Back to ResearchTopic: Protocol Fundamentals and On-Chain Data

Lightning Channel Open and Close Cost Tables

C

Research Desk • Organizational attribution

Source Standard
6 source notes
Last Reviewed
No edits since publication
Not Advice
Educational research, not a recommendation to buy or sell

Opening a Lightning channel from one native SegWit coin and closing it cooperatively costs 322 sats in on-chain fees at 1 sat/vB, 6,435 sats at 20 sat/vB and 64,350 sats at 200 sat/vB. A force close costs more, and keeping open plus close at or under 5 percent of the channel needs at least 128,700 sats at 20 sat/vB. Every figure below is computed from the transaction weights in BOLT 3, the Lightning specification.

The one input nobody controls is the fee rate. On November 3, 2026 at 17:57 UTC, with Bitcoin at block 974,419, the mempool.space recommended-fee endpoint returned 1 sat/vB for every confirmation target, and its projected-block endpoint showed a median of about 0.12 sat/vB for the first block. That is a single observation, not a trend, and it is the reason the tables below run from 1 to 200 sat/vB: a channel lives for months, and its closing transaction is priced on the day it is broadcast, not on the day the channel opened. At 1 sat/vB, the rate the endpoint recommended, the round trip is 322 sats. At 200 sat/vB the same transactions cost 64,350 sats.

Lightning payments do not pay on-chain fees. The on-chain bill sits at the two ends of a channel's life: one funding transaction to open it, then either one cooperative closing transaction or a longer chain of transactions after a force close. Pricing both ends from the specification's own weights shows which part of the bill depends on the channel, which on the fee rate, and which on choices such as the type of coin you fund from. For the fee mechanics that apply to every Bitcoin transaction, including how sat/vB is set and how a stuck transaction is replaced, see our guide to Bitcoin transaction fees and the mempool.

What Goes On-Chain, and Who Pays for Each Piece

A channel starts as an ordinary Bitcoin transaction. BOLT 3 says the funding output is a pay-to-witness-script-hash (P2WSH) output locked to a 2-of-2 multisignature of the two nodes' funding keys. Whoever builds that transaction pays its fee. In a single-funded channel that is the opener. In a dual-funded channel BOLT 3 splits it: "fees are paid by each party to the transaction, at a feerate determined during the initiation, with the initiator covering the fees for the common transaction fields."

After funding, each side holds a signed commitment transaction that spends the funding output and pays out the current balances. Nobody broadcasts it in normal operation. It exists as the fallback. BOLT 3 delays the broadcaster's own money on purpose: "all outputs that return funds to the owner of the commitment transaction (a.k.a. the "local node") must be delayed for to_self_delay blocks." The fee for that fallback is already settled in the channel's accounting: "Base commitment transaction fees and amounts for to_local_anchor and to_remote_anchor outputs are extracted from the funder's amount." BOLT 2 states the same rule from the other side: "the party which requested the channel creation always pays the fees for the commitment transaction."

A cooperative close replaces all of that with one new transaction that spends the funding output directly to each side's chosen address. BOLT 2 requires the channel to be empty of pending payments first: "Once shutdown is complete, the channel is empty of HTLCs, there are no commitments for which a revocation is owed, and all updates are included on both commitments, the final current commitment transactions will have no HTLCs." A node also "MUST NOT send an update_add_htlc after a shutdown." The legacy negotiation opens with the same sentence. In both versions the closing transaction is built only after the channel is empty of HTLCs, so pending HTLCs have to settle or fail first, which is why the HTLC counts in this article appear only in the force-close tables.

Who pays the closing fee depends on the negotiation version. In the legacy closing negotiation BOLT 3 says each node "MUST subtract the fee given by fee_satoshis from the output to the funder." In the newer simple close, "the node sending closing_complete ("the closer") pays the fees." The tables assume the opener is both funder and closer, so one party pays everything. In a channel where the other side closes, the bill moves with it.

TransactionWho paysAppears when
Funding transactionThe opener in a single-funded channel; each side for its own inputs and outputs in a dual-funded oneAlways, once per channel
Cooperative closing transactionThe funder (legacy negotiation) or the closer (simple close)Both sides agree to close and no HTLCs remain
Commitment transaction, broadcastThe funder, out of its channel balanceForce close only
HTLC-timeout or HTLC-success transactionOut of the HTLC value when there are no anchors; zero fee inside the transaction with anchorsForce close with HTLCs on the commitment
Sweep of delayed outputsWhoever owns the delayed outputAfter a force close, once the delay has passed
Child transaction on an anchorThe party that wants the commitment confirmedAnchor or zero-fee channels that need a rate bump

Payer rules from BOLT 2 and BOLT 3 at the commits listed in the sources. The HTLC-transaction row is explained in the force-close sections.

BOLT 5 explains why a cooperative close is preferred: "There exists a bias towards preferring mutual closes over unilateral closes, because outputs of the former are unencumbered by a delay and are directly spendable by wallets. In addition, mutual close fees tend to be less exaggerated than those of commitment transactions (or in the case of commitments using anchor outputs, the commitment transaction may require a child transaction to cause it to be mined)." The numbers below put sizes on both halves of that sentence.

The Weights Behind Every Figure

A transaction's weight is its non-witness bytes times four plus its witness bytes, and its virtual size in vB is the weight divided by four. A fee rate in sat/vB times the vB gives the fee. The specification works in satoshis per 1,000 weight units, written sat/kw, and BOLT 2 notes that this is "1/4 the more normally-used 'satoshi per 1000 vbytes'". So 1 sat/vB equals 250 sat/kw, 20 sat/vB equals 5,000 sat/kw, and 200 sat/vB equals 50,000 sat/kw.

BOLT 3 fixes the weights it needs for fees in one place: "Commitment weight (no option_anchors): 724 + 172 num-untrimmed-htlc-outputs" and "Commitment weight (option_anchors): 1124 + 172 num-untrimmed-htlc-outputs", with HTLC-timeout at 663 or 666 and HTLC-success at 703 or 706. The spec is candid that these are simplified: they assume "Signatures are 73 bytes long (the maximum length)", "There are always both a to_local output and a to_remote output", and, with anchors, both anchor outputs. It adds that "The actual fee may be higher than the amount calculated here, due to rounding and trimmed outputs." The tables here use the formula weights, so they are slightly high for a freshly funded channel that has no output for the remote side.

The 724 is 125 bytes of non-witness data times four, which is 500, plus 224 weight units for the segwit marker, flag and the two-signature witness. The anchor variant has 225 bytes: two 43-byte anchor outputs add 86, the to_remote output grows from a 31-byte P2WPKH to a 43-byte P2WSH for another 12, and the spec counts two more bytes in the output-count field. That is 100 bytes, or 400 weight units, which is the whole difference between 724 and 1,124.

ItemWeight (WU)vBWhere it comes from
Commitment transaction, no anchors, no HTLCs724181BOLT 3, Fee Calculation
Commitment transaction, anchors, no HTLCs1,124281BOLT 3, Fee Calculation
Each untrimmed HTLC output on the commitment17243BOLT 3, Fee Calculation
HTLC-timeout transaction (no anchors / anchors)663 / 666165.75 / 166.50BOLT 3, Fee Calculation
HTLC-success transaction (no anchors / anchors)703 / 706175.75 / 176.50BOLT 3, Fee Calculation
Funding transaction, set A (1 P2WPKH input, change)611152.75BOLT 3, Appendix A formula, input type chosen here
Cooperative close, two P2WPKH outputs676169.00Derived from Appendix A component sizes
Cooperative close, one P2WPKH output552138.00Derived from Appendix A component sizes
Sweep of one delayed output into one P2WPKH output490122.50Derived; witness size from BOLT 5, Appendix A
Each further delayed input in the same sweep32481.00BOLT 5 to_local penalty input weight; the delayed spend has the same witness length
CPFP child spending one anchor plus one wallet input719179.75Derived from the BOLT 3 anchor script

Rows from BOLT 3 are quoted values. Rows marked derived are computed in the same script from the byte sizes in BOLT 3 Appendix A and BOLT 5 Appendix A, because the specification gives no expected-weight constant for those transactions.

Two rounding rules apply. BOLT 3 rounds commitment and HTLC fees down: "Multiply feerate_per_kw by weight, divide by 1000 (rounding down)." Its funding-transaction example in Appendix G rounds up, giving 155 sat for a 609-weight transaction at 253 sat/kw where rounding down gives 154. The tables follow both: commitment and HTLC fees are rounded down, and every other fee (funding, closing, sweeps, child transactions) is rounded up, which is what wallet software does to meet its target rate. Any figure moves by at most one satoshi per transaction under the other convention.

These weights reproduce the specification's own examples. The Appendix A byte counts rebuild 724 and 1,124 for the commitment and 663 and 666 for HTLC-timeout. The Fee Calculation example (5,000 sat/kw, a 546 sat dust limit, four HTLCs) comes out at fees of 3,315 and 3,515 sat for the two HTLC transaction types, a commitment weight of 1,068, a base fee of 5,340 sat and an actual fee of 7,140 sat. The Appendix G opener comes out at 609 weight units and 155 sat. The trim boundaries encoded in the Appendix C test vectors also come out right: with a 546 sat dust limit, the 1,000 sat received HTLC survives at 647 sat/kw and is trimmed at 648, the 2,000 sat received HTLC at 2,069 and 2,070, the 2,000 sat offered HTLC at 2,194 and 2,195, the 3,000 sat offered HTLC at 3,702 and 3,703, and the 4,000 sat received HTLC at 4,914 and 4,915. One detail from the spec is worth knowing: it notes the HTLC-success weight is "really 702 and 705, but we use these numbers for historical reasons", so 703 and 706 are the values that determine real fees on real channels.

Assumptions That Change the Numbers

Every table depends on a short list of choices. Change one and the totals move, so they are stated here once.

Coin type. The funding transaction spends native SegWit (P2WPKH) inputs and returns change to a P2WPKH address. Each P2WPKH input adds 273 weight units: 164 for the 41 non-witness bytes and 109 for a witness holding a signature at the 73-byte bound the spec uses. Other coin types change the witness size; they are not priced here. Our guide to Bitcoin address types explains how those formats differ.
Who pays. The opener is the funder and also the closer, so it pays the funding fee, the commitment fee in any force close, and the closing or sweeping fees.
Closing outputs. Both parties hold a balance above dust, so a cooperative close has two P2WPKH outputs. A one-output variant is shown separately.
Fee rate. One rate applies to every transaction in a scenario. For channels without anchors the commitment is assumed to have been signed at that rate, which is the best case: the rate in a legacy commitment is whatever was last agreed through an update_fee message, not whatever the market charges at close time.
HTLCs. Each pending HTLC is 50,000 sat, large enough to stay on the commitment at all six rates (the trim section shows the thresholds). All of them are offered by the closer and have timed out, which is the worst case for the closer. Each needs a second-stage transaction, and the closer sweeps the results in one batched transaction.
Dust limit. Trim thresholds use 354 sat, the smallest dust_limit_satoshis BOLT 2 allows a peer to announce. Higher limits shift each threshold up one for one.
Anchors. The anchor rows price each HTLC-timeout transaction at its 666 weight units and the target rate, but leave out the wallet input and change output that must be added to pay for it, and they do not subtract the two 330 sat anchor outputs from the funder's balance. With pending HTLCs they are therefore lower bounds.
Units. Everything is in satoshis. Converting to dollars needs a dated price, and the fee mechanics do not depend on one.

Funding Transaction Weight for Four Input and Output Sets

The funding transaction is the only cost every channel pays, so its size is worth computing before anything else. BOLT 3 Appendix A gives the formula: four times the sum of 43 bytes for the funding output, 41 bytes per input, 9 bytes plus the script length per other output and 10 bytes of transaction fields, plus 2 weight units of segwit marker and flag, plus the witness weight of every input.

SetInputs and outputsWeight (WU)vB1 sat/vB2 sat/vB5 sat/vB20 sat/vB50 sat/vB200 sat/vB
A1 P2WPKH input, funding output, 1 P2WPKH change output611152.751533067643,0557,63830,550
B2 P2WPKH inputs, funding output, 1 P2WPKH change output884221.002214421,1054,42011,05044,200
C1 P2WPKH input, funding output only (no change)487121.751222446092,4356,08824,350
D1 P2WPKH input, 3 funding outputs (3 channels), 1 change output955238.752394781,1944,77511,93847,750

Fees in sats, rounded up, at the fee rates shown. Set A is used in every cost table below. All inputs are P2WPKH with the 73-byte signature bound; all change outputs are P2WPKH.

Set A is a single P2WPKH coin larger than the channel, with change returned: 611 weight units, or 152.75 vB. Dropping the change output (set C) removes 124 weight units, a saving of 31 vB, but it requires a coin that matches the channel size almost exactly, so the channel capacity is whatever the coin holds minus the fee. Adding a second input (set B) adds 273 weight units, which is 68.25 vB. At 50 sat/vB that second input costs about 3,412 sats, so the cost of assembling a channel from several small coins is real and grows with the fee rate.

Set D prices a funding transaction that opens three channels at once from one input, with one change output. Its 955 weight units work out to 318.33 per channel, against 611 for a stand-alone channel. At 50 sat/vB, three stand-alone openings cost 22,914 sats and the batched transaction costs 11,938, a difference of 10,976 sats. The saving comes from sharing the input, the transaction fields and the change output. Batching pays off when several channels are opened together.

The Appendix G opener example uses a witness weight of 107, two units below the 109 used here, because it takes the minimum signature length rather than the maximum. That makes the example transaction 609 weight units against 611 for set A with the longer signature, and it is also the reason the funding fee on a real transaction can land one or two satoshis under or over these figures.

If the coin comes from an exchange withdrawal first, that withdrawal is its own on-chain payment with its own fee and is not part of these tables. Our walkthrough of moving coins from an exchange to self-custody covers what that step costs and how to check it.

Open Plus Cooperative Close at Six Fee Rates

The cooperative closing transaction spends the funding output and pays the two balances out. Its size is not given as a constant in BOLT 3, so it is built from the same Appendix A parts: the 222-byte two-signature witness plus the 71-byte multisignature script for the input, 41 non-witness bytes of input, 31 bytes for each P2WPKH output and 10 bytes of fields. With two P2WPKH outputs that is 676 weight units, or 169.00 vB. With one output it is 552, or 138.00 vB.

Closing transactionWeight (WU)vB1 sat/vB2 sat/vB5 sat/vB20 sat/vB50 sat/vB200 sat/vB
One P2WPKH output552138.001382766902,7606,90027,600
Two P2WPKH outputs676169.001693388453,3808,45033,800
Two outputs with 43-byte scripts (P2WSH or P2TR)772193.001933869653,8609,65038,600

Fees in sats, rounded up. The wider-script row applies when either party's closing address is P2WSH, or a taproot address where the nodes have negotiated support for it.

The wider-script row matters because BOLT 2 lets a node choose its payout script from several forms: version 0 pay-to-witness-pubkey-hash, version 0 pay-to-witness-script-hash, and "if (and only if) option_shutdown_anysegwit is negotiated" a later segwit version. If both payouts use 43-byte scripts, the closing transaction is 96 weight units bigger than the two-P2WPKH case, 24 vB, which is the price of a script type that may suit the wallet receiving the funds.

Adding the funding fee from set A to the cooperative close gives the round trip.

Fee rateFunding (set A)Close, two outputsOpen + closeOpen + close, one output
1 sat/vB153169322291
2 sat/vB306338644582
5 sat/vB7648451,6091,454
20 sat/vB3,0553,3806,4355,815
50 sat/vB7,6388,45016,08814,538
200 sat/vB30,55033,80064,35058,150

Fees in sats at a constant rate. The last column assumes the other side's balance is below dust, so the closing transaction has a single output, as happens when a channel is closed before any payment has moved to the other side.

At 1 sat/vB the round trip is 322 sats, split 153 for opening and 169 for closing. At 20 sat/vB it is 6,435. At 200 sat/vB it is 64,350, and the close alone is 33,800, a little more than the open because the two-signature witness is larger than the single-signature witness of the funding input. Cost is proportional to the rate, so the rows scale: the 200 sat/vB total is 199.8 times the 1 sat/vB total, and the small gap from exactly 200 comes from rounding up each transaction.

The one-output column saves 31 vB, which is 620 sats at 20 sat/vB. BOLT 2 describes how the shortfall is handled: "The lesser-paid peer (if either is) can opt to omit their own output from the closing transaction." It also warns that dust must be omitted "to ensure that the resulting transaction can be broadcast". A channel closed right after opening, with the entire balance on the opener's side, falls in that case.

Force Close With 0, 5 and 30 Pending HTLCs

A force close is the opposite of the cooperative path. One side broadcasts its latest commitment transaction, and BOLT 5 spells out when it may. A node that has received a valid closing_signed message with a sufficient fee "SHOULD use this fee to perform a mutual close". Otherwise, if the node "knows or assumes its channel state is outdated", it "MUST NOT broadcast its last commitment transaction", and if not, it "MUST broadcast the last commitment transaction, for which it has a signature, to perform a unilateral close." The broadcaster then waits out the delay on its own balance and sweeps it, and any HTLC on the commitment needs its own follow-up transaction.

For a channel without anchors, the closer pays three things. First, the commitment fee, from the funder's balance: BOLT 3 gives the base fee as the feerate times 724 weight units plus 172 for each untrimmed HTLC output, rounded down. Second, for each HTLC the closer offered, an HTLC-timeout transaction that BOLT 3 prices at the feerate times 663 weight units, taken out of the HTLC's own value. Third, a sweep of the delayed outputs into the closer's wallet. BOLT 5 recommends spending the delayed output: the node "SHOULD spend the to_local output to a convenient address", after waiting out the delay.

The weights for the three HTLC counts are below, followed by the fees.

Pending HTLCsCommitment weight (WU)Commitment vBSweep weight (WU)Sweep vB
0724181.00490122.50
51,584396.002,110527.50
305,8841,471.0010,2102,552.50

Commitment weight is 724 plus 172 per HTLC (BOLT 3). Sweep weight is 490 for the closer's own delayed output plus 324 for each timed-out HTLC output (derived; the 324 matches the BOLT 5 to_local input weight).

Pending HTLCsFee rateCommitment feeHTLC-timeout feesSweep feeClose totalOpen + close
01 sat/vB1810123304457
02 sat/vB3620245607913
05 sat/vB90506131,5182,282
020 sat/vB3,62002,4506,0709,125
050 sat/vB9,05006,12515,17522,813
0200 sat/vB36,200024,50060,70091,250
51 sat/vB3968255281,7491,902
52 sat/vB7921,6551,0553,5023,808
55 sat/vB1,9804,1402,6388,7589,522
520 sat/vB7,92016,57510,55035,04538,100
550 sat/vB19,80041,43526,37587,61095,248
5200 sat/vB79,200165,750105,500350,450381,000
301 sat/vB1,4714,9502,5538,9749,127
302 sat/vB2,9429,9305,10517,97718,283
305 sat/vB7,35524,84012,76344,95845,722
3020 sat/vB29,42099,45051,050179,920182,975
3050 sat/vB73,550248,610127,625449,785457,423
30200 sat/vB294,200994,500510,5001,799,2001,829,750

Fees in sats. Each pending HTLC is 50,000 sat, offered by the closer and timed out; the commitment was signed at the same rate as the close; the sweep is one batched transaction. The HTLC-timeout fees come out of the HTLC values rather than the funder's balance, so the Close total column adds them to show what it costs to get everything back.

With no HTLCs, the closer pays 3,620 sats for the commitment and 2,450 sats for the sweep at 20 sat/vB, a total of 6,070. A cooperative close at the same rate costs 3,380, so the force close is 1.8 times as expensive before any HTLC is involved. The gap comes from the commitment being a bigger transaction than a closing transaction and from the extra sweep step. At 1 sat/vB the same comparison is 304 sats against 169.

Pending HTLCs move the bill much more than the base transaction does. Each untrimmed HTLC adds 172 weight units to the commitment, which at 20 sat/vB is 860 sats. If the closer offered the HTLC and it has to be timed out, the HTLC-timeout transaction costs 3,315 sats and the sweep gains another 324-weight input, 1,620 sats. Together that is 5,795 sats per HTLC at 20 sat/vB. Thirty such HTLCs add 173,850 sats to the 6,070 sat base, which is why the 30-HTLC close at 20 sat/vB reaches 179,920 sats and the open-plus-close total reaches 182,975. At 50 sat/vB the 30-HTLC total is about 20 times the no-HTLC total.

Thirty simultaneous HTLCs is a stress case chosen to show the slope, which is steep in the fee rate and in the HTLC count and which the cooperative path never meets, because those HTLCs must clear first. BOLT 2 caps the number a peer may accept at 483 for channels that do not use zero-fee commitments, so the table sits well inside the allowed range.

If the HTLCs were instead received by the closer and the closer holds the preimages, the second-stage transaction is HTLC-success at 703 weight units, 40 more than HTLC-timeout, and it has a locktime of 0 in BOLT 3, so it does not wait for the HTLC to expire.

The legacy format has a second assumption baked in that deserves an explicit warning. BOLT 2 says: "For the legacy commitment format, there is no general method for the broadcasting node to use child-pays-for-parent to increase its effective fee." If the rate in the last signed commitment is below what miners want when the force close happens, the node cannot raise it unilaterally. BOLT 2 suggests the fee payer "keep a good margin (say 5x the expected fee requirement) for legacy commitment txes; but, due to differing methods of fee estimation, an exact value is not specified." The legacy rows are therefore the case where the signed rate happened to match the market.

Anchor Channels: A Bigger Commitment and a Child Transaction

The anchor format answers the stuck-commitment problem by adding two small outputs, one for each side, that either party can spend to attach a child transaction. BOLT 3 gives the amount: "The amount of the output is fixed at 330 sats, the default dust limit for P2WSH." It also says when they appear: "Both anchor outputs are always added, except for the case where there are no htlcs and one of the parties has a commitment output that is below the dust limit. In that case only an anchor is added for the commitment output that does materialize. This typically happens if the initiator closes right after opening (no to_remote output)." The exception means a force close on a brand-new channel carries a smaller commitment than the 1,124-unit formula.

The formula weight rises by 400 weight units, 100 vB, from the extra outputs. At 20 sat/vB that is 2,000 sats of extra cost on the commitment before any child is attached (5,620 against 3,620), and the two 330 sat anchor values come out of the funder's balance, 660 sats in total. The funder can reclaim its own anchor by spending it; BOLT 3 says "any anchor that remains unspent can be spent by anyone after the commitment tx confirms", so the other anchor is not the funder's to count on. The trade is that the commitment's own fee rate stops being the thing that decides when it confirms.

The HTLC transactions change in the same format. BOLT 3 sets their fee to nothing: "If option_anchors or zero_fee_commitments applies: MUST be 0. Otherwise, MUST be calculated to match: Multiply feerate_per_kw by 663 and divide by 1000 (rounding down)." The fee for those transactions then has to come from additional inputs. The anchor table prices each HTLC-timeout transaction at its 666 weight units and the target rate, which is a lower bound because the added input and change output are not counted. The HTLC amounts also no longer shrink by the second-stage fee when the spec decides what to trim, a point the trim section returns to.

To price the bump, the table assumes the commitment was signed at 1 sat/vB, the lowest rate in this article, and that the closer attaches a child at the target rate. The child spends one anchor and one wallet coin and returns change: 719 weight units, or 179.75 vB, derived from the anchor script in BOLT 3 (a public key, a signature check and a 16-block fallback) and the same P2WPKH sizes used above. A miner judges the pair together, so the package fee at the target rate is the combined weight of commitment and child times the rate, and the small fee the commitment already carries counts toward it.

Pending HTLCsFee rateCommitment fee if signed at that rateCommitment plus child, from a 1 sat/vB commitmentHTLC-timeout fees (lower bound)Sweep feeOpen + close, with child (lower bound at 5 and 30 HTLCs)
01 sat/vB2814610123737
02 sat/vB56292202451,473
05 sat/vB1,4052,30406133,681
020 sat/vB5,6209,21502,45014,720
050 sat/vB14,05023,03806,12536,801
0200 sat/vB56,20092,150024,500147,200
51 sat/vB4966768305282,187
52 sat/vB9921,3521,6651,0554,378
55 sat/vB2,4803,3794,1602,63810,941
520 sat/vB9,92013,51516,65010,55043,770
550 sat/vB24,80033,78841,62526,375109,426
5200 sat/vB99,200135,150166,500105,500437,700
301 sat/vB1,5711,7514,9802,5539,437
302 sat/vB3,1423,5029,9905,10518,903
305 sat/vB7,8558,75424,96012,76347,241
3020 sat/vB31,42035,01599,90051,050189,020
3050 sat/vB78,55087,538249,750127,625472,551
30200 sat/vB314,200350,150999,000510,5001,890,200

Fees in sats. The third column assumes the anchor commitment was signed at the same rate as the close, so no child is needed. The fourth shows the package fee for commitment plus child when the commitment was signed at 1 sat/vB, and the last column uses it. The HTLC-timeout column prices each 666-weight transaction at the target rate and leaves out the wallet input and change output that must be added to pay it, so the 5 and 30 HTLC totals are lower bounds. The 660 sat anchor values are also excluded. The sweep covers the closer's own output and every timed-out HTLC output.

At 20 sat/vB with no HTLCs, the bumped route costs 9,215 sats for the package plus 2,450 for the sweep, 11,665 sats for the close, against 6,070 for the legacy row. That is not an argument against anchors. The legacy row only reaches its figure if the commitment was signed at the right rate. The anchor row pays for the ability to pick the rate on the day of the close. With 30 HTLCs at 20 sat/vB the anchor total of 189,020 sats is within 3.3 percent of the legacy total of 182,975, because the 30 timed-out HTLC transactions and the 31 sweep inputs dominate both totals. The 5,595 sat gap between the legacy commitment fee (3,620) and the anchor package (9,215) at 20 sat/vB is the 400 extra weight units of the anchor commitment plus the 719 of the child, 1,119 weight units in all.

BOLT 5 makes the child a requirement rather than an option: the broadcaster "MUST spend any to_local_anchor or shared_anchor output, providing sufficient fees as incentive to include the commitment transaction in a block." It also adds that the node "SHOULD use replace-by-fee or other mechanism on the spending transaction if it proves insufficient for timely inclusion in a block", so a first child that is too cheap can be replaced with a higher one.

Zero-Fee Commitments in the Current Specification

The BOLT 3 text on the main branch changed on May 4, 2026, when a commit titled "Zero-fee commitments using v3 transactions (feature 40/41)" was merged. It adds a channel type, zero_fee_commitments, in which the commitment transaction pays no mining fee at all. BOLT 3 states it directly: "Note that feerate_per_kw is set to 0 when using zero_fee_commitments: in that case the commitment transactions and HTLC transactions don't pay any fee, which must be paid using child-pays-for-parent (for the commitment transactions) or by adding inputs (for the HTLC transactions)."

The commitment uses a single shared anchor that anyone can spend. BOLT 3 describes its amount: "If the difference between the commitment transaction input and the sum of all other commitment transaction outputs is smaller than 240 sat, we set the anchor amount to this value. This ensures that the commitment transaction doesn't pay any on-chain fees whenever the anchor amount is smaller than 240 sat. Otherwise, we set the anchor amount to 240 sat, and the remaining difference between the commitment transaction input and its outputs directly contributes to the transaction's mining fees." BOLT 2 also cuts the pending-HTLC limit for this channel type to 114, explaining that "v3 transactions are limited to 10kvB", which reduces how many outputs a commitment can have.

For cost purposes, the commitment itself pays nothing, so the whole mining fee falls to a child transaction. BOLT 3 Appendix A gives no expected-weight constant for the zero-fee commitment, so no cost row is shown for it.

Which HTLCs Survive: Trim Thresholds by Fee Rate

A pending HTLC only appears on the commitment if it is worth enough. BOLT 3 describes the rule for the rest: "A trimmed output is considered too small to be worth creating: it is instead either added to the commitment transaction fee or, if zero_fee_commitments applies, to the shared anchor output." For each offered HTLC, the output is omitted "if the HTLC amount minus the HTLC-timeout fee would be less than dust_limit_satoshis set by the transaction owner"; for each received HTLC the HTLC-success fee is subtracted instead.

The threshold therefore moves with the fee rate in channels without anchors. For an HTLC-timeout at feerate_per_kw the fee is 663 times the rate divided by 1,000, and the smallest HTLC that stays on the commitment is that fee plus the dust limit.

Fee rateSat/kwHTLC-timeout feeSmallest offered HTLC keptHTLC-success feeSmallest received HTLC keptSmallest HTLC kept with anchorsTimeout fee as share of a 50,000 sat HTLC
1 sat/vB2501655191755293540.3%
2 sat/vB5003316853517053540.7%
5 sat/vB1,2508281,1828781,2323541.7%
20 sat/vB5,0003,3153,6693,5153,8693546.6%
50 sat/vB12,5008,2878,6418,7879,14135416.6%
200 sat/vB50,00033,15033,50435,15035,50435466.3%

Legacy thresholds use a 354 sat dust limit; the 546 sat limit in the BOLT 3 worked example would add 192 sat to each threshold. The last column is the HTLC-timeout fee divided by a 50,000 sat HTLC.

At 20 sat/vB a legacy channel drops every offered HTLC below 3,669 sats and every received HTLC below 3,869 sats from the commitment. At 200 sat/vB those cutoffs reach 33,504 and 35,504. The money in the trimmed HTLCs is not lost to an attacker or to the counterparty. BOLT 3's worked example shows where it goes: at 5,000 sat/kw, two of the four HTLCs fall below the cutoff. The base commitment fee is 5,340 sats and "the actual fee (which adds the 1000 and 800 satoshi HTLCs that would make dust outputs) is 7140 satoshi." The trimmed amounts are added to the commitment fee, so the miner collects them, and the real fee ends up well above the formula's 5,340 sats.

In channels with anchors the second-stage transactions pay no fee, so the cutoff is the dust limit alone. BOLT 3 flags the consequence: "since the introduction of option_anchors, the second-stage HTLC transaction's weight is not taken into account when deciding whether outputs should be included in the commitment transaction or not. It thus makes sense to use a dust_limit_satoshis that takes into account the cost of those second-stage HTLC transactions, to ensure that outputs added to the commitment transaction can actually be claimed on-chain, otherwise they may pollute the utxo set indefinitely." An output can therefore sit on an anchor commitment that costs more to claim than it holds.

The last column of the table shows how this plays out when fees are high. At 200 sat/vB the HTLC-timeout fee alone is 66.3 percent of a 50,000 sat HTLC. A force close at that rate is a bad moment to be holding many mid-sized HTLCs on a legacy channel.

The Smallest Channel for a 5 or 10 Percent Fee Ceiling

The question behind the tables is usually whether a channel is big enough for its fees not to matter. If open plus close costs C sats and you want that to stay at or under a share p of the channel, the channel capacity S must satisfy S ≥ C / p. At 5 percent that means at least 20 times the cost; at 10 percent, at least 10 times. The first table applies it to the cooperative round trip.

Fee rateOpen + cooperative closeSmallest channel for 5%Smallest channel for 10%
1 sat/vB3226,4403,220
2 sat/vB64412,8806,440
5 sat/vB1,60932,18016,090
20 sat/vB6,435128,70064,350
50 sat/vB16,088321,760160,880
200 sat/vB64,3501,287,000643,500

Channel size in sats. The smallest size such that open plus cooperative close, from the table above, is at or under 5% or 10% of the channel, rounded up to a whole satoshi.

At 20 sat/vB the round trip costs 6,435 sats, so the smallest channel for 5 percent is 128,700 sats and for 10 percent is 64,350. At the 1 sat/vB the endpoint recommended, the same thresholds are 6,440 and 3,220 sats. At 200 sat/vB, 5 percent needs 1,287,000 sats and 10 percent needs 643,500. The minimums scale with the fee rate almost exactly, which means that a rule of thumb stated in sats ages as quickly as the fee market moves.

The force-close version is stricter because the close itself is bigger. With no HTLCs, the legacy force close plus funding costs 9,125 sats at 20 sat/vB, and the anchor route with a child costs 14,720 sats.

Fee rateForce close, no anchors, 0 HTLCsFor 5%For 10%Force close, anchors with child, 0 HTLCsFor 5%For 10%
1 sat/vB4579,1404,57073714,7407,370
2 sat/vB91318,2609,1301,47329,46014,730
5 sat/vB2,28245,64022,8203,68173,62036,810
20 sat/vB9,125182,50091,25014,720294,400147,200
50 sat/vB22,813456,260228,13036,801736,020368,010
200 sat/vB91,2501,825,000912,500147,2002,944,0001,472,000

Same calculation applied to the force-close totals with no pending HTLCs. The anchor columns use the bumped commitment from a 1 sat/vB signature; the legacy columns assume the commitment was signed at the shown rate.

A worked case: a 1,000,000 sat channel at 20 sat/vB. Opening costs 3,055 sats and a cooperative close 3,380, 6,435 in all, which is 0.64 percent of the channel. A legacy force close with no HTLCs brings the total to 9,125 sats, 0.91 percent. Both are well under 5 percent because the channel is about 7.8 times the 5 percent minimum. The same channel at 200 sat/vB would pay 64,350 sats for the cooperative round trip, which is 6.44 percent and already above the 5 percent line.

Percent of capacity is not the only constraint on how small a channel can be, and two limits come straight from the specification. First, BOLT 2 says "1% of the channel total is suggested" as the reserve each side keeps, so a channel's spendable balance is smaller than its capacity. Second, the opener must be able to afford the commitment fee at open. BOLT 2 requires the receiving node to fail the channel if "the funder's amount for the initial commitment transaction is not sufficient for full fee payment". At 200 sat/vB the legacy commitment fee with no HTLCs is 36,200 sats, so a funder cannot open a legacy channel smaller than that at that rate, whatever the percentage rule says. BOLT 2 adds that the fee payer should keep a "fee spike buffer" on top of its reserve, so the practical minimum is higher than the arithmetic.

The Delay and the Penalty: Why a Force Close Is Not Just Pricier

Money is only one cost of a force close. The other is time and exposure, and both come from the same rule that sets the price.

The broadcaster's own balance sits behind a relative timelock. BOLT 3 describes the reason: the local node "must also wait to_self_delay before accessing these funds, allowing for the remote node to claim these funds if the transaction has been revoked." The delay is negotiated per channel, so the table cannot give a universal figure; the Appendix C test vectors use 144 blocks, which is a test value rather than a recommendation. A cooperative close has no such delay, which is the first half of the BOLT 5 sentence quoted earlier.

The second half is the penalty. If a node broadcasts a commitment that has already been replaced, the commitment is revoked and the penalty path opens. That is why BOLT 5 says a node that "knows or assumes its channel state is outdated" must not broadcast its last commitment transaction. BOLT 3 describes what a revoked broadcast exposes: the other party "can spend this output immediately" using the revocation signature. What is at stake is the publisher's balance, not a few thousand sats of fees.

An HTLC on the commitment adds a third wait. An offered HTLC cannot be timed out until its expiry height, and BOLT 3 notes that the HTLC-timeout transaction "has cltv_expiry as its specified locktime". Then the output of that second-stage transaction has the same delay as the first. A force close with pending HTLCs therefore settles in stages, and the fee at each stage is paid at that stage's rate, not the rate when the channel was closed. The tables use one rate for all stages. Real closes may pay different rates for the commitment, the second-stage transactions and the sweep.

Adjusting the Tables for Your Own Transaction

Because the tables come from additive weights, they can be adapted without rerunning anything. The building blocks, in weight units:

Each extra P2WPKH input in the funding transaction adds 273 weight units, 68.25 vB.
Each extra P2WPKH output adds 124 weight units, 31 vB. Each extra output with a 43-byte script (P2WSH or P2TR) adds 172 weight units, 43 vB.
Each pending HTLC on a commitment adds 172 weight units, 43 vB.
Each delayed input added to a sweep adds 324 weight units, 81 vB.
Moving from a no-anchor to an anchor commitment adds 400 weight units, 100 vB.

Multiply the vB by the fee rate in sat/vB to get sats. A funding transaction with two P2WPKH inputs and no change output, for example, is set B less the change output: 884 minus 124, or 760 weight units.

The calculation does not need the price of bitcoin, and neither does the percentage rule. It does need a fee rate. The sat/vB figure on a given day is available from any public mempool explorer, and our guide to reading a transaction in a block explorer shows how to check the fee a transaction actually paid after the fact. For the background on how blocks and transactions fit together, see how a blockchain works, and for where Lightning sits among scaling approaches, Layer 1 versus Layer 2. The Learn hub lists both.

Costs Outside These Tables

Several costs and variations sit outside the tables, and each could change a real bill.

Splicing, which lets a channel add or remove funds without a separate close and open, changes the on-chain footprint of resizing a channel. BOLT 2 refers to splicing in places; neither its transactions nor its fees are priced here. BOLT 3 defines a P2WSH funding output, so any other funding script is outside these figures.

Liquidity purchases, channel-opening fees charged by a service, and routing fees paid on payments are not on-chain fees and are not included. A 5 percent threshold computed here refers only to on-chain fees measured against the funded capacity.

The tables use one fee rate for every transaction in a scenario. Fee rates move between the open and the close, and in a force close they can differ between the commitment, the second-stage transactions and the sweep. A channel opened at 1 sat/vB and closed cooperatively at 50 sat/vB costs 153 plus 8,450, which is 8,603 sats, a figure no single row of the tables shows.

The zero-fee commitment channel type is described only from the specification text, with no cost row, and the tables say nothing about which node software supports it. The HTLC scenarios choose one HTLC size and one direction. The specification defines the weights per output, but a real channel's mix of offered and received HTLCs, trimmed and untrimmed, and timed-out and fulfilled, decides which second-stage transactions are broadcast. The anchor rows leave out the fee inputs the second-stage transactions need.

Finally, the weights are the specification's expected weights, which assume maximum-length signatures. A real transaction may be a few weight units lighter, and a wallet may add its own safety margin. Neither changes the structure of the result.

Frequently Asked Questions

How much does it cost to open a Lightning channel?

The opener pays one on-chain funding transaction. With a single P2WPKH input and a change output, it is 611 weight units, 152.75 vB, so 153 sats at 1 sat/vB, 3,055 sats at 20 sat/vB and 30,550 sats at 200 sat/vB. A coin set with more inputs or a batched transaction changes the figure, as the funding table shows.

Why does a force close cost more than a cooperative close?

A force close publishes the full commitment transaction, which is larger than a closing transaction, then needs a sweep after the delay, and adds second-stage transactions for any HTLCs. With no HTLCs at 20 sat/vB it costs 6,070 sats against 3,380 for a cooperative close, and each pending HTLC adds about 5,795 sats in the modelled worst case.

Can I close a channel cooperatively while payments are pending?

No. BOLT 2 says that once shutdown is complete the channel is empty of HTLCs, and a node must not send a new HTLC after sending shutdown. Pending HTLCs have to settle or fail before the closing transaction can be built. If they cannot, the remaining route is a force close.

Does the other side pay anything to open a channel?

Not in a single-funded channel. BOLT 2 describes the fee as one-sided, with the party that requested the channel paying the commitment fees, and the funding transaction is the opener's. In a dual-funded channel, BOLT 3 says each party pays for its part and the initiator covers the common transaction fields.

What channel size keeps open and close fees at or under 5 percent?

Divide the round-trip cost by 0.05. At 20 sat/vB, a cooperative round trip costs 6,435 sats, so the channel must be at least 128,700 sats. At 1 sat/vB the figure is 6,440 sats. For 10 percent, halve those numbers. A force close needs a larger channel for the same share.

Do anchor channels cost more to close?

The commitment weighs 1,124 units against 724, which is 100 vB more. If the commitment must be bumped by a child transaction, the package also includes that child's 719 weight units. In return the rate is chosen on the day of the close instead of being fixed in the last signed commitment.

Is the 1 sat/vB figure on this page what a wallet will charge today?

No. It is one reading from one source at 17:57 UTC on November 3, 2026. A wallet applies its own fee estimate when the transaction is built, may use different inputs, and rounds in its own way. Use the weights and the rule for adjusting them to price a specific transaction at the rate your wallet shows.

This page is general educational research from CryptosEyes. It is not investment, tax or legal advice, and it recommends no wallet, node software, service or channel size. Weights and rules are quoted from the Lightning specification at the commits listed in the sources; implementations may differ, and fee rates change.

Sources

Primary documents, market context, and methodology references behind this article.

Reviewed against source notes and calculations
Education
Research note: This article is educational market research, not financial advice. Crypto and public equity data can change quickly; see our methodology and editorial policy for sourcing, review, and correction standards.
Important: Educational Purposes OnlyThe data, charts, treasury tracking metrics (including mNAV and SPS), and research provided on CryptosEyes.com are for informational and educational purposes only. They do not constitute certified financial, investment, or trading advice. Digital assets like Bitcoin and Ethereum are highly volatile. Always conduct your own research and consult with a registered financial advisor before making investment decisions.