Lightning’s limits, Ark and the new payment layers

Wonder

1. Why does a payment channel that holds ten bitcoin still fail to send nine?
2. What happens to a user who goes offline for two days while their counterparty watches the chain?
3. Can a network holding a few thousand bitcoin of public capacity serve billions of potential users?

Explore

Part ILightning’s limits

1. A payment channel is a finite buffer, not an infinite pipe. The total capacity is fixed at opening; as funds move from one side to the other, the outbound balance on the sender’s end shrinks while the inbound balance grows. A user who spends aggressively depletes their ability to send further until they receive payments or rebalance through another route. Large transfers that exceed a channel’s remaining outbound capacity fail instantly, forcing retries or fallbacks to slower methods.
2. Receiving liquidity is structurally harder to obtain than sending liquidity for new participants. Opening a channel with one’s own funds creates only outbound capacity; the counterparty holds all the inbound balance. To receive payments, a user must either spend first (moving funds out), ask someone else to open a channel toward them, or purchase inbound liquidity from a service provider. This asymmetry creates a persistent dependency on third parties or prior transaction history for new nodes entering the network.
3. Non-custodial security requires constant vigilance. If a node goes offline while its counterparty broadcasts an outdated commitment state, the offline user has a limited window to respond with a penalty transaction using revocation secrets. Missing the window lets the cheater keep whatever the old state paid them. Watchtowers can monitor on users’ behalf, though the user must rely on the tower staying online and acting; a tower cannot steal funds. For many users, keeping a node online twenty-four hours a day is a practical burden that custodial services absorb by managing infrastructure internally.
4. The network’s topology is unevenly distributed. Routing efficiency depends on large, well-funded nodes holding substantial liquidity across multiple channels to facilitate high-volume transfers. Smaller nodes often lack the capacity or connectivity to route significant payments independently, creating bottlenecks. Probing attacks can reveal channel balances by sending test payments that fail at specific hops, allowing observers to map where liquidity sits, which weakens payment privacy.
5. On-chain costs remain a hard constraint on adoption. Every channel opening and closing requires a Bitcoin transaction, incurring fees that spike during periods of high mempool activity. For users who open and close channels frequently, these fees can erode the savings gained from off-chain payments. Designs like channel factories and covenants aim to reduce this overhead by allowing multiple users to share a single on-chain output or enforce complex scripts without bloating the chain, but widespread implementation is still pending.
6. Public network capacity remains modest relative to Bitcoin’s total supply. As of recent observations, the sum of all public channels hovers in the low thousands of BTC, while private channels, which are not visible to the graph, account for a significant but unknown portion. This limited public visibility makes it difficult to assess true network health or liquidity depth. The replacement-cycling attack disclosed in 2023 showed that an attacker could exploit mempool replacement rules to keep a routing node’s timeout transaction from confirming until an upstream deadline passed, then take the funds; mitigations reduce but have not eliminated the risk.

Part IIArk and the new payment layers

7. Burak Keceli proposed Ark in May 2023 as a second-layer protocol that allows many users to share UTXOs coordinated by an operator. Unlike Lightning, which requires two-party channels with pre-funded liquidity, Ark aggregates balances into shared outputs managed by the operator. This structure reduces the on-chain footprint per user significantly, aiming for global self-custody adoption without the complexity of channel management.
8. The core mechanism relies on the operator coordinating payments off-chain while maintaining a valid state on Bitcoin. Users can exit unilaterally to their own addresses at any time before their shared outputs expire by submitting an on-chain transaction. This exit path is critical: it prevents the operator from freezing funds indefinitely, distinguishing Ark from custodial models where the issuer holds sole control over the keys.
9. Ark Labs launched Arkade, making its first mainnet payments in August 2025 and opening publicly in October 2025. Simultaneously, Second developed Bark, a distinct implementation that hit mainnet on June 9, 2026. The two systems are not interoperable; users cannot move funds directly between Arkade and Bark without exiting to the base layer first, creating fragmented liquidity pools.
10. Compared to Lightning, Ark eliminates the need for inbound liquidity to receive payments. On Lightning, a user must open channels with peers to accept payments, requiring careful liquidity management. Ark simplifies this by allowing the operator to manage channel-like states internally, and its shared transactions give users some CoinJoin-like privacy.
11. The trade-offs involve operator liveness and exit costs. If an operator goes offline, users can still exit on-chain, but they must pay transaction fees and wait for confirmation. Timeouts govern how quickly a user can force an exit if the operator fails to sign cooperative transactions. Shared outputs also expire: in the original design users must come online to refresh their balances before a timeout, often weeks, or the operator can reclaim the output.
12. Ecash systems offer privacy but are custodial: Cashu relies on a single mint, Fedimint on a federation, and both hold the bitcoin. Ark aims for non-custodial self-custody with similar privacy benefits. The debate continues on whether this complexity is justified compared to Lightning’s mature ecosystem, as both protocols coexist and may complement each other in future payment stacks.

Key facts

Exercise nine, six, three, one

Fact check

Dates, figures and claims checked against the published sources listed below.

• A. M. Antonopoulos, O. Osuntokun & R. Pickhardt, Mastering the Lightning Network (2021)
• A. Riard, replacement cycling disclosure (bitcoin-dev mailing list, October 2023)
• Public network statistics from 1ML and mempool.space Lightning
• Ark protocol documentation (ark-protocol.org)
• Bitcoin Magazine, 'Bark Launches on Mainnet' (June 2026)
• Ark Labs announcements (August and October 2025)