Cardano’s Leios upgrade demonstrated sixfold throughput gains in its first public testnet run, but the numbers that matter for ADA holders are economic, not technical. The staking model still depends on users paying fees, and the reserve that subsidizes rewards is running down on schedule.
Input Output, the network’s development firm, reported that Leios peaked at 26.8 transaction kilobytes per second during a 41-day test, against a 4.51 TxkB/s ceiling for the current Ouroboros Praos system. The network produced more than 127,000 blocks and about 30,000 Endorser Blocks, with 63 stake pools taking part. In the stable final days, Leios carried 54% of traffic and processed 18 times the transaction count seen on Cardano mainnet over the same period.
There is a catch that the headline numbers hide. All of that traffic was synthetic, generated to stress the system rather than produced by real users paying to transact. Capacity and demand are two different things, and only one of them has been proven so far.
How Leios actually works
Leios runs as an overlay on Cardano’s existing consensus rather than a replacement. When demand rises, an elected slot leader produces an additional endorser block that travels in parallel with the standard block. Committee-based validation and cryptographic techniques, including SNARKs for compact correctness checks and data availability sampling, let the network verify more data without every node downloading everything. UTXO sharding spreads state maintenance across the network so consumer-grade hardware can still participate.
Input Output product manager Carlos Lopez de Lara has said the initial rollout starts at two to five times current throughput, with the full ceiling available as demand grows. The long-term goal is scaling from roughly 800,000 transactions per month to more than 27 million, a 30x increase, while the network becomes economically self-sufficient as reserve funding declines. Leios is the consensus upgrade meant to provide the capacity for that growth.
The design went through a formal specification process under CIP-164, which the community merged after review. Input Output structured the 2026/2027 cycle around a release candidate that moves through software readiness levels five to eight, including a rewrite of consensus components, conformance testing against the formal specification, and full integration into the primary node by the fourth quarter.
Why capacity alone does not pay delegators
Cardano’s staking rewards blend transaction fees with releases from its ADA reserve, and the reserve share declines over time by design. A cost model in the Leios specification assumes 48 million ADA in monthly rewards falling 43% by 2029, which would leave 20.64 million ADA to be covered by fees. At an average fee of roughly 0.221 ADA, that requires about 36 sustained transactions per second, or closer to 45 after the 20% treasury deduction.
Leios supplies the capacity for that transition. Whether users fill the blockspace is the open question. If usage falls short, the pressure moves elsewhere: smaller pools with fixed infrastructure costs face pressure to consolidate or accept lower earnings, and delegators see thinner rewards after operators take their declared costs and margins. Higher transaction fees are another lever, but only if users keep paying them.
Developers are now in the Water testnet phase, one of five stages named after chapters of Miyamoto Musashi’s Book of Five Rings, with mainnet readiness targeted by the end of 2026. A governance proposal for ₳27.7 million to mature Leios passed with over 84% support from delegated representatives, and a hard fork could arrive as early as November. Separately, the Amaru project, an open-source Rust rewrite of the Cardano node, is already syncing the chain and targets block production on mainnet in November, reducing the network’s reliance on its historical Haskell client.
| Measure | Ouroboros Praos | Leios testnet peak |
|---|---|---|
| Throughput | 4.51 TxkB/s | 26.8 TxkB/s |
| Blocks produced | 127,000+ | |
| Endorser Blocks | About 30,000 | |
| Stake pools in test | 63 |
The price backdrop is unforgiving
None of this arrives in a friendly market. ADA trades near $0.20, down roughly 35% over 30 days earlier this summer and sitting about 95% below its September 2021 all-time high of $3.09. The analytics platform TapTools shut down earlier this year, Cardano cancelled its 2026 Singapore Summit, and founder Charles Hoskinson warned of a wave of failures among Cardano DeFi projects.
The network kept shipping while the price went nowhere, which is an unusual contradiction. Leios removes the most persistent technical criticism of the chain, that the base layer cannot scale. What it cannot remove is the need for demand. Speed does not pay delegators, transactions do.
There is also a governance wrinkle worth noting. The constitutional committee renewal cleared its voting thresholds in early September by a margin of just 0.18% among stake pool operators, a sign that the community’s appetite for contested votes is thin even as the roadmap demands more of them.
The treasury funding structure adds another constraint. Every ada disbursed for Leios development is milestone-gated and independently assured, with unspent funds returned to the reserve. That protects delegators from open-ended spending, but it also means the ecosystem’s biggest engineering project consumes treasury budget at the same moment fee income is weakest.
For ADA holders, the question is whether the market reprices during the remaining testnet months or waits for mainnet. Either way, the technical story and the economic story have now separated. The chain can move six times faster. Whether anyone will pay to use that speed is the harder problem, and no protocol upgrade solves it by itself.