The Bitcoin mining industry is undergoing the most severe structural transformation in its history. With Bitcoin trading near $64,000–$68,000 — roughly 20–25% below the estimated average all-in production cost of $87,000 — the majority of miners are now operating at a loss. The Hash Ribbon capitu...
"We are no longer a Bitcoin company." — Ben Gagnon, CEO, Bitfarms (now Keel Infrastructure)
The Bitcoin mining industry is undergoing the most severe structural transformation in its history. With Bitcoin trading near $64,000–$68,000 — roughly 20–25% below the estimated average all-in production cost of $87,000 — the majority of miners are now operating at a loss. The Hash Ribbon capitulation indicator has been flashing since late January, the Miner Profit and Loss Sustainability Index has cratered to 21, and hashprice has fallen to a record low of $23.9 per petahash per second.
What makes this cycle different from previous miner capitulations is the emergence of a permanent exit door. Rather than simply shutting down and waiting for better prices, publicly traded mining companies are systematically converting their energy infrastructure into AI and high-performance computing (HPC) data centers. Bitdeer has liquidated its entire Bitcoin treasury. Bitfarms is rebranding as Keel Infrastructure and "winding down" mining entirely. Hut 8 has signed a $7 billion, 15-year lease with Google. The industry that was supposed to be Bitcoin's security backbone is becoming the backbone of the AI economy instead.
This report examines the convergence of three forces — a price-cost squeeze, a climate-driven stress test, and a structural pivot to AI — that are fundamentally reshaping Bitcoin mining's economic model and raising urgent questions about the network's long-term security budget.
The numbers are stark. According to Checkonchain data cited by CoinDesk, the average all-in production cost for a single Bitcoin — encompassing electricity, hardware depreciation, hosting, and overhead — sits at approximately $87,000. Bitcoin's spot price has spent most of February 2026 trading between $62,800 and $68,000, a sustained 20–27% deficit.
This is not a temporary dip below marginal cost. It is a structural squeeze driven by compounding forces:
Post-halving economics. The April 2024 halving cut the block subsidy from 6.25 BTC to 3.125 BTC, instantly doubling the cost per coin for every miner on the network. Transaction fees have not compensated. Revenue per exahash has fallen 35% since the halving, and ROI timelines for new ASIC deployments have stretched beyond 1,000 days.
Difficulty at all-time highs. On February 19, 2026, Bitcoin's mining difficulty spiked 14.73% to 144.4 trillion — the largest absolute increase in network history and the biggest percentage jump since China's 2021 mining ban. More hashrate competing for the same 3.125 BTC reward means thinner margins for everyone.
Hashprice collapse. Hashprice — the industry's standard measure of daily revenue per petahash — fell to approximately $23.9/PH/s, a multi-year low. The Miner Profit and Loss Sustainability Index dropped to 21, a level reflecting "total exhaustion of profit margins for all but the most efficient, low-cost operators."
The only miners still profitable are those running latest-generation S21-class ASICs with sub-$0.05/kWh power, producing Bitcoin at $34,000–$43,000 per coin. Everyone else — which means the majority of the global hashrate — is underwater.
The result is textbook capitulation. Miners are selling Bitcoin holdings to fund operations, cover energy bills, and service debt. The Hash Ribbon indicator — which compares the 30-day and 60-day moving averages of hashrate to detect financial stress — has been signaling capitulation since late January 2026.
On February 6–8, 2026, a severe winter storm swept across the southern and central United States, forcing widespread power curtailments. For Bitcoin, the impact was immediate and dramatic: approximately 200 exahashes per second of mining power disconnected from the network within 48 hours, cutting total hashrate by nearly 40% — from over 1 ZH/s to roughly 826 EH/s.
The damage was concentrated. Foundry USA, the largest Bitcoin mining pool by market share, saw a 60% drop in associated hashrate. Block times stretched to 20 minutes — double the 10-minute target — as the network struggled to process transactions.
Most shutdowns were technically voluntary. Many large U.S. miners participate in energy curtailment programs that pay them to release electricity back to the grid during demand spikes. But "voluntary" is a generous framing when the alternative is operating at a loss in a blizzard while selling power is the only profitable activity available.
The episode triggered a historic 11% downward difficulty adjustment on February 9 — the largest drop since the post-China-ban reset of 2021. When the storm passed and miners reconnected, hashrate surged back to 1 ZH/s, prompting the record 14.73% upward correction ten days later.
Beyond the technical volatility, the Great Hash Freeze exposed an uncomfortable truth: Bitcoin's hashrate is dangerously concentrated in U.S. geography. A single regional weather event temporarily compromised 40% of the network's computational power. IndexBox analysis published in the aftermath explicitly flagged this as a "centralization risk" — a term Bitcoin maximalists have historically reserved for criticizing proof-of-stake chains.
If the price-cost squeeze is pushing miners out, the AI opportunity is pulling them through a door they are unlikely to return from.
The economics are unambiguous. A megawatt of power dedicated to Bitcoin mining at current hashprices generates volatile, low-margin revenue with no contractual guarantee. The same megawatt leased to a hyperscaler for AI training generates 80–90% operating margins under a 10–15 year fixed-rate contract. For publicly traded companies accountable to shareholders, the choice is not difficult.
Bitdeer has been the most dramatic case study. The company liquidated its entire Bitcoin treasury — selling 1,132.9 BTC in a single week, including all 189.8 BTC mined during the period — to reach zero holdings as of February 20. It is now the largest publicly traded miner by self-mining hashrate to hold no Bitcoin on its balance sheet. The proceeds are funding powered land acquisitions and the rollout of NVIDIA GB200 NVL72 AI systems in Malaysia. Bitdeer recently raised $368.5 million through convertible notes and equity to fund datacenter expansion.
Bitfarms went further, announcing on February 6 that it is no longer a Bitcoin company. It will rebrand as Keel Infrastructure, relocate from Canada to the United States, wind down Bitcoin mining operations over two years, and focus exclusively on AI/HPC infrastructure. Its stock jumped 16% on the news. A shareholder vote is scheduled for March 20, with the transition expected to close April 1, 2026.
Hut 8 signed a 15-year, $7 billion lease for 245 MW of AI data center capacity at its River Bend campus in Louisiana, with Google providing a financial backstop. The deal could scale to 2,300 MW and $17.7 billion in total contract value. CEO Asher Genoot called it "the first domino to fall."
Riot Platforms sold $200 million worth of Bitcoin to fund operations and AI expansion. MARA Holdings is expanding into HPC through a planned 64% stake in France-based Exaion.
The scale is staggering. In 2025 alone, public Bitcoin miners signed over $65 billion worth of AI and HPC contracts with hyperscalers including Amazon and Microsoft. Industry analysts project that mining revenue will drop from 85% to under 20% of total revenue for these companies by late 2026.
This is not a hedge. It is an exodus.
Bitcoin's security model depends on miners being economically incentivized to dedicate computational resources to securing the network. The block subsidy — currently 3.125 BTC per block — is the primary incentive, supplemented by transaction fees. With the subsidy halving approximately every four years, Bitcoin's long-term security budget depends on transaction fees growing to compensate.
They are not growing. Fee revenue remains a small fraction of miner income, and the structural forces documented in this report are actively redirecting mining infrastructure away from Bitcoin.
This creates a slow-motion security concern. If the most well-capitalized, efficient mining operations continue converting megawatts from Bitcoin to AI, the network's hashrate will increasingly depend on smaller, less efficient operators — or on the price of Bitcoin rising high enough to make mining profitable again for industrial players. Neither outcome is guaranteed.
The Hash Ribbon signal suggests — based on historical precedent including the FTX collapse and the mid-2024 yen carry trade unwind — that a price recovery may form within 2–4 months of capitulation. But past cycles did not feature a competing $65 billion AI opportunity pulling mining infrastructure through a permanent exit door. This time, the hashrate that leaves may not come back.
The Bitcoin mining industry is not experiencing a cyclical downturn. It is experiencing a phase transition. The combination of post-halving economics, a price well below production cost, and the gravitational pull of AI data center economics has created conditions where rational actors are permanently reallocating capital away from Bitcoin mining.
For Bitcoin's network security, the implications are profound. The protocol's incentive model assumed that miners would always find it economically rational to secure the chain. That assumption held when Bitcoin mining was the best use of cheap energy. It no longer is.
The hashrate may recover if Bitcoin's price surges above $87,000. But the infrastructure — the powered land, the grid interconnections, the cooling systems, the long-term power purchase agreements — is being contractually committed to AI for 10–15 years. The most important question in Bitcoin's security model is no longer "what is the price?" It is: "when the miners leave, what brings them back?"