Bitcoin’s Broken Production Cost Floor Splits Miners: Survivors and Sellers Emerge
Bitcoin’s production cost floor has cracked. Not softened. Cracked. Miners are now sorting into two camps: those that can keep running and those that need to sell. BTC is trading just above $60,000, while the estimated all-in cost to produce one coin is near $84,300. That leaves a rough 25% gap, which puts much of the network underwater on a full-cost basis. The old theory was neat enough to fit in a trader note: Bitcoin should not trade far below its production cost for long. My take: that theory is looking less like a floor and more like a stress test.
For years, traders treated production cost almost like a hard floor. BTC falls too far, miners unplug machines, selling pressure fades, price bounces. Clean idea. Too clean, honestly. BTC has now spent weeks below that line, and the network is still running. Why does this matter? Because the mid-June difficulty adjustment showed the actual mechanism, not the slogan. Difficulty fell 10.09%, from 138.96 trillion to 124.93 trillion. Galaxy Research called it the second-largest downward adjustment of 2026 and the eleventh-largest in Bitcoin’s history. That epoch lasted 15.6 days instead of the usual two weeks because blocks slowed after higher-cost machines went offline. The protocol noticed and made mining easier for the operators still plugged in. So yes, the correction mechanism works. It just does not defend price the way the “cost floor” story suggested.
The sharper number is hashprice, the daily revenue a miner earns for each unit of computing power. It falls when BTC drops, when difficulty rises, or when transaction fees dry up. It rises when BTC rallies. It can also rise when weaker miners shut down and difficulty resets lower. In July 2025, hashprice was near $63 per petahash per day. By early June 2026, it had fallen into the high $20s. Hashrate Index and many operators treat that high-$20s zone as gross breakeven before debt and overhead. After the June difficulty cut, hashprice pushed back above $30. CoinShares, in its Q1 2026 mining report, estimated that public miners had a weighted average cash cost of about $79,995 per Bitcoin in Q4 2025, while hashprice slid from the $36-$38 range toward $29. CoinShares also projected that 15% to 20% of the global fleet would become unprofitable once power costs rose enough.
Averages blur the thing that actually matters. That is why production cost fails as a floor. A miner with latest-generation hardware below 15 joules per terahash and power under 5 cents per kilowatt-hour can still make decent money. An older fleet paying 6 or 7 cents can lose money on every block it finds. Same Bitcoin price. Different business entirely. When BTC falls, revenue per unit of hash falls with it. The most expensive machines become uneconomic first, and their operators start reaching for whatever is left: sell BTC, switch off rigs, delay expansion, renegotiate power contracts. Sometimes they raise capital because the alternative is worse. Counter to the usual advice, the cheapest miner is not just “more profitable”; it is playing a different game. Once enough hash rate leaves, difficulty moves lower and the miners still online get a larger share of the block subsidy. That helps. Slowly. Unevenly. It does not stop BTC from falling while the adjustment plays out. Production cost decides who survives the slide. It does not decide where the slide ends.
This matters for the wider crypto market, especially for the “Bitcoin as safe haven” pitch. I have never loved that framing, and this is one reason why. BTC may trade like digital gold during some macro scares, but miners still have bills, lenders, power contracts, payroll, and shareholders asking awkward questions. When public miners cut their holdings by more than 15,000 BTC from peak levels, those coins have to land somewhere. Core Scientific sold about 1,900 coins in January and plans to clear most of what remains. Bitdeer cut its balance to zero in February. Riot sold 1,818 coins in December. In Q1 2026, public miners sold more BTC than they did in all of 2025, even more than during the Terra-Luna panic. Is that just profit-taking? No. It is operating cash. That kind of supply can smother a rally, especially when demand is already soft. Bitcoin is decentralized, but its producers are still companies. When those companies get squeezed, price feels it.
In older downturns, a stressed miner usually had two choices: keep hashing or shut down. The largest public operators now have a third option: shift toward AI and high performance computing. Most guides frame this as diversification. That is only half right. It is also a rescue route, and in some cases a capital-intensive one. CoinShares says public miners have announced more than $70 billion in cumulative AI and HPC contracts, and it projects that listed miners could get as much as 70% of revenue from AI by the end of 2026, up from about 30% now. The deals are not small. Core Scientific’s expanded CoreWeave agreement totals $10.2 billion over twelve years. TeraWulf has booked $12.8 billion in contracted HPC revenue. Hut 8 signed a $7 billion, fifteen-year AI infrastructure lease. Bitfarms even dropped Bitcoin from its name. That is not subtle.
The sector now has three rough groups, though I would not treat them as equal buckets. Some miners already have AI contracts and are moving capacity over, often with debt. Cipher is one example: its $1.7 billion in senior secured notes pushed quarterly interest expense to $33.4 million. A second group is testing frameworks and pilots but has little or no revenue from them yet. The third group is still tied almost entirely to Bitcoin and lives or dies with hashprice. Investors are pricing these groups differently now. Hybrid infrastructure companies trade partly on contract execution and balance sheet risk. Pure Bitcoin miners are a cleaner bet on BTC and difficulty. Treasury policy matters too. Small operators with cheap power can also do well when difficulty resets, because cheaper energy gives them room that levered public peers may not have.
What this means
The Bitcoin mining industry is going through a painful sorting process. “Maturing” is the polite word, but it mostly looks like stress. The broken production cost floor changes how investors should think about Bitcoin price support. Cost of production sorts miners. It does not hold up BTC. Yes, this contradicts the old miner-capitulation shorthand. Bear with me: the longer BTC stays below the average production cost, the sharper the split gets. Efficient operators with cheap power and newer machines get room to breathe. Other revenue lines help. Operators without those advantages start running out of moves. That keeps pressure on Bitcoin’s supply side because stressed miners sell coins when they need cash. The AI pivot gives some companies another path, but it also adds a new risk. If the AI infrastructure cycle cools, those hybrid miners could get hit before Bitcoin gives them any relief.
For traders and investors, the numbers to watch are hashprice, difficulty adjustments, public-miner treasury balances, and how many coins miners send to exchanges. I would put hashprice first. If Bitcoin moves back toward $100,000, hashprice could return toward $37, treasury sales may slow, and hardware upgrades could restart. If BTC chops sideways near production cost, the sector probably keeps grinding. Public miners sell coins. They chase AI deals. They wait for difficulty to give them a little help. If BTC falls again, more high-cost hash rate goes offline, the gap between hybrid and pure-play equities widens, and miners with the cheapest power gain share. None of this breaks the network. The mid-June drop partly reversed, block times moved back near 10 minutes, and some curtailed capacity returned as price steadied. That suggests the hash rate that left was reacting to thin margins, not leaving forever. The next difficulty adjustment, usually every two weeks, is worth watching. It is one of the cleaner real-time reads on miner stress.