Paul Sztorc Says Lightning Becomes Banking. The Channel Data Says He's Half Right.

Last week Paul Sztorc, the founder of Bitcoin Hivemind and the architect of the drivechain proposal, said out loud what much of the Lightning community would rather bury: that the network's custodial wallets are self-deception, and that Lightning is drifting toward the very institution Bitcoin was built to escape. I have spent enough nights staring at channel graphs to know he is describing something real. I have also spent enough years auditing protocols to know that a real observation and a correct conclusion are two different animals. What Sztorc is pointing at is not a bug. It is a market outcome — and market outcomes are harder to patch than code. Bitcoin's block space was never going to absorb the world's payments. That question was settled in 2017, when the scaling wars split the community and Lightning emerged as the winning answer: keep the base layer scarce and expensive, move speed and volume to a second layer built on payment channels. The pitch was elegant. Two parties lock funds into a multisig output, then rebalance the channel off-chain as often as they like. Settlement only touches L1 when the channel closes. Cheap, fast, and — in theory — trustless. For a base layer obsessed with verifiability, it looked like the clean way out. Paul Sztorc is not a neutral commentator on that outcome. He built Bitcoin Hivemind, he has championed drivechains as a competing scaling philosophy, and he has been skeptical of L2s for years. That history matters, and I'll come back to it. But the specific claim he made — that the UX demands of Lightning inevitably produce custody, and that custody is just banking with extra steps — deserves to be tested on its mechanics rather than dismissed on its source. Here is the geometry, and geometry is the right word for it. A Lightning channel has two directions. You need outbound liquidity to send and inbound liquidity to receive, and those two capacities are not symmetric. When you open a channel from your own node, you buy outbound capacity and almost no inbound. To receive your first payment, someone else's node must have already committed capital toward you — or you must pay a service to arrange it. That asymmetry is the seed of everything that follows. The market solved it the way markets do. First it built LSPs — Lightning Service Providers — that sell you inbound liquidity for a fee. Then it built custodial wallets that skip the problem entirely: you hand your sats to an operator, they run a fat node with deep liquidity, and your so-called Lightning wallet becomes a thin client pointed at someone else's balance sheet. That is not a Lightning wallet. That is a bank account wearing a Lightning-shaped interface. The balance isn't the point; who owns it is. The public data supports the shape of the criticism, with a caveat I want to be precise about. Studies of Lightning's public graph have repeatedly shown that a small number of large nodes — exchanges, LSPs, well-capitalized routing hubs — account for a disproportionate share of visible capacity. That is real. But the public graph is not the network. Private channels never appear in it, and they carry a large and fundamentally unmeasurable share of traffic. So when someone tells you 'Lightning is centralized, the data proves it,' the honest reply is: the visible layer is concentrated, and the invisible layer is unknown. Unknown is not the same as fine. When I audited early prediction-market oracles back in 2017, I learned to separate the protocol's guarantee from the operator's behavior. The contract said one thing; the deployment quietly did another. Lightning has the same split. The protocol is trustless. The mobile experience is not. And users, overwhelmingly, choose the experience over the guarantee. They always do. Nobody opens a node to buy coffee. Picture the day in the life of an ordinary user. She downloads a wallet, taps 'receive,' and expects it to work like a bank app. Behind that single tap sits a chain of dependencies she will never see: a routing node with enough inbound capacity, a path with enough outbound capacity on the other side, a fee estimate that must beat the base layer's mempool, and an online device to sign the final hop. Every one of those dependencies is a place where the trustless version can fail. The custodial version removes all of them at once — and in doing so, removes her custody. That trade is not irrational. It is exactly the trade banking has offered for centuries, and it keeps winning for the same reason: it is the only version that works on a phone that goes into a pocket. This is where the critique stops being about Lightning and starts being about a pattern. I watched the same thing happen in RWA tokenization, where the pitch was 'bring real assets on-chain' and the reality was a permissioned wrapper around a custodian who already held the assets and never needed a public chain in the first place. I watched it in DAO governance, where 'decentralized autonomous organization' described a smart contract that, in most jurisdictions, has the legal standing of a group chat — so when something breaks, the 'members' can discover they hold unlimited personal liability for a treasury they never actually controlled. The technology promised to dissolve the intermediary. The market kept rebuilding the intermediary, because the intermediary solves a problem the technology refused to. In Lightning's case, the problem the technology refused to solve is liquidity bootstrapping. Routing payments requires capital sitting idle in the right places, and idle capital has a cost. Someone has to eat that cost. If the user won't, the operator will — and the operator will want something in return. That something is custody. This is not a conspiracy. It is an invoice. So Sztorc is right that custody is winning. But the frame he reaches for — 'Lightning becomes banking, therefore Lightning failed' — smuggles in an assumption I don't accept. Banking is not a failure mode of payments. Banking is what happens when a system has to handle liquidity, risk, and uptime at the same time, and no purely peer-to-peer design has solved all three at retail scale. The real question is never 'custody or no custody.' It is who carries the liability when the custodian fails, and whether the user can still leave. A custodial wallet with a working exit is a different animal from one without. And the conflict of interest runs under the surface of the argument. Sztorc is a drivechain advocate. A well-argued case that L2s inevitably centralize is, conveniently, a case for his alternative. That does not make him wrong — it makes him interested. Open source isn't a guarantee of neutrality; it's a philosophy of transparency, and transparency has to include seeing the incentives behind the critic. The sharpest version of his critique isn't 'Lightning is centralizing.' It's that Lightning never priced the cost of self-custody into the user experience, so the market quietly priced it out. That is a design debt, not a betrayal. Decentralization is not a tech stack; it's a set of incentives that either reward exit or punish it. So watch the non-custodial wallets — Phoenix, Zeus, and the LSP model that lets them compete on a phone — not because they are destined to win, but because they are the only live test of whether self-custody survives contact with a normal person's daily life. If it can't, Sztorc's warning stops being a prophecy. It becomes a description.