difficulty falls 19.9% as miners pivot to AI

Bitcoin mining difficulty has dropped 19.9% from its peak in the third deepest ASIC era decline on record, as miners sell bitcoin at record rates and redirect power capacity toward artificial intelligence data centers.

Bitcoin mining difficulty has dropped 19.9% from its all time peak. That single number captures a transformation that has been building for months but accelerated through the first half of 2026: the economics of mining bitcoin have deteriorated to the point where a meaningful share of the global fleet has shut down, and the operators that remain are increasingly looking beyond bitcoin for revenue.

The decline, tracked by Bitcoin Magazine Pro from the November 2025 peak of roughly 156 trillion to 126.23 trillion as of the July 25 adjustment, ranks as the third deepest drawdown since application specific integrated circuits became the standard mining hardware. Only the aftermath of China’s 2021 mining ban and a 2018 bear market contraction produced deeper declines. But unlike those episodes, this one has no single policy catalyst. It is the compound result of a lower bitcoin price, rising energy costs, post halving revenue compression, and a structural shift in how mining companies view their own business.

The capitulation is visible across every metric: hashrate, difficulty, miner selling, and hashprice. What makes this cycle different is what comes next. The miners who survive are not simply waiting for higher bitcoin prices. They are converting their facilities into AI data centers.

How difficulty measures mining health

Bitcoin’s difficulty adjustment is one of the protocol’s most elegant mechanisms. Every 2,016 blocks, roughly every two weeks, the network recalculates how hard it is to mine a new block. If blocks arrived faster than one every ten minutes during the previous epoch, difficulty increases. If they arrived slower, difficulty decreases. The system exists to keep block production steady regardless of how much computing power is pointed at the network.

When difficulty falls, it means hashrate has left the network. Miners have switched off machines, either because their operating costs exceed their revenue or because they have found more profitable uses for their power capacity. A falling difficulty makes mining easier for the operators who remain, temporarily improving their economics until the incentive draws hashrate back.

The current 19.9% decline from peak is notable for both its depth and duration. Bitcoin Magazine Pro’s data shows the downward trend extending approximately 287 days, making it one of the longest sustained mining contractions in bitcoin’s history. The July 25 adjustment of negative 0.74% was the ninth downward adjustment of 2026. The previous major drop in June was 10.09%, which ranked as bitcoin’s 11th largest single downward adjustment ever, reducing difficulty from 138.96 trillion to 124.93 trillion.

Difficulty has also turned negative on a year over year basis for only the second time in bitcoin’s history. The previous instance followed China’s 2021 mining ban, when authorities forced an estimated 50% of global hashrate offline in a matter of weeks. That comparison is instructive: the current decline has reached similar severity without any government ban, driven entirely by market forces.

The economics behind the shutdown

The fundamental problem is arithmetic. After the April 2024 halving, miners receive 3.125 $BTC per block, half what they earned before. That reduction was expected. What was not expected was that bitcoin’s price would fail to compensate.

Bitcoin traded near $63,100 on July 31, down approximately 47% over 12 months and nearly 50% below its October 2025 record. For miners, this price decline arrives on top of the halving’s structural revenue cut. The combined effect has been devastating for operators running older hardware or paying higher electricity rates.

The math is stark. Before the halving, a miner producing one block earned 6.25 $BTC. At bitcoin’s October 2025 peak near $120,000, that block was worth $750,000. Today, the same miner earns 3.125 $BTC per block at a price near $63,100, yielding approximately $197,000. That is a 74% decline in per block dollar revenue in less than a year. No industry can absorb that kind of revenue compression without significant operational fallout.

Transaction fees, which historically provide a secondary revenue stream for miners, have not offset the decline. Fee revenue as a percentage of total mining revenue has remained in the low single digits through most of 2026, well below the spikes that accompanied the inscription boom in late 2023 and early 2024. The fee market has normalized, removing what had briefly appeared to be a structural supplement to block rewards.

Hashprice, which measures the expected daily revenue from one petahash of computing power, stood near $32 per PH/s per day in late July. That figure sits below the breakeven threshold for many operations. CoinShares estimated in March 2026 that 15% to 20% of the global mining fleet was operating at a loss. Older machines, including models from the Antminer S19 generation, cannot generate positive cash flow at current prices unless operators have electricity costs below approximately five cents per kilowatt hour.

The result is a fleet rationalization. Miners with newer hardware, primarily the Antminer S21 and comparable models, continue to operate profitably at current prices. Miners with older hardware and higher power costs are shutting down, selling their bitcoin reserves, or converting their facilities to other uses. The 12% decline in hashrate from the late 2025 peak of over one zettahash per second to approximately 868 EH/s by late July reflects this ongoing culling.

Record bitcoin sales by miners

The selling pressure from mining companies has been extraordinary. Publicly traded miners sold more than 32,000 $BTC in the first quarter of 2026, a single quarter record that exceeded their combined sales for all of 2025. The total also surpassed the roughly 20,000 $BTC sold during Q2 2022, when the Terra Luna collapse sent bitcoin below $20,000.

The individual disclosures paint a clear picture of the pressure. Riot Platforms sold 3,778 $BTC in Q1 at an average price near $76,626, generating approximately $289.5 million, while producing only 1,473 coins in the same period. Core Scientific liquidated roughly 1,900 $BTC worth about $175 million in January alone. Cango sold 2,000 $BTC in March for approximately $143 million, using proceeds to retire bitcoin backed loans.

In a single week during Q1, MARA, Genius Group, and Nakamoto Holdings revealed combined sales of more than 15,000 coins. These were not routine sales of freshly mined production to cover electricity bills. They were drawdowns of treasury reserves that companies had previously chosen to hold.

The aggregate miner reserve, the total bitcoin held by mining companies, has been declining since 2023. It fell from more than 1.86 million $BTC at the end of that year toward roughly 1.8 million by mid 2026. The sustained drawdown suggests that this is not opportunistic selling but a structural shift in how mining companies manage their balance sheets.

The selling also reflects the debt burden that many miners accumulated during the 2024 and early 2025 expansion cycle. Companies borrowed against their bitcoin holdings and future production to finance fleet upgrades and facility construction. As bitcoin’s price fell and revenue declined, those loans required either refinancing at unfavorable terms or liquidation of the bitcoin collateral. Cango’s March sale of 2,000 $BTC was explicitly used to retire bitcoin backed loans, a pattern that has repeated across the industry.

The irony is that miner selling itself contributes to the price pressure that makes mining less profitable. When miners sell tens of thousands of bitcoin into the market over a single quarter, they add supply at a time when demand is already weakened by broader market conditions. The selling becomes self reinforcing: lower prices lead to more selling, which pushes prices lower, which forces more machines offline, which triggers more selling of treasury reserves to cover fixed costs.

The AI pivot

The most significant development in the mining industry is not about bitcoin at all. It is about artificial intelligence.

Mining companies operate large scale power infrastructure in locations with grid access, cooling capacity, and favorable energy contracts. Those same characteristics are exactly what AI data center operators need. The realization has transformed the investment thesis for publicly traded miners, turning them from pure bitcoin proxies into energy infrastructure companies.

Hut 8 provides the most dramatic example. The company signed a second 15 year lease on July 20 for 352 megawatts at its Beacon Point campus in Texas. The agreement raised the campus’s base term contract value to $19.6 billion and Hut 8’s total contracted AI portfolio to $26.6 billion. Initial delivery for the second phase is scheduled for Q2 2028. Hut 8’s shares more than quadrupled over the preceding 12 months and rose 11% after the announcement.

Core Scientific followed on July 28 with an AMD partnership anchored by 15 year agreements covering approximately 530 MW. The company said its total leased customer capacity had reached roughly 1.1 GW, representing more than $24 billion in potential contracted revenue.

TeraWulf’s transition is already generating revenue. The company reported $21 million in AI and high performance computing hosting revenue in Q1 2026, surpassing its bitcoin mining revenue of less than $13 million for the first time. HIVE Digital announced a $2.55 billion AI super factory project near Toronto designed to host more than 100,000 GPUs.

The scale of these AI commitments dwarfs the bitcoin mining operations they are displacing. Hut 8’s $26.6 billion in contracted AI revenue over 15 years exceeds what the company could plausibly earn from bitcoin mining over the same period at current prices and difficulty levels.

The pivot is not limited to North America. Mining operators in the Nordics, the Middle East, and parts of Central Asia are exploring similar conversions, attracted by the same logic: AI workloads pay more per megawatt hour than bitcoin mining and provide contractual revenue certainty that bitcoin mining cannot offer. A 15 year lease agreement with a hyperscaler eliminates the price volatility, halving risk, and difficulty uncertainty that define the bitcoin mining business.

The infrastructure requirements are different, however. AI data centers need higher power density, better cooling, more reliable uptime guarantees, and enterprise grade networking that most mining facilities were not built to provide. The conversion from mining to AI hosting requires significant capital expenditure, which is part of why miners are selling bitcoin reserves and issuing equity. The transition is not free, and companies that underestimate the engineering and capital requirements may find themselves stuck between a declining mining business and an AI hosting business that is not yet ready to generate revenue.

Why mining stocks diverged from bitcoin

The AI pivot has broken the historical relationship between mining stocks and bitcoin’s price. A basket of bitcoin mining equities gained 56% during the early months of 2026 while bitcoin fell 17%, according to research cited by industry analysts. That divergence would have been unthinkable two years ago, when mining stocks moved in lockstep with bitcoin’s price, only with greater amplitude.

Investors are now valuing these companies on their power contracts, real estate, and AI revenue potential, not on their bitcoin production. The market is pricing in a future where bitcoin mining is a secondary revenue stream for companies whose primary business is providing power and infrastructure for artificial intelligence workloads.

This creates an ironic dynamic for bitcoin’s network security. The same companies that built the infrastructure securing the bitcoin network are now economically incentivized to redirect that infrastructure toward AI. Every megawatt that moves from mining to AI hosting reduces the hashrate protecting bitcoin’s blockchain. The difficulty adjustment compensates for the loss automatically, but the trend raises questions about the long term security implications if mining becomes a marginal activity for what were once dedicated mining companies.

The counterargument is that the AI revenue stream makes these companies more financially resilient, which ultimately benefits the bitcoin network. A mining company with $26 billion in contracted AI revenue can afford to keep mining bitcoin through price downturns that would force a pure play miner to shut down entirely. The AI business subsidizes the mining operation.

The historical parallel is not perfect, but it is instructive. After the 2021 China ban, difficulty dropped more than 50% before recovering within months as displaced miners relocated and reconnected. That episode proved that bitcoin’s difficulty adjustment mechanism works as designed: when enough hashrate leaves, difficulty falls until mining becomes profitable again for the remaining operators, creating an economic incentive for hashrate to return. The current episode tests whether the same self correcting mechanism applies when the departure of hashrate is driven not by a ban but by a better economic opportunity. Miners who leave for AI may not return even if bitcoin prices recover, because the AI revenue exceeds what bitcoin mining can offer.

What capitulation historically signals

Miner capitulation has historically preceded bitcoin price recoveries. The logic is straightforward: when the weakest miners shut down and sell their reserves, the selling pressure eventually exhausts itself. Difficulty falls, making mining cheaper for survivors. The supply of newly mined bitcoin continues at a fixed rate regardless of hashrate, but the forced selling from distressed operators slows as those operators exit the market.

The 2022 capitulation followed this pattern. Miners sold aggressively through Q2 and Q3, difficulty fell, and by early 2023, bitcoin had begun a sustained recovery that eventually carried prices to new all time highs. Proponents of the capitulation thesis argue that the current period will resolve similarly: the pain is intense but temporary, and the difficulty adjustment ensures that mining always returns to profitability for the marginal operator.

The structural difference this time is the AI alternative. In previous cycles, sidelined mining capacity had no productive alternative use. It simply sat idle until bitcoin prices made mining profitable again. Today, that capacity has a buyer willing to pay more, which means the recovery mechanism may not function as cleanly as it has in the past.

What to watch

• The next difficulty adjustment. Whether difficulty continues to fall or stabilizes will signal whether the current round of miner shutdowns has run its course. A sustained difficulty increase would indicate that surviving miners are expanding or that sidelined operators are reconnecting.

• Q2 miner selling data. The 32,000 $BTC sold in Q1 set a record. Whether Q2 selling accelerated, stabilized, or declined will indicate the severity of the remaining financial pressure on listed operators.

• Bitcoin price relative to production cost. Some analysts estimate the average production cost for the global mining fleet near $80,000. Bitcoin trading at approximately $63,100 means a significant portion of miners are operating below cost. A price recovery above $80,000 would alleviate much of the current pressure.

• AI data center construction timelines. The announced deals from Hut 8, Core Scientific, and others involve multi year construction timelines. Whether these projects proceed on schedule and begin generating revenue will determine whether the AI pivot delivers on its promise.

• Regulatory treatment of dual use facilities. Mining companies that operate both bitcoin mining and AI hosting from the same campuses may face different regulatory frameworks for each activity. How jurisdictions classify and regulate these hybrid operations could affect the economics of the pivot.

Frequently asked questions

Disclaimer: This article is for informational purposes only and does not constitute financial or investment advice. The information presented reflects publicly available data as of August 1, 2026. Readers should conduct their own research and consult qualified professionals before making financial decisions.