Bitcoin’s Hashrate Came Back. Difficulty Took the Pay.

Bitcoin’s network did something in October that almost nobody writing about mining in the first half of 2026 expected: it rebuilt its own hashrate. After a year spent explaining why the number was shrinking (miners powering down, capital and megawatts draining toward artificial intelligence, the first voluntary decline in the network’s history), the aggregate hashrate climbed back above 1 zettahash per second, roughly 1,069 exahashes on a seven-day average by early November, its highest reading since March and within reach of the record set a year ago, according to Million Miner’s November mining update.

The problem is what came with it. The same machines that pushed the hashrate back up also pushed mining difficulty, Bitcoin’s self-adjusting measure of how hard each block is to find, up almost 12 percent in two steps across thirteen days, to 148.36 trillion, the highest since late 2025, per news.bitcoin.com. Difficulty is the network’s thermostat, and when it rises, every miner earns less for the same work. Hashprice, the dollar figure a miner expects per unit of computing power per day, fell to about $33, down 8.7 percent in a single week and down roughly 17 percent across October, The Currency Analytics reported. And it happened at the worst possible moment: Bitcoin, which started October above $83,000, slid toward $76,000 by 7 November after a late-October exchange breach and a Thursday rout in chip stocks dragged risk assets lower, crypto.news noted.

In other words, the recovery is already eating itself. The victory lap for the rebound, run in the crypto press over the past two weeks, left out the second half of the mechanism: on Bitcoin, the network answers every hashrate recovery with a difficulty bill, and this time the bill arrived just as the price that pays it began to fall. Here is how a number coming back became a squeeze coming back.

The number that came back

Start with the arc, because the round trip is the story. A year ago, in October 2025, Bitcoin’s hashrate peaked near 1,160 exahashes per second with the price at its all-time high above $126,000. Then came the long slide: through the first half of 2026 the network shed close to a quarter of its computing power, bottoming near 870 exahashes early in the year as weak margins and the pull of AI leasing idled older machines. For most of 2026 that decline was treated as the new normal, even as a permanent regime change.

October broke the pattern. The hashrate crossed back above 1 zettahash per second around 22 October, its highest since 3 March, and kept climbing: a seven-day average near 1,069 exahashes by 2 November, with a single-day peak of 1,088 exahashes on 26 October, Million Miner’s figures show. CryptoQuant data pointed the same way, with its chief executive describing miners as running at full capacity. In raw hash terms, the network had nearly undone a year of contraction in a single quarter.

Why it matters: hashrate is the single best proxy for how much energy and silicon is committed to securing Bitcoin, the physical cost of rewriting its ledger. A network defending itself with 1,069 exahashes is a vastly harder target than one defending itself with 870. HOGE Wire covered the up-leg of this move in detail in its look at why the recovery ran ahead of the price. This piece is about the part that comes next, the part the recovery narrative skipped: what that recovered hashrate does to the miners who built it.

2026 was supposed to be the year hashrate stopped growing

To see why the rebound surprised people, remember the thesis it broke. Bitcoin’s hashrate had fallen before, but always because someone made it fall: the 2021 China ban that knocked more than half the network offline overnight, the capitulations of 2018 and 2022 when crashing prices bankrupted miners. The 2026 decline was different, and that difference was the whole point of the year’s mining commentary. Nobody banned anything. Miners chose to power down, or to repoint their megawatts at artificial intelligence, where a signed lease pays more per watt than a volatile block reward.

The numbers behind that pivot were real and large. Research house CoinShares has tracked more than $70 billion in announced AI and high-performance-computing contracts across the listed mining sector, and its analysts estimated some public miners could draw as much as 70 percent of their revenue from AI by the end of 2026, up from roughly 30 percent, a shift James Butterfill of CoinShares called one of the most challenging periods the industry had faced since the 2024 halving. The logic looked airtight: if AI pays better, the hashrate leaves and does not come back.

Except it did come back, and the shape of the return shows the pivot was oversold as an exodus. It was a bifurcation. Million Miner’s November data splits the public miners cleanly: five AI-focused operators mined 47 percent less Bitcoin than a year earlier, while five Bitcoin-focused operators grew their combined output 16 percent. Bitdeer, running a hybrid strategy, held about 82.3 exahashes of self-mining hashrate even as it expanded AI operations. The dedicated miners simply scaled into the room the AI-pivoters vacated, and the efficiency of new silicon let them add hash without adding much power. Fred Thiel, chief executive of MARA, has long described this as the trap of the business: mining, he told CoinGeek, is “a zero-sum game. As more people add capacity, it gets harder for everybody else. Margins compress, and the floor is your energy cost.” In 2026 the floor held, the capacity came back, and it got harder for everybody.

The thermostat answered: difficulty’s two-step climb

Bitcoin recalibrates the hardness of its mining puzzle every 2,016 blocks, roughly every two weeks, with one goal: to keep the average time between blocks near ten minutes no matter how much or how little computing power is pointed at the network. More hashrate means blocks arrive too fast, so difficulty rises to slow them back down. Less hashrate means the opposite. The adjustment is automatic, written into the protocol, and answers to no committee. It is the closest thing Bitcoin has to a central bank, and the only thing it targets is time.

Through the summer, with hashrate depressed, the thermostat ran cold: difficulty bottomed at 124.93 trillion on 14 June, the year’s low. The autumn rebound flipped it hard. On 16 October difficulty jumped 5.92 percent to 140.58 trillion, the third-largest increase of 2026 and the highest reading since early March, news.bitcoin.com reported. Thirteen days later, on 29 October, it rose another 5.54 percent to 148.36 trillion. Back to back, that is a climb of almost 12 percent in under two weeks, and about 18.8 percent above the June floor.

One caution, because several trackers got it wrong in real time: 148.36 trillion is not a record. Some outlets, including hashprice monitors, labeled it an all-time high; it is not. The genuine record is 155.97 trillion, set around the hashrate peak in October 2025, which leaves the current reading roughly 5 percent short. The distinction matters, because the gap between here and the record is exactly the room the network still has to keep tightening if the hashrate holds.

Difficulty’s one real structural weakness, the timewarp bug that could let a dishonest majority of miners manipulate timestamps to force difficulty down, is the target of a live 2026 soft-fork effort, BIP-54, the Great Consensus Cleanup, which HOGE Wire examined alongside the odds on Bitcoin’s next fork as Taproot turned five. That fight is about closing a theoretical attack. The adjustment doing the damage this autumn is the ordinary one, working exactly as designed.

How difficulty quietly takes the reward back

The mechanism that turns a hashrate rebound into a pay cut is arithmetic, not sentiment. A miner’s revenue per unit of work is captured by hashprice, which can be written as the daily block rewards (about 144 blocks times the 3.125 BTC subsidy plus fees) multiplied by the Bitcoin price, then divided by the total network hashrate. Hashrate sits in the denominator. When it climbs and the price does not climb with it, hashprice falls, and every miner on the network, efficient or not, takes home less for the same machines.

Difficulty is the lever that makes this automatic and unavoidable. It is the network’s way of saying that the same block reward now has to be split among more committed computing power. Nikolaos Panigirtzoglou, the JPMorgan analyst who tracks mining economics, summarized the normal direction of travel when he told TFTC: “When bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower.” In the third quarter of 2026 that sentence ran in reverse. Bitcoin rallied through its best quarter since 2017, miners switched idle machines back on, the hashrate climbed, and difficulty adjusted higher. The trouble is the timing: the price leg that justified switching those machines on began to give way in late October, leaving miners holding elevated difficulty into a falling market.

Panigirtzoglou’s own estimate of the all-in production cost, around $78,000 per coin, with a difficulty-to-price sensitivity he pegs near 0.62, frames the danger. When the price sits below that cost, as it did for stretches of 2026, higher-cost miners bleed. The difficulty spike of late October raised the cost of producing a coin for the whole fleet at precisely the moment the market stopped paying above it. That is the squeeze, reopened not by a crash but by the network’s own success in rebuilding itself.

Hashprice back at the floor

The pay cut shows up fastest in hashprice, and it has been brutal. After touching above $41 per petahash per day twice in the thirty days around 2 October, hashprice closed October near $33.18 and printed $32.86 on 2 November, an 8.7 percent drop in a single week from $35.99, with a decline of roughly 17 percent across the month, according to Shattered’s mining dashboard and The Currency Analytics. For context, hashprice traded near $56 in the third quarter of 2025 and above $40 as recently as early October. The number miners live by has round-tripped to the lows that defined the first half of the year.

Here is the paradox that makes the rebound so treacherous. October was, by one measure, the mining industry’s best month since November 2025: total miner revenue reached about $1.215 billion, an average of $39.2 million a day, Million Miner calculated. Yet per-unit pay fell all month. Both facts are true because revenue equals hashprice multiplied by hashrate, and the hashrate grew faster than hashprice shrank. The industry as a whole earned more while each individual machine earned less, a classic commodity-glut signature: everybody produces more, and the price of what they produce falls. The miners who added the capacity are the ones competing their own margins away.

At $33 hashprice, the math turns unforgiving for anyone without cheap power or new machines. An operator running the latest hydro-cooled hardware near 9.5 joules per terahash can survive on power costs up to roughly 14.5 cents per kilowatt-hour, but an older Antminer S19j Pro at 29.5 joules per terahash needs power below about 4.7 cents just to break even on electricity, before accounting for the hardware itself. That is cheaper than most of the grid delivers. The least efficient quarter of the fleet, the machines that switched back on during the Q3 rally, are the first to flip underwater as hashprice falls, and they will be the first to switch off again if it stays there.

The price leg gave way at the worst time

A difficulty spike is survivable if the price is rising; it is a vice if the price is falling. In late 2026 the price began to fall. Bitcoin opened October around $83,565 and spent the month drifting lower before two shocks accelerated the slide. On 29 October, a breach at Binance drained billions from the exchange, a story HOGE Wire covered in its account of the $6.3 billion signature and the forensic work that followed, puncturing risk appetite across crypto. Then, on 5 November, a rout in semiconductor stocks spilled into digital assets; by the following morning Bitcoin had shed roughly 5 percent, about $1.03 billion in leveraged positions had been liquidated, and spot Bitcoin exchange-traded funds had bled some $430 million, crypto.news reported. By 7 November the price sat near $76,000 on CoinGecko.

For miners, the sequence is almost perfectly adverse. Hashprice near $33 already assumes a Bitcoin price around $76,000 to $78,000; every dollar the market sheds below that pushes the figure lower still, with no relief from difficulty until the next retarget two weeks out. The machines cannot un-commit in real time. A miner who energized capacity in September, when a coin fetched more than $83,000, is now selling that same production into a $76,000 market against a difficulty 12 percent higher than when the decision was made. The recovery that looked like vindication in October looks like a trap in November.

Q3 earnings landed into a reopening squeeze

The timing put a spotlight on the quarter just reported. Riot Platforms, one of the largest listed miners, released third-quarter results on 4 November, and they captured the whole tension of the moment: a strong quarter, printed just as its support was eroding. Riot posted revenue of $181.4 million and adjusted EBITDA of $191 million, with a GAAP loss per share of 24 cents that beat the 28-cent loss analysts expected, GuruFocus reported. The company produced 1,637 BTC, up from 1,406 a year earlier, and ended the quarter with 44.9 exahashes of deployed hashrate, about 4.9 percent of the global network.

The cost side is where the Q3 recovery is visible, and where its fragility hides. Riot’s cost to mine a Bitcoin fell to $39,265 from $46,324 a year earlier, helped by an all-in power cost of just 3.2 cents per kilowatt-hour and $25 million in power-curtailment credits, worth about $15,067 for every coin mined. At a Bitcoin price above $80,000, as held through much of the quarter, that is a comfortable cash margin. At $76,000 with difficulty 12 percent higher, the comfort thins fast, and Riot sits at the efficient end of the industry. CoinShares put the public sector’s weighted-average cash cost near $75,500 a coin in the second quarter, which means the median miner was already producing at a level the November price barely clears before depreciation.

The headline Riot wanted investors to read was not the mining line at all. It was the $38 billion in contracted data-center revenue it unveiled in the same presentation, the clearest sign yet that the company sees its future as an AI-infrastructure landlord that happens to mine Bitcoin on the side. HOGE Wire dug into how that reframing distorts the profit picture across the sector in its read of the Q3 profit line. For the purposes of the hashrate story, the point is narrower: even the miner with the strongest balance sheet is telling the market not to value it on mining margins, because mining margins are exactly what the difficulty rebound is squeezing.

The treasuries emptied out

If you want proof that miners themselves do not trust the recovery, watch what they did with their coins. Riot sold 10,599 BTC for $751.5 million during the third quarter and 20,264 coins for nearly $1.5 billion across 2026, news.bitcoin.com reported. Its treasury fell from 18,005 BTC at the start of the year to just 2,418 by 30 September. Management said plainly that it expects to sell substantially all newly mined Bitcoin to finance development and operating expenses rather than rebuild a sizable crypto reserve, a sharp break from the hoard-everything orthodoxy that defined the last cycle.

Riot was not alone. Hut 8 sold 954 BTC for $74.8 million in the quarter and another 2,148 coins for $175.4 million in October, roughly $250 million combined, to fund share buybacks, while still holding 16,787 BTC worth about $1.4 billion. MARA, the biggest miner of all, had already sold roughly $1.5 billion of Bitcoin earlier in 2026 to fund its own infrastructure pivot. The common thread is that the companies rebuilding the hashrate are also selling the coins that hashrate produces, into a falling market, to pay for data centers. That makes them structural sellers again, the forced-seller dynamic that has turned every past mining recovery into a drag on price just when the network looked strongest.

Why hashrate lags price and never leads it

The deepest lesson of the autumn round trip is one the market relearns every cycle and forgets by the next: hashrate is a lagging indicator, not a leading one. It does not predict the price; it responds to it, slowly, with a delay measured in the months it takes to order machines, secure power, and energize a site. The hashrate that surged in October was the delayed echo of the summer rally, when a recovering price made switching idle rigs back on look profitable. By the time that capacity came online, the price that justified it had already turned. Reading a hashrate high as a bullish signal gets the causality backwards; the high is the network catching up to a rally that is already fading.

Efficiency complicates the picture in a way that matters for 2026 specifically. A meaningful share of the rebound was not new power at all, but new silicon: machines like the hydro-cooled Antminer S23 class, rated near 9.5 joules per terahash, pack far more hashing into the same megawatt than the S19 generation they replace. That decoupling, hashrate rising while the power bill stays flat, is real, and it means part of the recovery is structural rather than a bet on price. But it cuts both ways. Each efficiency jump also raises the break-even bar for the older fleet, because the network as a whole gets harder to compete in. The table below shows how thin the margin for error has become at a $33 hashprice.

The takeaway for anyone reading mining stocks or network charts: treat a rising hashrate as evidence that yesterday’s economics were good, and treat a rising difficulty as the mechanism that will erase today’s. The two numbers that look like strength are the same force that reopens the squeeze.

A record in hash, not in dollars

There is a second gap hiding inside the rebound, and it reframes what the recovered hashrate actually buys. Hashrate measures physical commitment; it does not measure the dollar cost of attacking the network. Those two things have pulled apart in 2026. Bitcoin’s security budget, the total it pays miners to defend the ledger, is roughly the daily issuance of about 450 BTC multiplied by the price, annualized. At $76,000 that is on the order of $12.5 billion a year. A year ago, with the hashrate at a similar level but the price near $126,000, the same security was worth closer to $20 billion. The network is nearly as hard to out-compute as it was at the peak, but far cheaper, in dollars, to out-spend.

That distinction grows more consequential as the subsidy shrinks. More than 99 percent of the security budget still comes from the block subsidy, with transaction fees contributing under 1 percent of the reward in a typical week. The 2028 halving will cut that subsidy from 3.125 BTC to 1.5625, halving the dollar budget at a constant price unless fees or the price fill the gap. A hashrate record set on a shrinking dollar budget is a more brittle thing than it looks on a chart. It is worth contrasting with proof-of-stake, where security is priced in bonded capital rather than burned energy; Ethereum’s model faced its own stress test when a market crash tested whether staked-ETH derivatives would hold, which HOGE Wire traced in its look at why stETH kept its peg. Bitcoin buys its security with electricity, and electricity is priced in a currency that just fell.