The halving is a height, not a date

The Bitcoin halving is triggered by block height 1,050,000 and by nothing else. At the 2 October 2026 reading the chain stood at height 969,538, so 80,462 blocks remain to be found. Divide by the 144 blocks a day the protocol targets and the answer is 558.76 days, which is where the figure of 559 days used everywhere on this page comes from.

The projected calendar date, 12 April 2028, is the least reliable number on this page and it is printed here only because readers ask for it. Difficulty retargets every 2,016 blocks to pull the average interval back towards ten minutes, but it retargets after the fact, so a network that is gaining hashrate runs a persistently short block and arrives early. The table below is the whole of the uncertainty: it takes 80,462 blocks as given and varies only the average interval.

A ten-second difference in the average block interval moves the halving by roughly 10 days. That is the correct order of magnitude for the error bar on any halving countdown, including this one. What does not move is the height, which is why every projection on this site steps the subsidy at a block count rather than at a date.

At block 1,050,000 the subsidy drops from 3.125 BTC to 1.5625 BTC. Every projection on this site runs that step through subsidyAt(t), which walks the schedule forward in 1458-day intervals rather than holding a constant, so a four-year model contains one halving and a ten-year model contains two.

The second half of the reward is transaction fees, carried here as 0.71% of the subsidy, derived from the 0.0223 BTC 30-day average fee per block. Expressing fees as a share of subsidy means the fee term halves alongside the subsidy, which is a modelling choice worth naming rather than burying. The choice is defensible in 2026 for one reason: fees are negligible. Cointelegraph, on-chain data reported on 12 August 2026 that fees had fallen to 0.69% of miner revenue, a ten-year low. Whether that fraction of a percent halves or holds is not a decision a fleet turns on.

Because the reward halves exactly, the arithmetic downstream is unusually clean. Coins per day for a given hashrate share halve. Revenue at a fixed bitcoin price halves. The electricity bill does not move by a single cent, because a machine draws the same watts whether the subsidy is 3.125 BTC or 1.5625. So the price at which revenue meets the power bill doubles, and it doubles for every machine on the network at the same instant.

The shutdown price of the default fleet, fifty Antminer S21 Pro units drawing 193.1 kW at the wall at $0.05 per kWh, is $41,949 on 2 October 2026 and $83,898 after the halving. The ratio is 2.000, and it is that clean because the same subsidy term appears in the numerator of revenue and nowhere in the cost. Spot is $85,941, so today the fleet clears its power bill with 33% of gross to spare, and after the halving it would need bitcoin 33.1% above today's price simply to break even on electricity.

The shutdown price is the only hard floor in mining. Purchase price, financing and depreciation are already spent and cannot be recovered by continuing to run, so the question a fleet faces every morning is narrower than profitability: does today's revenue exceed today's power bill. Below the shutdown price the answer is no and the correct action is to switch off, which is why hashrate falls in a drawdown rather than grinding on at a loss.

Difficulty makes the picture worse rather than better. Holding difficulty at today's level is generous to the post-halving case, because the network has to lose hashrate for difficulty to fall and that only happens after miners have already been losing money. At this model's default drift of 0.2% per retarget, difficulty on day 559 is 1.090 times today's, and the fleet's shutdown price is $91,487 rather than $83,898.

Of the 27 machines in this catalogue, 17 clear their own power bill at the 15 August 2026 spot price of $85,941, at $0.05 per kWh with PUE 1.10 and 96% uptime. After the halving, with the price and the difficulty both held still so that the halving is the only change, 3 do. 14 machines that are cash positive today are cash negative the day after block 1,050,000.

The 3 survivors are Antminer S23 Hyd, SealMiner A4 Pro Air, Antminer S23, all of them 2026 hardware at 9.5 to 11.0 J/TH. Every 2024 and 2025 machine in the catalogue falls below its power cost, including the Antminer S21 Pro that this site uses as its default fleet and that vendors are still selling new at $1,938 a unit. A machine bought today at $8.28 per terahash therefore has 559 days of positive cash flow ahead of it at this power price and this bitcoin price, and then needs one of three things to change: a higher bitcoin price, a lower power price, or enough of the network switching off to bring difficulty down.

Running the default fleet forward 1,460 days with the subsidy stepped down at day 559 produces 5.307 BTC and $334,613 of gross revenue. Running the same fleet on the same difficulty path with the reward frozen at 3.125 BTC produces 7.354 BTC and $463,687. The difference is 38.6%, or $129,074 of revenue that the second run invents. Both runs use identical hardware, identical uptime, identical power price and identical difficulty drift of 0.2% per retarget. The subsidy schedule is the only input that differs.

The overstatement is not evenly spread and that is what makes it dangerous. For the first 559 days the two runs are the same number to the last decimal, because before the halving there is nothing to disagree about. All of the error lands in the final 853 days, which is the stretch a buyer is relying on to finish paying for the machines.

The cash result changes sign rather than degree. With the halving applied, four years of this fleet at $85,941 and $0.05 per kWh nets −$6,956.60 against $338,224 of electricity, which is a loss before the $96,900 of hardware is considered at all. With the reward frozen, the same four years net $120,826.38 and the fleet looks like a business. A swing of $127,783 on a $96,900 purchase is not a modelling nuance; it is the difference between a payback table and a write-off, and it is produced by one line of the model.

Day 559 is 1.66 years from the 2 October 2026 reading. A miner bought new today and written off straight line over three years still carries $43,245 of book value out of $96,900 on the day the subsidy halves, meaning 45% of the purchase is still waiting to be earned back when the revenue side is cut in half. This site's network production-cost model assumes a 4.0-year hardware life, on which the halving lands 42% of the way through.

Resale does not rescue the position, because every other owner faces the same block height on the same day. The three-method residual model on this site puts the default fleet's realisable value at $30,511 on day 559, against $96,900 paid, after the 15% liquidation haircut. That estimate is a model output rather than an observed price, and it already assumes a buyer exists. The honest version of the sentence is that the machines are worth what a buyer can earn with them, and on day 559 every buyer is looking at the same halved subsidy.

The practical consequence is a question about ordering rather than about forecasting. A machine bought in August 2026 gets roughly 559 days at the current subsidy and then several years at half of it, so the purchase has to be justified by the second regime and not the first. Any payback figure quoted in months that does not name a halving date is describing the first 559 days and quietly assuming the rest.

When is the next Bitcoin halving?

The next Bitcoin halving happens at block height 1,050,000, not on a calendar date. As of the network reading taken on 2 October 2026 at height 969,538, 80,462 blocks remain. At a perfect ten-minute block that is 559 days, which projects to 12 April 2028. The projection is arithmetic, not a schedule: blocks have been arriving slightly faster or slower than ten minutes for the whole history of the network, and every tenth of a minute of average interval moves the projected date by about 6 days.

What does the halving do to Bitcoin mining profitability?

The halving cuts the block subsidy from 3.125 BTC to 1.5625 BTC and leaves the electricity bill exactly where it was, so the bitcoin price at which a machine stops covering its own power doubles. For the default fleet on this site, fifty Antminer S21 Pro units at $0.05 per kWh, the shutdown price moves from $41,949 to $83,898. At the 2 October 2026 spot price of $85,941 that fleet earns $116.53 a day today and would lose $57.57 a day on the same difficulty after the halving.

How many mining machines stop being profitable at the halving?

Of the 27 machines in this catalogue, 17 cover their own electricity at the 2 October 2026 spot price of $85,941, at $0.05 per kWh with PUE 1.10. After the halving, holding both the price and the difficulty still, 3 do: Antminer S23 Hyd, SealMiner A4 Pro Air, Antminer S23. That is 14 machines that pay for their power today and would not pay for it the day after block 1,050,000. All 3 survivors are 2026 hardware under 11.0 J/TH.

How much does ignoring the halving overstate a four-year mining projection?

By 38.6% of the bitcoin mined, on this model, over 1,460 days from 2 October 2026. Stepping the subsidy down at day 559 yields 5.307 BTC and $334,613 of gross revenue for the default fleet. Freezing the reward at 3.125 BTC on the same difficulty path yields 7.354 BTC and $463,687, an overstatement of $129,074. The sign of the answer changes, not just its size: with the halving applied the four-year cumulative net is −$6,956.60, and with the reward frozen it is $120,826.38.

Does the halving cut transaction fee income as well?

No, but on this model fee income is too small to matter either way. Fees are carried here as a percentage of the subsidy, currently 0.71%, drawn from the 0.0223 BTC 30-day average fee per block. Cointelegraph reported on 12 August 2026 that fees had fallen to 0.69% of miner revenue, a ten-year low. Because fees are modelled as a share of the subsidy, the fee term halves with it, and the total reward per block therefore halves exactly. If fees instead held their absolute level, the post-halving reward would be a fraction of a percent higher than half. That correction is smaller than the rounding on the difficulty estimate.