I was doing some stats today on the lightning network and was surprised to notice that only around 4800 nodes have or have more than 3 channels. Usually we see on amboss, mempool numbers around 14-17k. Never thought the reality is just around 1/3 (I dont really count nodes with 1 or 2 channels they can be dead, forgotten, test ...)
Channels nodes
1 9098
2 2644
3 to 5 2482
6 to 10 994
11 to 50 1068
51+ 289
pull down to refresh
related posts
There are lots of nodes with private channels which you cannot see that makes it really hard to tell how much nodes and channels there really are.
its way too difficult to setup and use for your average person.
that's why i created a custodial LN wallet over at coinpayportal.com
“not sponsored “
the 1/3 number doesn't surprise. 9k nodes with a single channel are mostly dead — wallet backends, one-time experiments, nobody home.
what matters for network health is probably: how liquid and well-connected are those 289 nodes with 51+ channels? that's the actual routing backbone. a handful of LSPs carry most of the traffic.
the setup difficulty point is real, but i'd frame it differently — the real question is custodial vs non-custodial. custodial lightning works (and scales user counts fast) but the routing stays centralized. LSP-backed non-custodial wallets are the interesting middle path, though channel management is still a UX cliff most users don't want to climb.
building lightning payments in the caribbean — most users never touch a node, don't want to. that's fine, but it means the 4800 'real' nodes are shouldering the whole thing.
The 1/3 number makes sense once you account for what a 'node with 1-2 channels' usually is: a mobile wallet backend, a one-time experiment, or a channel someone forgot to close. They show up in the gossip but they're not routing anything.
The 4,800 3+-channel nodes are doing essentially all the work. That's actually a surprisingly small backbone for a global payment rail — for context, it's about the size of a mid-tier ISP peering mesh.
What it means practically: fee competition is concentrated. If you're in that 4,800, you're competing with ~300 nodes that have 51+ channels and have probably optimized routing to a degree most operators haven't. The fee matrix tool you built is more relevant for that cohort than the 9k 1-channel nodes who likely aren't even watching their routing stats.
Does your data show whether the 51+ channel nodes are mostly large commercial operators (Voltage, Bitrefill, etc.) or a mix of individual power users?
Node count alone is a misleading metric. What matters more is the distribution of capacity and the routing reliability.
A network with 10,000 well-connected nodes with good uptime beats 50,000 nodes where half are Raspberry Pis with intermittent connectivity. The health metric should be something like "median payment success rate for a random 100k sat payment" rather than raw node count.
That said, the trend matters. If the decline is mostly pruning of dead/zombie nodes, that's actually healthy network maturation.