Hashrate Growth Reverses as Bitcoin Falls Below Cost
Bitcoin's hashrate growth just went into reverse. A slide below the cost of production triggered the biggest difficulty cut since June, exactly as the models said it would.
For most of the past decade, writing about Bitcoin’s hashrate meant writing about growth. The combined guessing speed of the mining network climbed from a few kilohashes on a hobbyist laptop in 2009 to more than a zettahash, a one followed by twenty-one zeros, by late 2025. The line on the chart went up and to the right so reliably that it became a kind of cliche. In the middle of November 2026, that story turned inside out. Bitcoin’s price slid below the cost of producing a coin, the least efficient machines began switching off, and on 13 November the network printed its largest downward difficulty adjustment since June. Hashrate growth did not merely stall. It reversed.
This is not a crisis, and it is not a surprise. It is the most predictable thing Bitcoin’s mining network does. When the revenue from mining falls below what it costs to keep rigs powered, high-cost machines go dark, total hashrate drops, and roughly every two weeks the protocol quietly makes mining easier so that blocks keep arriving about every ten minutes. Economists have a name for the mechanism, analysts have modeled it to the basis point, and in November 2026 it is running in reverse in real time. This explainer walks through what hashrate growth actually is, why it just turned negative, what the reversal does and does not mean for the security of the network, and what to watch as the year closes. It belongs to the same macro story as Bitcoin’s push toward its first back-to-back losing year, only told from the miners’ side of the ledger.
The thermostat just ran in reverse
Start with the hard numbers. On 13 November, at block 975,744, Bitcoin’s mining difficulty fell from 148.36 trillion to 134.82 trillion, a drop of roughly 9.1 percent, according to CoinWarz difficulty data. That is the steepest single adjustment since a 10 percent cut in June, and Crypto Briefing had flagged it coming as blocks slowed to an average near 10 minutes and 48 seconds instead of the 10-minute target. Slow blocks are the fingerprint of a shrinking hashrate: fewer machines are racing for each block, so the network takes longer to find one, and the next adjustment lowers the bar to compensate.
The backdrop was a brutal month for the price. Bitcoin traded around $64,000 on 16 November, down from roughly $79,000 at the start of the month and about 49 percent below the record $126,080 set in October 2025, per CoinGecko. It was the sixth consecutive weekly loss and the first weekly close below the 200-week moving average since the summer, with a weekend dip toward $62,000 and more than $1.9 billion in leveraged positions wiped out in a single day, Crypto Briefing reported. Fear, not greed, set the tone. For miners, a price near $64,000 is not an abstraction. It is below the number at which a large share of the industry breaks even.
| Metric | Mid-November 2026 |
|---|---|
| Bitcoin price | around $64,000 |
| Market capitalization | roughly $1.29 trillion |
| Network hashrate | about 960 to 1,000 EH/s (off the 26 Oct peak near 1,088) |
| Mining difficulty | 134.82 trillion (after a 9.1 percent cut on 13 Nov) |
| Difficulty vs record | about 13 percent below the 155.97 trillion peak |
| Hashprice (estimate) | roughly $30 per PH/day, near June’s record low |
| Estimated production cost | around $78,000 per coin (JPMorgan) |
| Next difficulty retarget | about 28 Nov, estimated near minus 7 percent |
What hashrate growth normally means
Hashrate is the combined speed at which all the computers on the Bitcoin network make guesses, measured in hashes per second. Mining is a guessing game: machines repeatedly run a block’s data through the SHA-256 hash function, trying to land on an output below a target set by the current difficulty. The more guesses per second, the more often that network finds a valid block. Hashrate growth simply means more machines, faster machines, or both are joining the race. It is the clearest single proxy for how much real-world capital, meaning rigs, electricity, cooling, and real estate, is committed to securing Bitcoin at any moment.
Because the raw numbers are so large, the industry climbs a ladder of prefixes. A modern application-specific integrated circuit, or ASIC, produces hundreds of trillions of hashes per second on its own. Stack a few hundred thousand of them across the planet and you reach the exahash and zettahash scale the network now operates at. The ladder below puts the units in order, which matters because the difference between a terahash and an exahash is a factor of a million, and casual coverage mixes them up constantly.
| Unit | Hashes per second | Rough reference point |
|---|---|---|
| KH/s (kilohash) | 1,000 | A 2009 laptop CPU |
| MH/s (megahash) | 1 million | Early GPU miners, 2010 to 2011 |
| GH/s (gigahash) | 1 billion | First-generation ASICs, 2013 |
| TH/s (terahash) | 1 trillion | A single modern ASIC today |
| PH/s (petahash) | 1 quadrillion | A mid-size mining facility |
| EH/s (exahash) | 1 quintillion | A top public miner’s whole fleet |
| ZH/s (zettahash) | 1 sextillion | The entire network at its 2025 peak |
Growth has four engines. The first is price, which flows through to mining revenue and is the cyclical, fast-moving one. The second is ASIC efficiency, the slow and structural engine that lets each watt do more work every year. The third is energy, the availability of cheap power to plug machines into. The fourth is capital, the willingness of operators and their lenders to fund new rigs and data centers. When all four point the same way, hashrate rises in a near-straight line. When one reverses hard enough, as price just did, it can drag the whole number down despite the others. Understanding the reversal means understanding which engine failed.
The difficulty thermostat, explained
Bitcoin is designed to produce one block about every ten minutes, no matter how much computing power is pointed at it. It holds that pace with a feedback loop. Every 2,016 blocks, which is roughly two weeks, the protocol looks at how long those blocks actually took. If they arrived faster than ten minutes on average, difficulty rises to slow things down. If they arrived slower, difficulty falls to speed them up. The adjustment is capped at a factor of four in either direction, a clamp that has never come close to binding in modern times. Think of it as a thermostat for block production: it does not care why the hashrate changed, only that blocks drift back toward the ten-minute setpoint.
This feedback loop is also why nobody can measure Bitcoin’s hashrate directly. There is no central meter. Analysts infer it from difficulty and how fast recent blocks arrived, using the relationship that hashrate is approximately difficulty multiplied by 2 to the 32nd power, divided by the 600-second block target. Plug in the post-cut difficulty of 134.82 trillion and you get a network running near 965 exahashes per second, which lines up with the slower-than-target block times of mid-November. Every hashrate figure you read, including the ones in this article, is an estimate smoothed over some window. That is a feature, not a flaw, and it is the reason the thermostat has kept Bitcoin’s block schedule remarkably steady for sixteen years through booms, bans, and crashes alike.
Why a price below the cost of production pulls hashrate down
Miners earn Bitcoin but pay their bills in dollars. Electricity, hosting, debt service, and worn-out hardware are all priced in fiat, while the block reward arrives in BTC. The pivotal number is therefore the all-in cost of producing one coin. When Bitcoin trades above that cost, mining is profitable and machines stay on. When it trades below, the least efficient operators lose money on every block they help find, and the rational move is to switch off until conditions improve. That decision, multiplied across thousands of facilities, is what makes the aggregate hashrate fall.
Analysts at JPMorgan, led by managing director Nikolaos Panigirtzoglou, described the loop precisely months before it played out. In their framework, as summarized by TFTC, “When bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower.” The bank pegged Bitcoin’s production cost near $78,000 per coin and has noted that the network is more sensitive to price than in past cycles, with hashrate reacting more sharply to each swing because so many miners now operate close to break-even. CoinShares has put the industry’s average cash cost a little lower, near $75,500, but the conclusion is the same. At $64,000, Bitcoin is trading well under both figures.
That is the whole story of the November reversal in one sentence. Price fell through the cost floor, the marginal miner powered down, the hashrate slipped, and the thermostat answered with a cut. Panigirtzoglou’s team expects this behavior to persist, with larger and more frequent difficulty adjustments for as long as Bitcoin stays well below its production cost. The model is not a forecast anymore. It is a description of the present.
The 13 November cut was the biggest since June
To see how sharp the turn was, look at the difficulty record across 2026. The year opened with difficulty near its all-time high, slid through a long spring and summer of cuts as the stall set in, then staged a genuine autumn recovery. Two large upward adjustments in October, roughly 5.9 percent on 16 October and 5.5 percent on 29 October, pushed difficulty to 148.36 trillion and briefly revived the growth narrative. The 13 November cut gave back most of that gain in a single step. The next retarget, expected around 28 November, is estimated near minus 7 percent, which would drag difficulty back toward 125 trillion, almost exactly where it bottomed in June. In difficulty terms, the network is on course to erase the entire autumn rebound.
| Date (2026) | Adjustment | Resulting difficulty |
|---|---|---|
| 14 June | minus 10.1 percent | 124.93 trillion (year’s low) |
| 27 June | plus 7.2 percent | 133.87 trillion |
| 5 September | plus 1.3 percent | 127.45 trillion |
| 19 September | plus 4.2 percent | 132.76 trillion |
| 16 October | plus 5.9 percent | 140.58 trillion |
| 29 October | plus 5.5 percent | 148.36 trillion |
| 13 November | minus 9.1 percent | 134.82 trillion |
| 28 November (est.) | about minus 7 percent | roughly 125 trillion |
Two things stand out. First, the swings are getting bigger: a 5.9 percent hike and a 9.1 percent cut inside a single month is a wide band by historical standards, exactly the elevated volatility JPMorgan warned about. Second, the thermostat is symmetric. The same mechanism that made mining harder as idle rigs re-energized in October is now making it easier as they switch back off. Nothing broke. The feedback loop is simply pointing the other way.
A year-long round trip
The reversal lands at the end of the strangest year in the hashrate’s history. For a decade the number only knew one direction. In 2026 it made a full round trip. Hashrate set its all-time high near 1,156 exahashes per second on 18 October 2025, according to Luxor’s Hashrate Index, after the seven-day average first held above a full zettahash in early September 2025, a milestone CoinDesk recorded at the time. Then it rolled over. The seven-day average fell back below one zettahash on 17 January 2026 at about 988 exahashes per second, as Yellow reported, and bottomed near 850 exahashes per second in early February.
That was the first decline of its kind. When the network’s power dropped in 2021, a government had just banned mining. When it dropped in 2018 and 2022, prices had collapsed. The early-2026 slide was different: a voluntary retreat as operators redirected megawatts toward artificial-intelligence contracts even with the network near its peak. That first-quarter slide was the network’s first quarterly decline in years, a clean break from a decade of near-uninterrupted growth. Then autumn brought a real recovery, the seven-day average crossing back above one zettahash on 11 October and peaking near 1,088 exahashes per second on 26 October. Mid-November is the fourth act: a price-driven decline on top of the structural one.
| Milestone | Approximate date | Level |
|---|---|---|
| First sustained zettahash (7-day avg) | Early September 2025 | around 1,000 EH/s |
| All-time high hashrate | 18 October 2025 | about 1,156 EH/s |
| Back below one zettahash | 17 January 2026 | about 988 EH/s |
| Cycle trough | Early February 2026 | about 850 EH/s |
| Recovery above one zettahash | 11 October 2026 | around 1,000 EH/s |
| Autumn peak | 26 October 2026 | about 1,088 EH/s |
| Mid-November reversal | November 2026 | about 960 to 1,000 EH/s |
The longest stretch ever below the record
Difficulty tells the same story with even more drama. Its record of 155.97 trillion was set on 29 October 2025. By late October 2026 the network had gone 374 days without reclaiming that peak, the longest such gap in Bitcoin’s history, as Bitcoin.com News documented. The 13 November cut and the expected late-November cut push the streak past 390 days and widen the gap to the record rather than closing it. For context, the drought now exceeds the 364-day stretch that followed the 2011 crash and dwarfs the 253-day gap after China’s 2021 mining ban.
That comparison carries weight because those earlier episodes were far more violent on the surface. Economist Saifedean Ammous, author of The Bitcoin Standard, observed that neither the roughly 95 percent price crash of 2011 nor China’s abrupt expulsion of the entire mining industry in 2021 kept difficulty below its peak for as long as the current, quieter stretch has. The 2026 drought is not the product of a single shock. It is the cumulative weight of a soft price, thin margins, and a steady leak of power to AI, none of them dramatic on their own, together enough to keep the network’s hardest number pinned below its high for more than a year.
Hashprice and the miner squeeze
The metric that captures the squeeze is hashprice, the daily revenue a miner earns per unit of hashrate. It bundles the price of Bitcoin, the block subsidy, transaction fees, and difficulty into a single dollar figure per petahash per second per day. Hashprice sat near $33.26 on 2 November, when Bitcoin was still around $77,000, according to Million Miner’s November update. Work the arithmetic at $64,000 and it drops to roughly $30 per petahash per day: the network issues about 450 BTC a day in subsidy, which at the current price is close to $29 million spread across nearly a zettahash of capacity. That puts hashprice within sight of the record low near $27 set in June. The 9 percent difficulty cut helps a little, since survivors split the reward among fewer machines, but it cannot offset a price that has fallen this far.
The revenue picture follows. October was actually the industry’s best month since late 2025, with miners collecting about $1.215 billion as the autumn rally lifted prices. November will come in sharply lower. The network’s annualized security budget, the total dollars paid to miners to keep attacking the chain unprofitable, works out near $10.5 billion at $64,000, down from roughly $14 billion when Bitcoin traded in the mid-$80,000s. That budget is funded almost entirely by the block subsidy today, which is why the long-run question for mining is whether transaction fees can replace it after future halvings. For now, the squeeze is cyclical, and it falls hardest on operators with old machines and expensive power. The staking side of the ledger faces its own version of this yield compression, which is why comparing the returns you actually keep across liquid-staking providers has become its own cottage industry.
What kind of decline is this?
Not all hashrate declines are the same, and sorting them is the fastest way to understand what November means. There are three classic varieties. The first is price capitulation, seen in 2018 and 2022, when a falling price pushed miners below break-even and they switched off. The second is coercion, seen in 2021, when China banned mining and roughly half the network went dark within weeks regardless of profitability. The third is the one that defined early 2026: a voluntary migration, as operators chose to redirect power toward AI and high-performance computing even while mining remained marginally profitable and the network sat near its record.
Mid-November is a price capitulation stacked on top of that voluntary migration. The old-fashioned, cyclical kind of decline has returned, with the price breaking through the cost floor, while the newer structural drain toward AI never paused. Both the price engine and the capital engine are now pulling the hashrate in the same direction. That combination is what makes this reversal more than a routine dip. It is the first time the two forces that have defined 2026, a weak price and a permanent reallocation of power, have aligned to push the number down together rather than fighting each other to a stall.
Efficiency keeps falling even as the fleet shrinks
Here is the counterintuitive part. Even as machines power down, the ones that stay on are dramatically more efficient than the ones they replaced. Efficiency is measured in joules per terahash, the energy a rig burns to produce a trillion hashes, and lower is better. The 2016-era Antminer S9 drew about 98 joules per terahash. The current flagship hydro-cooled units land near 9.5, with Bitmain’s S23 Hydro and Bitdeer’s SealMiner A4 Ultra both crossing below ten joules per terahash for the first time in mass production. That is roughly a tenfold efficiency gain in a decade.
| ASIC (year) | Efficiency (J/TH) | Hashrate per unit |
|---|---|---|
| Antminer S9 (2016) | about 98 | 13.5 TH/s |
| Antminer S19 (2020) | about 34 | 95 TH/s |
| Antminer S19 XP (2022) | about 21.5 | 140 TH/s |
| Antminer S21 XP (2024) | about 13.5 | 270 TH/s |
| Antminer S23 Hydro (2026) | about 9.5 | 580 TH/s |
| SealMiner A4 Ultra (2026) | about 9.45 | 886 TH/s |
This efficiency curve is why a difficulty cut does not simply shrink the network; it reshuffles it. When difficulty falls, the efficient machines that stayed online capture the share vacated by the high-cost rigs that switched off, exactly the rebalancing JPMorgan described as a floor that stops the decline from spiraling. It is also why today’s newest hardware tends to replace aging fleets rather than expand the network, with sub-ten-joule machines increasingly used to upgrade existing sites instead of adding net-new capacity. A flat or falling hashrate can therefore hide a complete churn of the machines underneath it, with the most efficient silicon quietly taking over behind a single aggregate number.
Does a falling hashrate weaken Bitcoin’s security?
The honest answer is: at the margin, yes, but far less than the headlines imply. Bitcoin’s security rests on the cost of acquiring a majority of the hashrate. A lower network hashrate means fewer machines to out-compute, and a lower Bitcoin price means the dollar cost of buying or renting that power falls too. On paper, a downturn makes an attack cheaper on both axes at once. In practice, the number is still enormous. Campbell Harvey, the Duke University finance professor, has modeled a one-week majority attack at roughly $8 billion, a figure he frames as “about 50 basis points of the value of bitcoin,” as reported by Bitcoin.com News. His newer twist is that the attack could in principle pay for itself if the attacker shorted Bitcoin first, since, in his words, “the difference today is the derivatives markets.”
Even so, most practitioners dismiss the scenario as economically self-defeating. Matt Prusak, an executive at American Bitcoin, put it bluntly to the same reporting that economic feasibility kills the 51 percent thesis: acquiring that much hardware would take years, exchanges would freeze suspicious activity, and the attack would destroy the value of the very asset it targeted. Bitcoin has never suffered a successful 51 percent attack in sixteen years, even as smaller proof-of-work coins like Ethereum Classic and Bitcoin Gold have been hit repeatedly. The self-healing difficulty adjustment is part of why: as high-cost miners leave, the survivors earn more per unit of work, which stabilizes the floor and pulls hashrate back once price recovers. A proof-of-stake network defends itself differently, pricing security in locked capital and slashing rather than electricity and silicon, a contrast explored in our look at validator economics.
The AI drain that didn’t pause
Underneath the cyclical crash runs a structural story that will outlast it. Over the past two years, the largest listed miners have discovered that leasing their power and data centers to artificial-intelligence and high-performance-computing tenants pays far more than mining Bitcoin with it. James Butterfill, head of research at CoinShares, has called this one of the most challenging periods the mining industry has faced, projecting that up to 70 percent of listed-miner revenue could come from AI by the end of 2026, per Bitcoin.com News. The economics are stark: a megawatt leased to AI can generate several times the revenue of a megawatt pointed at ASICs, which is why multibillion-dollar lease deals with cloud and model-training customers have become routine across the sector.
This matters for hashrate growth because it changes the meaning of a downturn. In past cycles, miners that powered down during a price slump plugged back in when the price recovered, and the hashrate snapped back. Now, some of that idled power does not come back to mining at all; it is contracted to AI tenants on multi-year terms. The same surge in demand for compute that is reshaping projects like decentralized GPU marketplaces is competing directly for the cheap, firm power that Bitcoin miners spent a decade securing. So even a strong price recovery in 2027 may not restore the hashrate to its old trajectory. The capital engine has a permanent leak, and the November reversal is widening it.
How the SEC sees mining
For readers in the United States, the regulatory picture is comparatively settled. In March 2025, the SEC’s Division of Corporation Finance issued a staff statement concluding that proof-of-work mining, including both solo mining and participation in mining pools, does not involve the offer or sale of securities. In the staff’s view, miners do not satisfy the Howey test because their rewards come from their own computational work rather than the efforts of others. Mining itself, in other words, is not a security, and the more than a third of global hashrate based in the United States operates outside the registration regime that governs token offerings.
The SEC’s real interest in mining runs through the public miners’ equities, their disclosures, and the spot Bitcoin exchange-traded funds whose value tracks the asset miners produce. What moves the price, and therefore the hashrate, is macro policy, not mining rules: interest rates, the dollar, and the risk appetite that drains out of crypto when the Federal Reserve leans hawkish. That calendar is the subject of the lame-duck rulebook race into January, and it will shape whether Bitcoin reclaims its cost floor far more than any statement about Howey ever could.
What to watch into December
The reversal is a process, not an event, and a short watchlist tells you where it is headed. The first marker is the late-November difficulty retarget: if it prints near the estimated minus 7 percent, the network will have given back its entire autumn recovery and difficulty will sit near its June low. The second is the price relative to the cost of production. As long as Bitcoin trades below roughly $75,000 to $78,000, the pressure to power down persists and further cuts are likely. A clean move back above that zone would flip the thermostat again.
The third marker is capitulation behavior: watch for public miners selling larger shares of their production to cover costs, and for the hashrate to carve out a clear trough the way it did near 850 exahashes per second in February. The fourth is the AI pipeline, which could accelerate in a downturn as struggling miners sign leases to survive, permanently removing power from the network. And the fifth is the macro calendar into December, where rate expectations and liquidity will decide the price. Whatever happens, the thermostat will keep adjusting every 2,016 blocks, indifferent to the narrative, quietly keeping blocks arriving about every ten minutes while the humans argue about what it all means.
Frequently Asked Questions
Why is Bitcoin’s hashrate falling in November 2026?
Bitcoin’s price fell below the estimated cost of producing a coin, which is near $78,000 according to JPMorgan. When mining becomes unprofitable for high-cost operators, they power down their machines, so the total network hashrate declines. The protocol then lowered difficulty by about 9 percent on 13 November to keep blocks arriving near the ten-minute target.
What was Bitcoin’s 13 November 2026 difficulty adjustment?
Difficulty fell about 9.1 percent, from 148.36 trillion to 134.82 trillion, at block 975,744. It was the largest downward adjustment since June and followed two weeks of slow blocks averaging close to 10 minutes and 48 seconds, a sign that hashrate had dropped.
Does a lower hashrate make Bitcoin less secure?
Only at the margin. A majority attack on Bitcoin is still estimated to cost billions of dollars and has never succeeded in sixteen years. The difficulty adjustment is self-healing, because as high-cost miners leave, the survivors earn more per unit of work, which stabilizes the network and pulls hashrate back once the price recovers.
What is Bitcoin’s production cost and why does it matter?
Production cost is the all-in dollar cost of mining one Bitcoin, including electricity, hardware, and overhead. JPMorgan estimates it near $78,000 and CoinShares puts the cash cost near $75,500. When the market price trades below that level, as it did in November 2026 at around $64,000, high-cost miners lose money and switch off, which pulls the hashrate down.
Will Bitcoin’s hashrate grow again?
History strongly suggests it will. The hashrate has recovered from every previous decline, including the 2021 China mining ban and the 2022 crash. Growth tends to resume once the price climbs back above the cost of production or more efficient machines replace older ones, though the ongoing shift of power toward AI could slow the pace of the next expansion.
Marcus Okafor covers Bitcoin mining, energy, and network security for HOGE Wire.