Bitcoin mining is experiencing its most severe profitability crisis since China's 2021 ban. The industry-average all-in cost to produce one bitcoin has reached approximately $87,000, according to Checkonchain data, while spot BTC has spent most of February 2026 trading between $60,000 and $68,000...
"When power becomes scarce, the highest-value use of a megawatt tends to win." — BlackRock, 2026 Global Outlook
Bitcoin mining is experiencing its most severe profitability crisis since China's 2021 ban. The industry-average all-in cost to produce one bitcoin has reached approximately $87,000, according to Checkonchain data, while spot BTC has spent most of February 2026 trading between $60,000 and $68,000 — a 20-30% deficit below breakeven. Hash ribbons have been flashing capitulation for three months, publicly traded miners sold 8,200 BTC in January alone (the highest monthly total since the FTX collapse), and the network experienced its largest difficulty drop since China's mining exodus.
But this is not merely a cyclical squeeze. A structural transformation is underway. AI data centers are outbidding miners for cheap electricity, BlackRock has warned that AI could consume 24% of U.S. electricity by 2030, and the most forward-looking mining companies are pivoting from hashing to hosting. The question is no longer whether Bitcoin mining will survive — it's whether the industry that emerges on the other side will still be recognizable as mining at all.
The math is unforgiving. Checkonchain's industry-wide all-in production cost estimate sits at roughly $87,000 per BTC. JPMorgan's crypto research team puts the figure slightly lower at $77,000, reflecting a brief period of difficulty relief before the latest spike. Either way, with Bitcoin trading in the $60,000–$68,000 range for most of February 2026, the average miner is operating 15–25% underwater.
The damage is visible across every metric that matters:
The result is a two-speed industry. Efficient operators mining at $34,000–$43,000 per BTC (with access to power at $0.04–$0.06/kWh) maintain margins of up to 71%. Everyone else is bleeding.
MARA Holdings (formerly Marathon Digital), the industry's largest publicly traded miner, disclosed a devastating $1.71 billion net loss for Q4 2025 — a stark reversal from the $528 million net profit recorded in Q4 2024. For the full year, MARA posted a $1.31 billion annual net loss, erasing the $541 million profit achieved in 2024. The stock has fallen to $8, its lowest level since 2023.
TeraWulf missed Q4 estimates by a wide margin, reporting a loss of $1.66 per share against analyst expectations of $0.13, on revenue of $35.8 million. Mining revenue specifically dropped to $26.1 million, with $9.7 million coming from the company's nascent HPC (high-performance computing) business.
Publicly traded miners collectively sold approximately 8,200 BTC in January 2026 — the highest monthly sell-off since the FTX-driven capitulation in November 2022. This forced selling creates a self-reinforcing cycle: miners dump BTC to cover operations, pushing prices lower, which further compresses margins, triggering more selling.
Analysts have warned that if Bitcoin drops another 10–15% from current levels, multiple major miners could face bankruptcy, potentially triggering a cascade of forced liquidations that would accelerate the downward spiral.
This cycle's mining crisis has a structural dimension that previous ones did not: competition from AI for the same scarce resource — cheap, reliable electricity.
BlackRock's 2026 Global Outlook delivered a stark assessment. AI-driven data centers could consume as much as 24% of U.S. electricity by 2030. The report argued that the AI buildout is "pushing against physical limits" and identified electricity as "the constraint investors are underpricing." For Bitcoin miners, the implication is existential: when power markets tighten, the highest-value use of a megawatt wins.
The data supports the warning. In Texas — the epicenter of North American Bitcoin mining — ERCOT's interconnection queue for large loads has surged to 226 gigawatts, with the overwhelming majority of applications coming from AI and high-performance computing projects. These facilities can pay more for power and carry stronger political backing. AI infrastructure is framed as essential to national competitiveness, defense, and productivity. Bitcoin mining is more easily characterized as optional.
The U.S. Energy Information Administration estimated that Bitcoin mining consumed 0.6%–2.3% of U.S. electricity in 2024. That share is under pressure from both directions: miners being priced out of existing contracts, and new power capacity being allocated to AI before miners can bid.
This is not a temporary inconvenience. It is a permanent reshaping of the competitive landscape for energy-intensive computation.
The most significant strategic response is the industry-wide pivot from pure-play mining to AI/HPC infrastructure hosting. This is the defining industrial story of 2025–2026, and the market has moved past announcements into execution.
MARA Holdings launched a partnership with Starwood Capital Group to convert existing mining sites into next-generation digital infrastructure targeting enterprise, hyperscale, and AI customers, with a target of 2.5 GW of capacity.
TeraWulf secured $12.8 billion in contracted long-term revenue across 522 MW of HPC capacity, with Google credit backing. The company enters 2026 with a 2.9 GW multi-regional platform designed specifically for long-term AI expansion.
Hut8 signed a deal to build an AI data center in Louisiana for Anthropic and a separate $7 billion lease agreement with Fluidstack, backed by Google. The company is expected to post Q4 revenue of $95.6 million — more than double the same quarter last year — largely on the strength of its hosting business.
The logic is straightforward: mining companies already own land, power purchase agreements, substation access, and cooling infrastructure. These are precisely the assets AI developers need most and cannot build fast enough on their own. Public Bitcoin miners collectively plan to expand to 30 gigawatts of AI-focused power capacity — nearly triple current levels.
But the pivot is harder than it appears. AI data centers require different cooling architectures, different network topologies, and critically different uptime guarantees. Mining can tolerate interruptions; enterprise AI customers cannot. The cost of retrofitting is enormous, and the competition includes specialist data center operators like Equinix and Digital Realty with decades of institutional relationships.
The crisis is accelerating a Darwinian consolidation. The industry is splitting into three tiers:
Tier 1: AI-Infrastructure Hybrids — Companies like TeraWulf, Hut8, and IREN that have successfully executed AI hosting contracts are thriving during the crypto crash. Their stock prices are rising even as Bitcoin falls, because Wall Street is pricing them as energy infrastructure companies, not crypto bets. TeraWulf's $12.8 billion in contracted HPC revenue dwarfs its mining income and provides the kind of cash flow visibility that institutional investors demand.
Tier 2: Efficient Pure-Play Miners — Operations running next-generation ASICs at $0.04/kWh or below, with all-in production costs of $34,000–$43,000 per BTC. These miners remain profitable but face the strategic question of whether to stay the course or follow the herd into AI hosting. Their competitive moat is narrow: energy contracts can expire, hardware depreciates, and the next difficulty adjustment is always coming.
Tier 3: The Walking Dead — Mid-tier and smaller operations running older hardware at average power costs. With production costs near or above $87,000 and no AI pivot in progress, these companies face a binary outcome: find a buyer, or shut down. The difficulty drop earlier in February offered temporary relief, but the subsequent 14.73% difficulty spike to 144.4 trillion eliminated any breathing room.
Morgan Stanley has begun covering the sector through an "infrastructure" lens rather than a "crypto" lens, viewing surviving miners as critical energy infrastructure assets needed to fuel the AI boom.
The Hash Ribbon indicator — which tracks the relationship between the 30-day and 60-day moving averages of Bitcoin's hashrate — has been signaling capitulation since late November 2025. During that period, Bitcoin fell from approximately $90,000 to a low near $60,000 in early February before rebounding to roughly $65,000.
As of late February, the indicator is approaching a recovery crossover. Since 2011, there have been approximately 20 mining capitulations, and most have coincided with local or major Bitcoin price bottoms — including January 2015, December 2018, and December 2022.
CoinDesk reported on February 25 that "one of the longest mining capitulations nears end, signaling potential BTC price bottom." The hashrate itself is rebounding, indicating that surviving miners are coming back online and that network stress is easing.
Historically, Bitcoin miners facing a margin crunch of this severity have preceded strong returns within 90 days. If the pattern holds, the current crisis may be setting the stage for a recovery — but for many miners, that recovery will come too late.
Bitcoin mining's February 2026 crisis is both familiar and unprecedented. The cyclical pattern — price drops, margins compress, weak miners capitulate, difficulty adjusts, survivors consolidate — has played out before. What makes this cycle different is the structural overlay: AI's insatiable appetite for electricity is permanently altering the economics of energy-intensive computation, and Bitcoin mining no longer operates in a world where cheap power is a given.
The companies that survive will not be the ones with the most hashrate. They will be the ones that understood earliest that owning a megawatt is more valuable than owning a petahash. The pivot from mining to infrastructure hosting is not a retreat — it is a recognition that the economic value in the Bitcoin mining stack is migrating from computation to energy access.
For Bitcoin itself, the hash ribbon recovery signal offers a familiar source of optimism. But for the mining industry as an industry, the message is stark: adapt to the AI energy economy, or become its casualty.