questions

What is GHOSTDAG?

The ordering rule Kaspa uses. The PHANTOM and GHOSTDAG paper describes it as a greedy algorithm that generalizes Nakamoto consensus to a directed acyclic graph of blocks, with a parameter k controlling tolerance for concurrently created blocks.

How fast does Kaspa produce blocks?

Ten blocks per second. The rusty-kaspa client's documentation states that the Crescendo hard fork on May 5, 2025 moved the network from one block per second to ten.

Was there a premine?

Kaspa states there was none, and publishes a verification of the genesis UTXO commitment against an empty set, with Bitcoin block hashes embedded as timestamps to prove no hidden premining.

How does Kaspa's halving work?

The client source halves the subsidy every twelve months, smoothed through a table of 426 monthly reward-per-second entries, reaching zero after 36 halvings.

Where can you buy Kaspa?

Availability depends on your country or US state. See where to buy Kaspa for the venues serving your jurisdiction and Exchanges to compare fees and kyc requirements.