The Bitcoin mining industry is experiencing its most dramatic structural transformation since China's 2021 ban. In the first quarter of 2026, at least nine publicly traded miners have announced partial or complete pivots from Bitcoin mining to artificial intelligence and high-performance computin...
"We're no longer a Bitcoin company." — Ben Gagnon, CEO, Bitfarms (now Keel Infrastructure), February 2026
The Bitcoin mining industry is experiencing its most dramatic structural transformation since China's 2021 ban. In the first quarter of 2026, at least nine publicly traded miners have announced partial or complete pivots from Bitcoin mining to artificial intelligence and high-performance computing (HPC) infrastructure. Collectively, these firms have signed over $43 billion in AI hosting contracts, sold more than 15,000 BTC from their treasuries, and begun the systematic conversion of mining facilities into GPU data centers.
This is not a temporary reallocation. CoinShares estimates that mining revenue will fall from 85% of total sector revenue to under 20% by late 2026. The economics are unambiguous: AI customers pay steady, high-margin rents with 80–90% profit margins, while Bitcoin mining profits swing with crypto prices and difficulty adjustments. The question is no longer whether miners will pivot — it is whether enough hashrate will remain to secure a $1.2 trillion network.
The few holdouts, most notably the Trump family-backed American Bitcoin (ABTC), are making a contrarian bet that the exodus itself creates opportunity. But the macro trend is clear: the companies that once secured Bitcoin's network are becoming the power backbone of the AI revolution.
The speed of the mining-to-AI pivot has been staggering. In a matter of months, the largest publicly traded Bitcoin miners have collectively announced more than $43 billion in AI and HPC hosting contracts. These are not speculative memoranda of understanding — they are binding, multi-year lease agreements backed by hyperscalers and major cloud infrastructure providers.
The three largest deals alone account for the majority of this figure:
| Company | Contract Value | Partner/Client | Duration | |---------|---------------|----------------|----------| | Core Scientific (CORZ) | $10B+ | CoreWeave | 12 years | | Hut 8 (HUT) | $7B | Fluidstack (Google-backed) | 15 years | | TeraWulf (WULF) | $6.7B | Hyperscaler (undisclosed) | Multi-year |
These contracts translate into predictable, high-margin annual revenue streams. Core Scientific's $10 billion contract alone represents an approximately $850 million average annual revenue run rate — dwarfing the economics of its former mining operations.
The remaining $19+ billion in contracts is distributed among Riot Platforms, Bitfarms (Keel Infrastructure), IREN, Cipher Mining, Bit Digital, and others, each securing their own hyperscaler relationships.
Core Scientific (CORZ) has committed to the most aggressive transformation. The company's Q4 2025 earnings presentation, released March 2, 2026, showcased its $10B+ contracted revenue pipeline and announced that by the end of 2028, every megawatt in its portfolio will be dedicated to AI colocation. Its AI colocation revenue soared 268% quarter-over-quarter. In January 2026, the company sold 1,900 BTC for $175 million, and management expects to monetize "substantially all" remaining Bitcoin holdings during Q1 2026.
Riot Platforms (RIOT) posted a 72% jump in 2025 fiscal year revenues to $647.4 million, driven by both mining and its new data center services. In early 2026, Riot began repurposing significant portions of its capacity for AI and HPC hosting, transitioning toward a "Power-as-a-Service" model that leverages its 2.4 GW of total power capacity across facilities in Texas, Kentucky, and Colorado.
Bitfarms (BITF → KEEL) represents the most symbolically dramatic pivot. In February 2026, the company declared it is "no longer a Bitcoin company," announced a move from Canada to the United States, and unveiled plans to rebrand as Keel Infrastructure (ticker: KEEL). Its Washington State facility — an 18 MW mining farm — is being retrofitted to support Nvidia GB300 GPUs with advanced liquid cooling, backed by a fully funded $128 million deal. A shareholder vote on March 20 will formalize the transformation.
Hut 8 (HUT) signed a landmark 15-year, $7 billion lease agreement with Fluidstack, a Google-backed cloud infrastructure provider, to supply data center space and power for AI computing workloads. This single contract exceeds the total lifetime mining revenue of most publicly traded miners.
HIVE Digital Technologies is investing $100 million in HPC expansion, while its executive chairman Frank Holmes has been vocal about the cost realities: cutting-edge HPC facilities cost more than 30 times as much as traditional Bitcoin mining infrastructure per megawatt.
The financial logic driving the exodus is stark. HIVE's Frank Holmes provided the clearest cost breakdown in January 2026:
But the revenue differential is even more extreme. HIVE estimates that 10 MW of Nvidia H100 GPUs yield revenue equivalent to 100 MW of Bitcoin mining. AI hosting contracts generate roughly 3x the revenue per megawatt compared to mining operations, with dramatically more predictable cash flows.
Meanwhile, Bitcoin mining profitability has collapsed. JPMorgan reported that miners earned an average of just $38,700 per EH/s in daily block reward revenue in December 2025 — down 7% from November and 32% year-over-year, representing the lowest level on record. CNBC reported in February 2026 that Bitcoin mining is "no longer profitable" for many operators after crypto's latest downturn, with BTC trading in the $62,000–$63,000 range.
The post-halving squeeze has turned what was already a difficult business into an unsustainable one for all but the most efficient operators. When a single 15-year AI lease from a creditworthy counterparty exceeds the total expected mining revenue from the same megawatts, the capital allocation decision becomes trivial.
The pivot has triggered a parallel phenomenon: the end of the miner HODL strategy. According to CoinDesk, public miners have sold more than 15,000 BTC from their treasuries since Q4 2025, with the pace accelerating into Q1 2026.
Key treasury liquidation events:
This represents a fundamental philosophical shift. For years, the miner HODL thesis — that miners were the smartest Bitcoin accumulators, buying at production cost — was a core narrative supporting BTC price expectations. The simultaneous unwinding of this thesis by the industry's largest players removes a structural source of demand and adds a new source of supply.
The irony is acute: the companies that secured Bitcoin's network are now its most motivated sellers.
The exodus has tangible implications for Bitcoin's security model. In January 2026, the network's hashrate plunged from highs of ~1.1 ZH/s to as low as 663 EH/s — a 40% drop — during Winter Storm Fern. While the storm was the acute catalyst, the underlying trend reflects structural hashrate withdrawal as miners redirect power capacity.
The difficulty adjustment mechanism — Bitcoin's elegant self-correcting design — partially mitigates the risk. Following the January drop, difficulty adjusted downward by 16–18%, one of the largest single adjustments in recent history, improving margins for remaining miners. But the long-term trajectory is concerning: if mining revenue falls to 20% of sector activity as CoinShares projects, the economic incentive to secure the network weakens considerably.
Bitcoin's security budget — the total revenue flowing to miners through block rewards and fees — was approximately $18.1 billion annually from issuance alone (per the webthreepedia economic value framework). If a significant portion of capacity permanently exits for AI, the network must either see substantially higher transaction fees or accept a lower security budget. Neither outcome is assured.
The geographic concentration risk also intensified. When Bitfarms exits Canada and multiple operators consolidate in U.S. power markets optimized for AI, the remaining mining hashrate becomes more geographically concentrated, potentially increasing regulatory and operational risk vectors.
Not every miner is leaving. American Bitcoin (ABTC), backed by the Trump family, announced on March 3, 2026 the purchase of 11,298 ASIC miners for deployment at its Drumheller, Alberta site — a 12% capacity expansion adding 3.05 EH/s. The company is explicitly betting that the exodus creates opportunity: fewer competitors mean lower difficulty and better economics for those who remain.
At current prices (~$68,000 BTC), the added capacity could produce roughly 42 BTC per month (~$2.9 million in gross monthly revenue). Whether this contrarian bet pays off depends entirely on two variables: Bitcoin's price trajectory and how much hashrate permanently exits the network.
Marathon Digital Holdings (MARA) occupies a middle ground, maintaining its position as the leader in pure hashrate (~60 EH/s) while acquiring French HPC firm Exaion for European AI exposure. Marathon's strategy hedges both outcomes — but its lack of Riot's vertical integration or Core Scientific's committed AI contracts leaves it vulnerable in either direction.
The Great Mining Exodus of 2026 represents more than an industry pivot — it is a stress test for Bitcoin's fundamental economic model. The network's security has always depended on the assumption that mining would remain the highest and best use of cheap electricity. AI has shattered that assumption.
The $43 billion in AI contracts demonstrates that the market has spoken: the power infrastructure built to mine Bitcoin is more valuable serving artificial intelligence. For the companies making this transition, the economics are compelling — predictable, high-margin revenue streams that make mining look like a rounding error.
But the second-order effects deserve serious attention. Bitcoin's security is not an abstract concept — it is a direct function of the economic incentives flowing to miners. As the most sophisticated operators exit for AI, the network's defense falls to a smaller, potentially less capitalized set of miners. The difficulty adjustment mechanism provides resilience, but it does not solve the fundamental question: can Bitcoin's fee market alone sustain adequate security as block rewards continue to halve?
The contrarian holdouts betting on this transition may prove prescient. In a world where most miners leave, those who remain inherit a less competitive environment. But this only works if Bitcoin's price and fee revenue justify the opportunity cost of not hosting AI workloads.
What is certain is that the relationship between Bitcoin mining and energy infrastructure has permanently changed. The companies that once defined themselves by hashrate now define themselves by megawatts available for AI. The Great Mining Exodus is not a temporary market dislocation — it is a structural realignment that will define Bitcoin's security economics for the next decade.