Leios scales Cardano sixfold; fees need 36-50 TPS

Input Output, Cardano’s core developer, ran a 41-day public testnet of the Leios scaling protocol. The testnet raised the network’s data-processing ceiling to a peak of 26.8 transaction kilobytes per second (TxkB/s), compared with a 4.51 TxkB/s ceiling for the existing Ouroboros Praos system.

During the stable final days of the phase, Leios carried 54% of traffic reaching the chain. The test produced more than 127,000 blocks and announced about 30,000 Endorser Blocks. Over a comparable period, the network processed 18 times the transaction count seen on Cardano mainnet; most of that traffic was artificially generated to stress the system rather than produced by fee-paying users.

Leios adds parallel capacity by using larger Endorser Blocks to carry additional transactions while the underlying Praos chain continues to operate. Committees validate Endorser Blocks before transactions are recorded, allowing higher throughput without changing Cardano’s security model.

Sixty-three stake pools registered for the test phase. Developers found and fixed implementation issues under load, including memory leaks, chain forks and occasional crashes. Input Output reported that none of the incidents exposed a flaw in the Leios protocol.

The project moved into a Water phase to test different configurations, a redesigned mempool and production cryptographic keys. Input Output has set a goal of making Leios mainnet-ready by the end of 2026 and noted that delivering a consensus upgrade on that timetable would be unusually fast.

Leios increases technical capacity but does not by itself generate fee revenue. Cardano funds staking rewards from transaction fees and periodic releases from an ADA reserve that declines over time. A cost analysis tied to the Leios specification uses a baseline of 48 million ADA in monthly rewards and models a 43% decline by 2029, leaving about 20.64 million ADA per month to be replaced through transaction fees under that scenario.

Using an assumed average transaction size of 1,500 bytes and a modeled fee of roughly 0.221 ADA per transaction, the analysis finds that replacing 20.64 million ADA would require about 36 transactions per second sustained continuously through a 30-day month. The Leios specification frames the relevant economic range at roughly 36 to 50 transactions per second.

TxkB/s and transactions per second are different measures: the testnet’s 26.8 TxkB/s peak measures raw data-processing capacity, while the economic calculation counts fee-paying transactions that must actually arrive. Cardano also directs 20% of its reward pot to the treasury before distribution; after that deduction the same assumptions raise the required throughput to closer to 45 sustained TPS.

Under the model, 50 TPS would produce about 28.7 million ADA in gross monthly fees; 20 TPS would produce roughly 11.5 million ADA; and 10 TPS would produce about 5.7 million ADA. The analysis notes that lower average fees would require more transactions to reach the same revenue, while higher fees raise revenue per transaction and could affect demand.

If fee-paying transactions do not meet the model’s targets, available staking rewards for delegators and pool operators could decline. The Leios design includes a mechanism to limit infrastructure costs when demand is low: ordinary Praos blocks can continue to handle traffic without activating additional Endorser Block work.

Input Output’s specification sets an ambition of increasing monthly transactions from roughly 800,000 to more than 27 million and moving Cardano toward economic self-sufficiency as reserve funding declines. For stake pool operators and ADA delegators, sustained user transactions and the fees they generate will provide the data needed to measure whether the network replaces falling reserve contributions.