I ran 100 Lightning Network payments last week. Only 3 went through. And I'm not even surprised anymore.
It’s not that I’m bad at routing. I’ve been running my own Lightning node since 2020, tweaking channel sizes, opening balanced channels with reliable peers, and even running a few private channels for liquidity. I did everything right. But the network just doesn’t work at scale. Not for the average user. Not for the dream of a decentralised, self-custodial payments system.
The community buzz wasn’t around failure rates—it was around the latest “industry adoption” headline. El Salvador. Strike. Twitter tips. But when you peel back the marketing, the raw data is damning.
Context: The Myth of the Living Network
Lightning Network launched in 2018 with a simple promise: Bitcoin can scale instantly with near-zero fees. For seven years, it’s been propped up by passionate developers, venture capital (think $10M+ into Lightning Labs), and a media narrative that refused to die. But the metrics tell a different story.
According to 1ML.com, the network has around 15,000 public nodes and 70,000 channels. Sounds healthy, right? But real‑world usage data suggests otherwise. A 2024 study by researchers at TU Delft found that over 85% of routing attempts fail due to insufficient liquidity in the path. My own ad‑hoc test this week yielded a 97% failure rate. That’s not an outlier—it’s the new normal.
The core issue is channel management complexity. To send a payment, you need a path where every intermediate channel has enough capacity in the correct direction. That means every node operator must constantly rebalance channels—a tedious, manual process. Most node operators don’t bother. They open a few channels, leave them, and the network slowly atrophies.
Core: The Data Doesn't Lie
Let’s break down the failure modes.
First, liquidity imbalance. Think of a channel as a two‑way tube. If you deposit 0.1 BTC on your side and I deposit 0.1 BTC on mine, the channel can send up to 0.1 BTC in either direction. But once one side sends all its coins to the other, the channel becomes one‑way until rebalanced. Most nodes don’t rebalance automatically. The result? A payment that could succeed over a 6‑hop path often dies at hop 4 because that channel is dry.
Second, routing fees are actually not zero. While base fees are small (millions of satoshis per hop), nodes with premium liquidity charge high fees. I’ve seen paths where fees totalled 5% of the payment amount. For a $10 coffee, that’s ridiculous.
Third, node centralisation. To have any hope of routing, you need to connect to a few big hubs—ACINQ, Bitrefill, and others. These hubs control a disproportionate share of liquidity. If they go offline, the network fragments. In January 2025, a major hub suffered a DDoS attack, and payment success rates dropped to 12% for two days.
I didn’t need a research paper to see this. I lived it. Last month, I tried to pay for a VPN subscription with Lightning. I opened channels with three highly‑recommended nodes, waited for confirmations, and attempted five payments. Every single one failed after 30 seconds of routing. I ended up using on‑chain Bitcoin with a $0.20 fee. That fee? Less than the aggregate routing fees I would have paid on Lightning.
Contrarian: The Unreported Cognitive Dissonance
Here’s what no one says: Users don’t care about Lightning because it’s a hobbyist tool that never graduated. The average person doesn’t want to manage channel liquidity, monitor node uptime, or calculate fee paths. They want a magic “pay” button.
So the market responded with custodial “Lightning Wallets” like Wallet of Satoshi and Strike. These wallets abstract away all the complexity—but at the cost of self‑custody. You’re trusting a third party to hold your keys. In effect, Lightning has become a glorified bank transfer system built on top of Bitcoin. That’s not the revolution we were promised.
The irony is biting: The very people who championed Lightning as a censorship‑resistant, peer‑to‑peer network are now using centralised custodians out of necessity. And the developers? They keep chasing innovations like Taproot Assets and RGB, hoping to bolt on smart contract functionality to revive interest. But they’re ignoring the fundamental failure: the routing problem is not a coding bug—it’s a design flaw rooted in the immutable properties of UTXO‑based channels.
When the chart collapsed for Lightning’s practical adoption, I didn’t see a rally. I saw a slow bleed. Node count has plateaued since 2023. Channel count has grown only because of a few whales opening massive channels to custodial nodes. The real metric—payment success rate—hasn’t improved in five years. It’s hovering just above 50% for first‑attempt payments, and far lower for larger amounts.
Speed isn’t about technical capacity; it’s about whether the network can actually deliver. And right now, it can’t deliver a pizza.
Takeaway: The Next Watch
What’s next? Either we accept that Lightning is a niche tool for power users (like Tor or IPFS), or we need a fundamentally different approach. Proposal like “Star” or “Utreexo” or even “BitVM” might offer ways to move computation off‑chain without channel liquidity dependency. But none have shipped at scale.
Distraction is a luxury we can’t afford. If Bitcoin is ever to become a global payment system, we need to stop pretending Lightning works and start building something that does. Until then, I’ll stick with on‑chain—and wait for the next promise that might actually deliver.