Economist and ‘The Bitcoin Standard’ author Saifedean Ammous published a detailed essay on 15 September 2026 arguing that Bitcoin mining may have already reached a peak in capital expenditure and electricity consumption.
The author presented the claim as a testable hypothesis rather than a certainty, tying it to the mathematics of the halving, slower price appreciation, and competition from artificial-intelligence data centres.
Ammous wrote that mining expands only while Bitcoin’s expected real price growth exceeds roughly 18.92% a year, the rate needed to offset the 50% cut in new coins every four years. Once that hurdle is no longer cleared, he said, miners at the margin reduce investment and power use.
Halving Economics and the Growth Threshold
The essay frames the issue in simple arithmetic. Because the daily subsidy halves every 210,000 blocks, the dollar value of mining output rises only if the BTC price more than doubles over four years after adjusting for dollar devaluation.
Ammous calculated that the current epoch’s subsidy revenue is only 28% higher than the previous epoch’s despite a 2.56-fold rise in the average BTC price. After dollar devaluation of 6–7% a year, the real increase approaches zero.
He noted that early epochs produced far larger multiples: the second epoch generated more than 30 times the dollar subsidy of the first, the third eight times the second, and the fourth 2.86 times the third. Those multiples financed rapid growth in hashrate and energy use. The fifth epoch, now more than two years old, has not repeated the pattern. Ammous argued that market participants examining five-year compound annual growth rates as low as 10% are less willing to commit new capital to mining equipment.
Slowing Subsidy Revenue Across Epochs
A striking data point in the essay is 14 March 2024. On that day the Bitcoin price reached about $73,800 while the subsidy was still 6.25 coins per block, producing a daily subsidy worth roughly $65 million. Two years into the current 3.125-coin subsidy era the price has not yet doubled to the $147,600 needed to match that dollar figure. Ammous observed that if the price fails to hit that level before the next halving, the fifth epoch will be the first in which daily mining revenue never exceeds the prior epoch’s high.
He projected the required prices forward: $295,200 in the sixth epoch, $590,400 in the seventh, and so on. Because Bitcoin cannot double in value every four years indefinitely once it captures a large share of global monetary demand, some future date must mark the all-time peak in subsidy value. Whether 14 March 2024 already holds that record remains an open empirical question, the author wrote.
The same margin pressure now visible in public filings aligns with Ammous’s hypothesis. A CoinShares Q2 2026 mining review found listed miners as a group fell below cash breakeven, with an average pre-tax production cost of about $75,500 against a quarter-end BTC price of $58,400 and a record-low hashprice of $27.7 per petahash per day in June.
Read: Bitcoin Mining Sector Faces Tightest Margins Since the 2024 Halving
Network hashrate sat roughly 50% below its longer-term trend after the first six-month decline since the 2021 China ban, while firms such as Core Scientific recorded a $41.9 million loss to cancel remaining next-generation hardware deliveries and shift sites toward high-performance computing and AI colocation, as disclosed in its Form 10-Q.
Those cash-cost figures and contract cancellations supply contemporaneous industry evidence that capital and electricity are already leaving Bitcoin mining for competing uses, the dynamic Ammous argued would follow once subsidy growth no longer outpaced halvings and rising opportunity costs.
Difficulty Stagnation and Falling Electricity Use
Network difficulty reached an all-time high on 30 October 2025 and has not surpassed it more than 300 days later, the second-longest such stretch in Bitcoin’s history. Earlier long draw-downs followed a 94% price crash in 2011 or the 2021 Chinese mining ban. The current episode, Ammous noted, has occurred after a milder 54% draw-down and without a comparable regulatory shock.
The most direct test of the hypothesis, he said, is the Cambridge Bitcoin Electricity Consumption Index. Preliminary figures from the Cambridge Centre for Alternative Finance put annualised consumption at approximately 190 TWh in December 2025. By mid-2026 the same series had fallen back into the high-130 TWh range. Ammous treated December 2025 as a candidate peak; a sustained rise above that level would falsify this particular claim.
AI Competition and Implications for Bitcoin
Ammous identified artificial intelligence as an accelerant. Data-centre demand for power and infrastructure has risen sharply, raising the opportunity cost of using the same sites for Bitcoin mining. Unlike Bitcoin, whose subsidy shrinks on a fixed schedule, AI demand can expand with productivity gains. Many stranded-energy locations first developed by miners are now better connected and therefore more attractive to AI operators.
The author stressed that a smaller mining industry would not threaten Bitcoin’s security. Supply issuance continues on its predetermined path, and transaction fees, though still a small share of miner revenue, remain available to route around any attempted censorship. He argued that capital previously destined for new mining farms is more likely to buy BTC itself, a development he viewed as supportive of the price. Reduced electricity use, he added, would also diminish environmental criticism of the network.
Ammous concluded that Bitcoin itself can continue to grow while the mining sector contracts. The practical recommendation he offered readers was straightforward: those who understand Bitcoin’s monetary properties should buy the asset rather than finance additional mining capacity; those who control cheap power should consider pivoting infrastructure toward AI. The hypothesis, he wrote, will be tested by whether Cambridge’s electricity series stays below its late-2025 high and whether difficulty remains below its October 2025 peak.
Also read: BIS Study – Bitcoin Transfer Estimates Differ by Up to Sixfold
