Bitcoin developer Paul Sztorc claims more than 80% of Lightning Network payments run through custodial services — and calls the protocol's promotion "a vast campaign of fraud on par with Elizabeth Holmes' Theranos." The allegation, posted October 4 on X, has reopened a long-running argument inside Bitcoin's developer community about whether Layer 2 payments actually work for ordinary users.

"Most people don't realize how bad the Lightning Network is," Sztorc wrote, "because there has been a vast, years long conspiracy to lie about the LN user experience." He did not respond to requests for comment.

There's a data problem at the heart of his claim. Lightning payments are private by design. Nobody can verify the custodial share from public network data alone, a point his critics were quick to make.

What does the criticism actually rest on?

Lightning was built to move Bitcoin faster and cheaper than the base chain. In practice, its payment channels demand liquidity in the right place at the right time, and routed payments can fail — especially larger ones crossing several nodes. Running a node and managing channels remains far too technical for users who just want to buy coffee.

Even Lightning's own head of liquidity, Alex Bosworth, conceded the point this month: "the painful truth is that in a setup with privacy and p2p etc it's extremely slow compared to tap to pay."

That friction explains why users drift toward custodial services. They're easier. They work. The cost is a trusted intermediary, which rather defeats the purpose for some Bitcoiners.

What do the reliability numbers say?

Jesse Shrader, CEO of Lightning infrastructure provider Amboss, pushes back on the doom. Liquidity requirements are "challenging" at first, he says, but "trivial to solve" with current tools or automation.

"For everyday payments of around $100, most users can achieve 99% reliability in the core network," Shrader says. Amboss simulations put the success rate for $100 payments at 99.66%.

That's a benchmark from one infrastructure provider, not a verified network-wide measure. Your results may vary.

Have developers fixed the user experience?

Danny Stagg, marketing lead at payments firm Breez, argues the criticism is outdated. "The criticism is not fair in 2026," he says, "but historically it was. For years, both developers and users were asked to understand the complexities of Lightning, like channels and inbound liquidity, just to get paid."

Over the past two years, newer infrastructure has abstracted that complexity away. "Non-custodial UX is now comparable with custodial," Stagg says. Users no longer need to manage channels or run nodes; developers handle it under the hood. Breez CEO Roy Sheinfeld adds that AI coding tools have made it easier to build Bitcoin-enabled applications, and partner inquiries have climbed.

Skepticism persists at street level, though. In an April 2026 Reddit thread, one user said he was "genuinely impressed" by the network's technical progress but still doubted its everyday utility: "Every coffee shop, grocery store and gas station I walk into runs on Visa or Mastercard terminals. NFC tap to pay is the standard. Nobody is running a Lightning node at the point of sale and nobody is going to."

Do the adoption numbers prove anything?

River's 2026 Bitcoin adoption report estimated Lightning processed $1.17 billion across 5.22 million transactions in November 2025 alone. That volume sits below the network's August 2023 peak of 6.6 million transactions — transaction counts have not kept pace with the value moving through the system.

Payment processor CoinGate reported 782,403 paid crypto orders in the first half of 2026, up 21.4% year over year, with Lightning accounting for 9.6% of the Bitcoin payments it processed.

Each metric measures something different — volume, capacity, developer interest — and none of them alone establishes how easy Lightning is to use.

Can the custody question be settled?

Shrader isn't buying Sztorc's 80% figure. Since network-level statistics are "impossible to produce unless you are party to the transactions themselves," he says, "I distrust anyone who claims they can prove an 80% custodial share either way from public data alone."

Stagg says the claim "describes yesterday's Lightning," before non-custodial services caught up. River estimated back in 2023 one self-custodial user for every eight custodial users. Without current hard data, the custody question stays open.

The sharpest criticism may not be that Lightning failed, but that its supporters conflate what the network can do with what people actually do on it. Until someone produces verifiable custody data — or a merchant point-of-sale footprint that rivals card rails — that gap between capability and adoption will keep fueling this fight.