A Quarter of Lightning’s Channels Closed. Capacity Barely Moved

Public Lightning capacity is down 7% over the last year. Channel count is down 24%. Both of those are true, both get quoted as evidence the network is failing, and both miss what the third number says: the average channel got 22% bigger. Lightning is not shrinking so much as consolidating — and the data everyone is arguing about is less reliable than either side assumes.

Lightning statistics have a peculiar problem. The network is designed so that most of it is none of your business, and the part you can measure is the part that opted into being public. Every chart you have seen, including ours, is a census of announced channels only. That is the honest starting point, and it should make you suspicious of anyone quoting Lightning capacity to three decimal places in either direction.

What the public graph did over twelve months

Measured from 22 September 2025 to 21 September 2026, on daily snapshots of the public network:

• Capacity: 4,044 BTC → 3,760 BTC. Down 7.0%.
• Channels: 41,604 → 31,680. Down 23.9%.
• Nodes: 17,152 → 16,103. Down 6.1%.
• Average channel size: 9.72M sats → 11.87M sats. Up 22.1%.

Capacity peaked at 5,851 BTC on 28 December 2025 and sits 35.7% below that today. Channel count peaked slightly earlier, at 42,397 on 12 December 2025.

A quarter of the channels disappeared and only 7% of the capacity went with them. The channels that closed were the small ones.

Two readings, and the honest one

The bearish reading writes itself: fewer channels, less capacity, fewer nodes, a third off the peak. Lightning had its moment and the moment passed.

The bullish reading is equally available: capacity per channel is up sharply, which is what a maturing payments network looks like when hobbyist channels close and routing consolidates into operators who actually move volume and earn fees.

The honest reading is that both are stories told about the same four numbers, and the data cannot arbitrate between them, because the thing that would settle it — payment volume — is exactly what Lightning is built not to reveal. Nobody publishes Lightning transaction counts because nobody can. A capacity chart tells you how much water is in the pipes, not how much flowed through them.

What can be said without straining: the public network is smaller and more concentrated than a year ago, and concentration is visible inside the numbers too. The average public channel holds 11.87M sats while the median holds 2.08M — the mean is nearly six times the median, so a small number of very large channels carry most of the public capacity.

The measurement problem nobody mentions

While building the Lightning cards on this dashboard we found something that should change how you read any Lightning statistic, including the ones above.

The public data comes from mempool.space, which publishes community mirrors serving the identical API. Queried within the same minute, three of those mirrors disagreed:

• One returned a snapshot dated 8 July 2026 showing 4,532 BTC.
• Another returned 21 September 2026 showing 3,754 BTC.
• A third returned 30 August 2026, also showing 3,754 BTC.

A snapshot 75 days stale, presented by a working API with a 200 response and perfectly well-formed data, reading 20% higher than reality. Nothing about the response says it is old unless you read the added field and compare it to today.

That is not a criticism of mempool.space, which publishes the timestamp honestly. It is a warning about everything downstream. Any dashboard, article or chart that pulled capacity without checking that date has been quoting a July number ever since, and would show Lightning as roughly flat over the year rather than down 7%. We now take the freshest snapshot across mirrors and print its date on the card, because we were briefly guilty of exactly this.

What the public number cannot see

Three blind spots, all of which push the same way — the real network is larger than the measured one.

Unannounced channels are invisible. A channel only appears in the public graph if both parties announce it. Private channels, which are the normal arrangement between a wallet and its service provider, never show up at all. The node count above includes 2,014 nodes that announce themselves but not their channels — visible participants whose capacity is missing from every total, including ours.

Big custodial flows bypass the graph entirely. When two users of the same exchange or wallet pay each other over Lightning, the payment may never touch a public channel.

Capacity is not throughput. A channel can route its own capacity many times in a day, or sit idle for a year. Falling capacity with rising utilisation would look identical to decline in every chart published.

How to read Lightning numbers

Treat public capacity as a floor and a trend line, never as the size of the network. Watch capacity and channel count together, because the gap between them is the only consolidation signal the public data offers. And check the date on any figure you are given — we found a 75-day-old number being served as current by a live API, which means the number in the article you are reading elsewhere might be older than the argument it is supporting.

The live Lightning page shows capacity, channels and nodes with the snapshot date attached, for exactly this reason.

Frequently asked questions

Is the Lightning Network growing? The public network is not, by capacity or channel count: capacity is down 7.0% and channels down 23.9% over the last year. Average channel size is up 22.1%, which is consistent with consolidation rather than abandonment. Private channels and payment volume are not measurable, so no public figure can settle the question.

How much Bitcoin is on Lightning? About 3,760 BTC in public channels as of 21 September 2026, roughly $322M. That is a floor rather than a total, because unannounced channels are excluded from the public graph by design.

Why do Lightning statistics differ between sites? Partly because they measure different things, and partly because the public data sources fall out of sync. Three mirrors of the same API, queried in the same minute, returned snapshots dated 8 July, 30 August and 21 September 2026, with the oldest reading 20% higher than the newest.

Related reading: Your Fee Is Not About Timing on what on-chain transactions cost when you cannot use a payment channel, and What Is Bitcoin Hashrate? on the base layer Lightning settles to.

Figures from mempool.space Lightning statistics, 357 daily snapshots covering 22 September 2025 to 21 September 2026, taken from the freshest available mirror on 21 September 2026. Dollar value uses a Bitcoin price of $85,645. Capacity and channel figures cover announced channels only; node figures include nodes that announce themselves without announcing channels.