Bitcoin miners are facing their most severe profitability crisis since the 2022 bear market — and this time, many are choosing not to wait it out. With Bitcoin trading at approximately $67,000 while the average all-in production cost sits near $87,000, the entire publicly listed mining sector is ...
"The decision to dispose of our BTC holdings should not be viewed as a bearish signal for the broader market. We are actively evaluating several powered land acquisitions and consider it prudent to lock in liquidity now." — Bitdeer Technologies, Corporate Statement on Treasury Liquidation (February 2026)
Bitcoin miners are facing their most severe profitability crisis since the 2022 bear market — and this time, many are choosing not to wait it out. With Bitcoin trading at approximately $67,000 while the average all-in production cost sits near $87,000, the entire publicly listed mining sector is operating at a loss. The Hash Ribbon capitulation indicator has been flashing for three consecutive months, one of the longest sustained capitulation events on record.
But unlike previous cycles where miners simply shut down machines and waited for prices to recover, the 2026 crisis has triggered something structurally different: a mass exodus from Bitcoin mining into artificial intelligence infrastructure. At least eight publicly listed mining companies have announced partial or full pivots toward AI data center operations in the past 18 months. Bitdeer has liquidated its entire 1,127 BTC treasury to fund AI expansion. Bitfarms has rebranded itself as "Keel Infrastructure" and dropped "Bitcoin" from its identity entirely. MARA Holdings has signed a deal with Starwood to convert mining sites into AI data centers targeting 2.5 gigawatts of compute capacity.
This report examines the economics driving this pivot, the implications for Bitcoin's security model, and what it means when the companies that secure a $1.3 trillion network decide the network itself isn't worth securing.
The numbers are stark. As of late February 2026, Bitcoin trades at roughly $67,000 — approximately 20% below the estimated average all-in production cost of $87,000 per coin. This gap represents the widest sustained margin compression since the post-FTX collapse in late 2022.
The crisis is multidimensional:
The February crash amplified these pressures. Bitcoin fell from $90,000 in late November 2025 to below $61,000 on February 5, 2026 — a 6.05σ move on the rate-of-change Z-score, placing it among the fastest single-day crashes in crypto history. Futures open interest collapsed from $61 billion to $49 billion in days, with $3-4 billion in total liquidations.
For miners, this wasn't just a price drop. It was the moment the math broke.
The network's response was immediate and dramatic. Bitcoin's hashrate plunged from its October 2025 all-time high of 1.1 ZH/s (zettahash per second) to 826 EH/s — a 25% decline driven by the dual shock of collapsing prices and widespread winter storm outages across U.S. mining operations.
The resulting difficulty adjustment told the story: an 11.16% downward correction, the steepest single drop since China's mining ban in 2021. Block times spiked to 20 minutes — double the protocol target — as unprofitable machines went dark across North America.
What followed was equally extraordinary. Difficulty subsequently surged 15% to a record 144.4 trillion, the largest percentage increase since 2021, as surviving miners — predominantly large-scale operators with sub-$0.04/kWh electricity — aggressively expanded their share. The hashrate recovered to approximately 1.13 ZH/s by late February.
The Hash Ribbon indicator, which tracks the relationship between 30-day and 60-day hashrate moving averages, has been in capitulation mode since late November 2025 — a three-month duration that ranks among the longest on record. Historically, the end of Hash Ribbon capitulations (there have been approximately 20 since 2011) has coincided with significant price bottoms, including January 2015, December 2018, and December 2022. As of late February 2026, analysts see early signs the capitulation is nearing its end, with the 30-day moving average beginning to approach the 60-day from below.
But the structural damage may already be done. The miners who turned off aren't simply waiting to turn back on. Many are converting their facilities to serve an entirely different customer.
The pivot from Bitcoin mining to AI infrastructure is no longer an experiment conducted by a few forward-thinking firms. It has become an industry-wide strategic realignment, and the pace is accelerating.
Bitdeer Technologies represents the most dramatic case. In February 2026, the company liquidated its entire Bitcoin treasury — 1,127 BTC including the 189.8 BTC produced that week — zeroing out its crypto balance sheet. Simultaneously, it priced a $325 million convertible notes offering and a $43.5 million equity raise to fund data center expansion, HPC, and AI cloud growth. Bitdeer is no longer a Bitcoin company that dabbles in AI. It is an AI infrastructure company that happens to have once mined Bitcoin.
Bitfarms went further, rebranding as Keel Infrastructure and formally announcing its exit from Bitcoin mining. Its 6-acre, 18MW Washington facility is being converted to HPC/AI hosting, with completion expected by December 2026. The company has dropped "Bitcoin" from its corporate identity entirely.
MARA Holdings (formerly Marathon Digital) struck a deal with Starwood Digital Ventures to develop AI data centers, sending its stock up 17% on the announcement. The partnership targets approximately 1 gigawatt of near-term computing capacity, with plans to scale beyond 2.5 gigawatts. Select MARA mining locations will be converted to serve enterprise cloud and AI customers.
Core Scientific signed a 12-year, 200 MW deal with CoreWeave for AI workloads and plans over 300 MW of AI capacity by 2026.
Riot Platforms sold approximately $200 million worth of Bitcoin in November and December 2025, reducing its holdings to 18,005 BTC. Analysts estimate the proceeds roughly equal the entire capital expenditure required for the first phase of Riot's Corsicana AI data center build.
The economic logic is straightforward. Bitcoin mining revenue is cyclical, halvings are deflationary shocks that compress margins every four years, and hashprice is at multi-year lows. AI compute contracts, by contrast, offer predictable, long-term revenue streams. A 12-year hosting deal with CoreWeave provides more earnings visibility than four halving cycles of Bitcoin mining. The market has rewarded this logic: mining stocks that announced AI pivots consistently outperformed those that remained BTC-only.
But there is a deeper irony here. These companies spent years building precisely the infrastructure — access to cheap power, cooling systems, land, grid interconnections, permitting — that AI companies now desperately need. Bitcoin mining, in retrospect, may have been the loss leader that built America's AI data center network.
The pivot raises an uncomfortable question for the Bitcoin network: what happens to security when the companies that secure it decide the economics no longer work?
Bitcoin's security model depends on miners spending real resources — primarily electricity — to validate transactions and produce blocks. The network currently operates on roughly $18 billion in annual mining subsidies (block rewards at current prices) but generates only approximately $115 million in annual transaction fees. This means that 99.4% of Bitcoin's security budget comes from inflationary issuance, not organic demand for block space.
Every halving cuts this subsidy in half. The April 2024 halving reduced the block reward from 6.25 to 3.125 BTC. By 2028, it drops to 1.5625 BTC. The long-term security model assumes transaction fees will eventually replace subsidies — but fees have shown no structural growth trajectory capable of filling a gap that widens with each halving.
The current crisis offers a preview. When Bitcoin's price dropped below production cost, 25% of the network's hashrate disappeared within weeks. The miners who left aren't marginal hobbyists — they're publicly listed companies with access to capital markets. Their departure to AI isn't temporary; it's a permanent reallocation of infrastructure.
If hashrate becomes structurally dependent on a smaller number of mega-miners with access to the cheapest electricity, Bitcoin's decentralization properties weaken. The top four publicly listed miners — Marathon, Riot, CleanSpark, and Hut 8 — already control 82.1% of the 115,335 BTC held across all public mining companies. Concentration of hashrate follows concentration of capital.
The mining companies that will remain in Bitcoin share common characteristics: sub-$0.04/kWh electricity costs, vertically integrated power assets (stranded gas, hydroelectric, nuclear), minimal debt, and diversified revenue through AI/HPC hosting that cross-subsidizes mining operations.
The pure-play Bitcoin miner — the company that mines BTC and only BTC — is becoming an endangered species. The survivors will be hybrid infrastructure firms that mine Bitcoin as one product line among several, treating hashrate as a swing load that absorbs excess power capacity between higher-margin AI workloads.
This is a fundamentally different security model than Satoshi envisioned. Bitcoin's security was supposed to be provided by a distributed network of miners incentivized by block rewards and fees. Increasingly, it will be provided by a handful of infrastructure conglomerates for whom Bitcoin mining is a side business — profitable only when the spread between electricity cost and BTC price is favorable, and abandoned the moment AI compute offers better returns.
The Bitcoin mining industry is undergoing its most significant structural transformation since the Chinese mining ban of 2021. But where China's ban was an external shock that redistributed hashrate geographically, the 2026 crisis is an internal economic reckoning that is redistributing hashrate functionally — from Bitcoin to AI.
The companies that built Bitcoin's physical security layer — the power contracts, the cooling systems, the grid interconnections — are discovering that this infrastructure has a more profitable tenant. The question is no longer whether Bitcoin mining is profitable at current prices. It is whether Bitcoin mining is the highest-value use of mining infrastructure at any price.
For now, the network continues to function. Difficulty adjusts. Blocks are produced. The protocol does what it was designed to do. But the composition of miners securing the network is changing in ways that compress decentralization, increase concentration risk, and accelerate the timeline on Bitcoin's unresolved security budget problem.
The miners aren't bearish on Bitcoin. They're rational about infrastructure. And that may be worse.