I work in the crypto self-custody space, and Bitcoin's security budget has a scheduled expiry date.
That is the uncomfortable premise behind a talk Peter Todd recorded at bitcoin++ in Toronto on 23 July 2026, published on 14 August 2026 (bitcoin++ video). His proposal is that Bitcoin should keep paying miners forever, and that if the network will not accept new coins to do it, it should tax the existing ones instead.
Within a day, Adam Back, whose Hashcash the Bitcoin whitepaper names as the model for its proof of work (whitepaper), was in the thread arguing against it. Below is what Todd argued, what the numbers support, and where the objections land.
The claim that proof of work was never about making coins
Todd opens with a concession about his own past. He says that back when he was, in his words, trying to be Satoshi and invent Bitcoin, he completely screwed this up, and that Adam Back did not get it either. Both of them thought of proof of work primarily as the mechanism that creates coins.
His current position is that issuance is a byproduct. Proof of work exists to produce consensus, to make the network agree on which block comes next. Paying miners in new coins is one way to fund that, not the point of it.
Everything that follows depends on this reframing. If the subsidy is just a funding mechanism, then removing it raises an engineering question rather than a moral one. If the 21 million cap is the product itself, the question never gets asked.
Where the security budget stands
Fees are a rounding error in miner revenue. On a 30-day average they were about 0.69 percent of it in August 2026, according to a Glassnode chart that Rafael Schultze-Kraft, a Glassnode co-founder (Glassnode), posted on 11 August, noting that Bitcoin was below 400 dollars the last time the fee share was this low (Schultze-Kraft on X). The same chart marks December 2015 at 0.69 percent with Bitcoin at 388 dollars, and an April 2026 low of 0.52 percent.
Replacing the subsidy with fees would take far more than blocks pay today. The 3.125 BTC subsidy was worth about 203,000 dollars at the 65,000 dollar price on that chart (Glassnode chart). The 150 blocks mined up to block 962,005 on 11 August 2026 carried about 4,700 transactions and 0.022 BTC in fees each, roughly 1,400 dollars per block (mempool.space). Covering the subsidy at that volume would take about 43 dollars from every transaction.
Todd's read on the fee spikes is that they do not last. In his view inscriptions were a flash in the pan, and once the inscription community moved from JPEGs to much smaller payloads, the fees went with it. The spikes were real while they lasted: on 20 April 2024, when the Runes protocol activated at block 840,000 (Runes specification), the 130 blocks of that day collected 1,258 BTC in fees across about 631,000 transactions, an average near 127 dollars per transaction at the day's 63,730 dollar price (mempool.space).
The subsidy falls to 1.5625 BTC at block 1,050,000, expected around April 2028, and reaches zero after block 6,929,999, around 2140 (Bitcoin Core). The trajectory is the issue, not the endpoint.
Why a fixed tail emission converges
The counterintuitive part of Todd's case is that a permanent block reward does not mean permanent inflation.
People lose coins. They die without passing on keys, drives fail, passphrases are forgotten. The rate of loss is proportional to the supply rather than a fixed quantity, because losing coins is a probability applied to holdings, not a headcount. Todd's framing is that whether you hold one coin or ten does not change your odds of a boating accident.
Set a fixed number of new coins per block against a proportional loss rate and the supply converges. In his 2022 essay he writes it as a differential equation, dN/dt = k - λN, which settles at k/λ, the point where coins are created exactly as fast as they disappear (Todd, 2022).
In the talk he skips the calculus for a leaky bucket. Water pours in at a constant rate, and the hole at the bottom drains faster as the water rises. Below equilibrium the level climbs, above it the level falls, and it ends up in the same place either way.
Monero is the working example: its tail emission began at the end of May 2022 at 0.6 XMR per block (Moneropedia), which Todd's essay puts at a 0.9 percent apparent inflation rate (Todd, 2022). Todd is careful not to endorse the reason Monero needed it, which was its variable block size design, and he notes there are scenarios where that design fails on consensus grounds.
The number Todd has not pinned down
In my view the convergence argument is sound math with a missing input.
Convergence tells you a stable point exists. It does not tell you where. That depends entirely on the real loss rate, which nobody can measure. Todd concedes this directly, comparing it to dropping a golf ball: he does not need to know the height to know it lands.
His own numbers have moved. The 2022 essay calls 1 percent per year "likely to be a good guess, quite possibly less" (Todd, 2022). In the Toronto talk he suggests something like 0.1 percent (bitcoin++ video). That is an order of magnitude of spread on the one input that determines whether the policy is trivial or expensive.
He supplies the cost framing himself, over a 50 year saving life. In the talk he says 1 percent compounded over 50 years "is like 64%" and 0.1 percent "is like 5%" (bitcoin++ video). The 64 is the growth figure: 1.01 to the power of 50 is 1.64. For a holder paying 1 percent a year, 0.99 to the power of 50 leaves 60.5 percent, so the loss is about 40 percent of the holding. At 0.1 percent the loss is about 5 percent either way. His argument is that paying 5 percent of your savings to keep those savings secure is a reasonable trade, and that Bitcoin's price swings dwarf the difference anyway.
His critics are arguing about whether 40 percent is reasonable too.
Demurrage as the soft fork workaround
Todd does not expect tail emission to happen. He asked the Toronto room to raise hands for anyone who thought a hard fork was likely within five years, and concluded that even with agreement on tail emission, nobody would hard fork for it.
So he proposes reaching the same place without minting anything. Inflation, he argues, is a tax on savings, redistributing purchasing power from holders to whoever receives the new units. If that is what tail emission does economically, then a direct tax on holding coins does the same job without touching the supply cap.
The mechanism works on coin-days destroyed. For each block, sum the value of the transaction outputs being spent, multiply by how long each has existed, and apply a rate. That product becomes a tax the block owes. A new consensus rule requires the coinbase transaction to assign that amount to a fund, and miners may withdraw only a capped percentage of that fund per block, which spreads the payout over time instead of handing it to whoever mines next.
No coins are created. The rule only constrains how existing value moves. That, in Todd's framing, makes it a soft fork.
The case against
The first objection after the video went up was political. On 14 August a user posting as ts_hodl argued that a fork to change the supply schedule would fail as hard as BIP-110, if not harder (ts_hodl on X). BIP-110 is a temporary data-limiting soft fork (BIP-110) whose best two-week signaling period reached 51 of 2,016 blocks, 2.53 percent, against a 55 percent activation threshold (BIP-110 Monitor). Adam Back replied the next day that the trick for such causes is to "trigger and rally people to your dangerously inadvisable cause with simple though false narrarives" (sic), and named the two narratives he says BIP-110 used (Back on X).
The sound money objection is the oldest and the least technical. It predates the Toronto talk, which Todd opened by saying it repeats the one he gave at the Hoover Dam edition of bitcoin++ in April 2026, listed there as "The Case for Tail Emissions" (bitcoin++ archive). On 25 November 2025 the account Beautyon wrote that Bitcoin's main feature is sound money, which "means no tail emission, no demurrage and no increasing the number of coins" (Beautyon on X).
That position treats the fixed supply as the product rather than a parameter of it. On that reading, a network that taxes holdings to pay for its own security has already lost the argument it was built to win, whatever the security budget looks like afterwards.
Todd answered the next day with a screenshot of that post (Todd on X).
In my view there is also a narrower technical objection to the soft fork label. A rule that reduces what existing holders can spend changes the economic position of every coin, whether or not the supply cap moves. Calling that a soft fork is accurate about the code and contested about the substance.
The pushback Todd accepted
For me the most useful part of the talk is the Q&A, because Todd concedes more there than in the presentation.
Asked to justify the convergence claim, he defends the direction but not the destination, and admits he does not know the numbers.
Asked whether an inflation tax and a demurrage tax are equivalent, he agrees there are cases where they are not. His use of equivalent means approximately, and he offers a hypothetical 50 percent demurrage rate to show where the equivalence breaks. He goes further and allows that you can make an argument for demurrage being the better solution because its rate can be set.
He also cuts off an argument he does not want. Demurrage has a long history in economics as a way to force spending and discourage hoarding. Todd says that argument is not relevant to his case, describes himself as relatively speaking an Austrian economist, and frames demurrage strictly as a funding mechanism for a public good.
What the fight is about
Set the monetary politics aside and a narrower argument remains. It gets the least attention.
The reorg concern is not that Bitcoin dies. Todd allows that a fee-only chain with occasional reorganisations might be survivable if people wait for more confirmations. The cost he cares about is who gets to mine.
Reorganising the chain to capture a fee-rich block is only rational if you have enough hash power to expect consecutive blocks. A miner in a garage has effectively zero chance and treats any block as a windfall. A miner at 49 percent has real opportunities. Todd argues that cooperative structures like P2Pool cannot coordinate a reorg at all, which means that in a high fee-variance world they would be at a disadvantage against pools that can.
Fee variance is not hypothetical either. Block discovery is a Poisson process, so long gaps between blocks are normal, and each long gap stacks up transactions bidding against each other. Todd cites Mara's Slipstream as evidence miners will build machinery to capture marginal revenue, while noting he does not consider Slipstream itself a significant threat.
The academic footing predates all of this. Carlsten, Kalodner, Weinberg and Narayanan published the fee-only instability result at ACM CCS 2016, showing that the variance of fee-based rewards makes forking a wealthy block attractive and produces an equilibrium with undesirable properties, and in some circumstances no equilibrium at all (Carlsten et al.).
In my view Andrew Poelstra's line, which Todd quotes, holds up better than either side's rhetoric. The subsidy going down is a scary phase change. Bitcoin has a system that seems to work, and it keeps marching towards a type of system nobody has examples of working.
Todd's answer is to pay for consensus permanently. His critics answer that a Bitcoin which taxes its holders to buy security is no longer the thing worth securing. Neither side can produce the evidence that would settle it, because the experiment has not run yet.
The part I would not wave away is the decentralisation argument. In my view the 21 million debate will get decided politically, probably by nobody doing anything. Whether fee variance pushes mining toward the operators large enough to reorg is a question with a technical answer, and I expect it to arrive long before 2140.
