The largest publicly traded Bitcoin miners are executing the most dramatic industrial pivot since the shale oil industry retooled for natural gas. In February 2026 alone, MARA Holdings announced a 1 GW AI data center joint venture with Starwood Capital, Bitdeer liquidated its entire 1,133 BTC tre...
"We entered a strategic agreement with Starwood to accelerate the delivery of cutting-edge hyperscale, enterprise, and AI-capable digital infrastructure." — Fred Thiel, CEO, MARA Holdings
The largest publicly traded Bitcoin miners are executing the most dramatic industrial pivot since the shale oil industry retooled for natural gas. In February 2026 alone, MARA Holdings announced a 1 GW AI data center joint venture with Starwood Capital, Bitdeer liquidated its entire 1,133 BTC treasury to fund AI expansion, and Cipher Mining shut down Bitcoin operations at its Black Pearl facility entirely. These are not hedging strategies. They are existential pivots.
The math is unforgiving. Post-halving hash prices have collapsed to roughly $0.045 per TH/s per day, while network difficulty hit 139.33T on February 20. Meanwhile, AI workloads generate up to 25 times more revenue per kilowatt-hour than Bitcoin mining. JPMorgan reports that the 14 U.S.-listed miners it tracks ended January 2026 with a combined market cap of $60 billion — up 23% month-over-month — even as Bitcoin itself declined 4%. The market is no longer pricing these companies as Bitcoin proxies. It is pricing them as AI infrastructure plays.
This report examines the economics driving this transformation, the companies leading the charge, and what this means for Bitcoin's security model when its miners no longer need it.
Bitcoin's April 2024 halving cut block rewards from 6.25 BTC to 3.125 BTC. The industry called it "the harshest margin environment of all time." They were right.
Network difficulty adjusted upward to 139.33 trillion on February 20, 2026, while hash price — the revenue earned per terahash per second per day — has fallen to a structural low near $0.045. At current difficulty, 1 TH/s earns approximately 428 satoshis daily. Transaction fees, which constitute just 0.48% of block rewards, have dropped to 12-month lows.
The industry's average all-in production cost now sits between $34,000 and $43,000 per BTC for the most efficient operators. With Bitcoin trading near $63,000–$67,000 in late February 2026 — a 50% decline from its October 2025 peak of roughly $125,000 — margins have evaporated for all but the best-capitalized miners. To sustain the current network hashrate of over 1,054 EH/s at post-halving economics, Bitcoin likely needs to trade between $90,000 and $160,000.
MARA Holdings, the largest public miner by hash rate, reported a $1.71 billion Q4 2025 loss. Revenue fell 6% quarter-over-quarter to $202.3 million, driven by a 14% decline in the average price of Bitcoin mined. Full-year revenue reached $907.1 million — a 38% increase — but the quarterly trajectory signals trouble ahead.
The bottom line: mining Bitcoin is becoming economically irrational for companies that have alternative uses for their most valuable asset — cheap, grid-connected power at scale.
The pivot is driven by a single, overwhelming economic fact: AI workloads generate dramatically more revenue per unit of power than Bitcoin mining.
According to industry data, AI and HPC contracts generate approximately three times the revenue on a per-megawatt basis compared to mining operations. In optimal configurations, the multiple reaches 25x per kilowatt-hour. TeraWulf's decade-long hosting agreements with Google-backed Fluidstack price at roughly $1.85 million per MW per year. IREN's landmark deal with Microsoft covers 200 MW of IT load and generates an expected $1.94 billion in annual recurring revenue at approximately 85% project EBITDA margins.
Compare this to Bitcoin mining. At current hash prices, a 200 MW mining facility generates approximately $70–90 million in annual revenue with margins compressed to single digits after power and depreciation. The same facility, converted to AI hosting, can generate $300–400 million with margins above 50%.
The kicker: AI contracts are typically 5–12 years in duration, with hyperscaler counterparties like Microsoft, Google Cloud, and Amazon Web Services. These are investment-grade revenue streams, not speculative bets on hash price recovery.
Nvidia's Q4 fiscal 2026 earnings — $68.1 billion in quarterly revenue, up 73% year-over-year, with data center revenue of $62.3 billion — confirmed that demand for AI compute infrastructure is accelerating, not plateauing. This is the demand signal that miners are chasing.
An estimated 70% of the top publicly traded Bitcoin miners are now generating some revenue from AI and high-performance computing. But the depth of commitment varies dramatically.
The Full Converts:
Bitdeer made the most symbolic move of the month. On February 20, it liquidated its entire Bitcoin treasury — all 1,133 BTC — to fund AI infrastructure expansion. The company simultaneously priced a $325 million convertible notes offering and a $43.5 million equity raise. This is not a hedge. A miner selling every satoshi it owns to fund AI data centers is a company that has decided Bitcoin mining is a transitional business.
Cipher Mining ceased all Bitcoin mining at its Black Pearl facility in February 2026. Morgan Stanley issued an "overweight" rating with a $38 price target, projecting 66% revenue growth in 2026, driven entirely by AI contract execution. Cipher has landed multi-billion dollar contracts with Google Cloud and Amazon Web Services.
The Hybrid Operators:
MARA Holdings announced a joint venture with Starwood Capital Group on February 26 to develop 1 GW of near-term IT capacity across MARA's power-rich sites, with a long-term roadmap to 2.5 GW. Starwood Digital Ventures will lead design, construction, tenant sourcing, and facility operations. Despite the $1.71 billion quarterly loss, MARA's stock surged 10% on the announcement. The market rewarded the AI story, not the mining results.
TeraWulf secured contracts valued at approximately $6.7 billion, backed by Google's financial frameworks. Morgan Stanley assigned a $37 target price with projected 132% revenue growth in 2026. Its Fluidstack hosting agreements run for a decade.
The Scale Leaders:
IREN (formerly Iris Energy) operates with a $14 billion market capitalization and projects $3.4 billion in annual recurring revenue by end of 2026, with $2.3 billion already under contract. Its anchor deal — a $9.7 billion AI cloud contract with Microsoft — covers a 3 GW power pipeline, though conversion timelines face scrutiny.
Core Scientific has accumulated approximately $10.2 billion in projected cumulative revenue over 12-year contract terms with CoreWeave. It is positioned to provide approximately 500 MW of critical IT load for HPC infrastructure by the second half of 2026.
The Domino Effect:
Riot Platforms recently sold $200 million in Bitcoin to fund AI expansion. Bitfarms is dropping its "Bitcoin company" identity entirely. The pivot is no longer a strategy — it is an industry-wide migration.
The most telling indicator of this transformation is the decoupling of mining stocks from Bitcoin's price.
JPMorgan reported that the 14 U.S.-listed miners and data center operators it tracks ended January 2026 at $60 billion in combined market capitalization — a 23% monthly increase while Bitcoin declined 4%. Twelve of the fourteen miners outperformed Bitcoin that month. IREN rose 42%.
Historically, mining stocks were leveraged Bitcoin proxies. Over the prior three years, while Bitcoin surged more than 450%, MARA stock rose less than 50% and Riot advanced less than 240%. Miners consistently underperformed the commodity they mined.
That relationship has broken. Equity markets are re-rating these companies based on contracted AI revenue, not speculative hash price recovery. This is a fundamental regime change in how the market values Bitcoin mining infrastructure.
Mining revenue is projected to fall from approximately 85% of total revenue in early 2025 to less than 20% by the end of 2026 for companies that have secured AI contracts. The "Bitcoin miner" label is becoming a misnomer.
Here is the question that the industry prefers not to answer: if the most efficient mining operations on Earth find it more profitable to host AI workloads than to mine Bitcoin, what happens to network security?
Today, Bitcoin's hashrate stands at 1,054 EH/s. But this is sustained by miners who are, in many cases, cross-subsidizing mining operations with AI revenue or burning through cash reserves. As AI contracts ramp and conversion timelines accelerate through 2026–2027, the opportunity cost of dedicating power to mining will increase.
The webthreepedia economic value framework has consistently highlighted that Bitcoin requires $54–72 billion annually in subsidies to secure just $115 million in user fees. The mining sector's AI pivot does not solve this structural imbalance — it may accelerate it. As sophisticated operators reallocate power capacity to higher-returning AI workloads, the miners who remain may be smaller, less efficient, and more concentrated in jurisdictions with the cheapest — and least reliable — power.
The optimistic scenario: hash price recovers if enough capacity exits mining, restoring profitability for remaining operators through difficulty adjustments. The pessimistic scenario: a positive feedback loop where falling hash price drives further defections to AI, concentrating hashrate among fewer participants and degrading Bitcoin's decentralization — the one property that underpins its entire value proposition.
The Bitcoin mining industry is not dying — it is metamorphosing. The companies that built the world's largest grid-connected power infrastructure to solve SHA-256 hashes have discovered that the same infrastructure is worth multiples more when solving transformer matrices for hyperscalers.
This is rational economic behavior. When a 200 MW facility generates $300–400 million hosting AI workloads versus $70–90 million mining Bitcoin, the decision is not strategic — it is arithmetic. The market agrees: mining stocks have decoupled from Bitcoin and are trading on contracted AI cash flows.
But rationality at the company level creates risk at the protocol level. Bitcoin's security has always depended on mining being the highest-return use of specialized power infrastructure. That assumption is now being tested by the single largest compute demand surge in history. The miners are not abandoning Bitcoin out of malice — they are following the economics. The question is whether Bitcoin's incentive structure can adapt before its most capable defenders have fully pivoted to serving a different master.