The Bitcoin mining industry is undergoing the most dramatic identity crisis in its fifteen-year history. On February 6, 2026, Bitfarms — one of the largest publicly traded Bitcoin miners in North America — announced it would rebrand as Keel Infrastructure, relocate from Canada to Delaware, and tr...
The Bitcoin mining industry is undergoing the most dramatic identity crisis in its fifteen-year history. On February 6, 2026, Bitfarms — one of the largest publicly traded Bitcoin miners in North America — announced it would rebrand as Keel Infrastructure, relocate from Canada to Delaware, and trade under the ticker KEEL. The company's declaration that it is "no longer a Bitcoin company" was not an isolated case of corporate drift. It was the loudest signal yet that an entire industry is fleeing its founding product.
The numbers tell a stark story. CoinShares projects that mining revenue will plummet from 85% of total revenue in early 2025 to under 20% by late 2026 for companies that have secured AI contracts. Hut 8 signed a $7 billion, 15-year Google-backed lease to build AI data centers for Anthropic. Core Scientific rejected a $9 billion buyout to maintain its position as the dominant AI colocation provider. IREN (formerly Iris Energy) locked in a $9.7 billion, five-year AI cloud agreement with Microsoft. In total, public Bitcoin miners signed over $65 billion worth of AI and HPC contracts in 2025 alone.
Meanwhile, Bitcoin's own network dynamics are flashing distress. On February 19, 2026, mining difficulty surged 14.73% — the largest percentage jump since China's 2021 mining ban — to 144.4 trillion. The average cost to mine one Bitcoin has risen to approximately $87,000, while the spot price hovers near $67,000. Miners are, on average, operating 20% underwater. The question is no longer whether Bitcoin mining is profitable. It is whether Bitcoin mining will exist as a standalone business at all.
On February 19, 2026, Bitcoin's mining difficulty adjusted upward by 14.73% to 144.4 trillion — the largest percentage increase since the Chinese mining ban created a hashrate vacuum in mid-2021. In absolute terms, the adjustment represented a gain of roughly 18.5 trillion, the largest raw increase in Bitcoin's history.
The spike was mechanical, not organic. A severe winter storm across the United States in early February forced an estimated 200 EH/s of mining capacity offline, dropping the network hashrate from around 1.0 ZH/s to 826 EH/s. As weather conditions eased, miners brought machines back online, restoring hashrate and triggering the difficulty rebound. The network has since recovered to approximately 961 EH/s.
But the recovery masked a deeper problem. Hashprice — the estimated daily revenue miners earn per unit of hashrate — remains at multi-year lows of $23.9/PH/s. At current prices, the difficulty surge compresses already-negative margins further. The cost to mine one Bitcoin now exceeds the price of one Bitcoin for the average operator.
According to data from CoinDesk, Bitcoin traded at approximately $70,000 in early February while the average production cost stood at $87,000 — meaning the network's average miner was operating at a 20% loss. Only operators with access to power at $0.04–$0.06/kWh can achieve production costs in the $34,000–$51,000 range, maintaining profitability. The rest are bleeding cash.
The profitability crisis is not new — it was baked in by the April 2024 halving, which cut the block reward from 6.25 to 3.125 BTC. But what's new is the simultaneous emergence of a vastly more lucrative alternative: AI infrastructure.
The core economic comparison is devastating for Bitcoin maximalists:
But the revenue gap is only half the story. The quality of revenue is fundamentally different. Bitcoin mining revenue is volatile, correlated to a single asset's price, and subject to difficulty adjustments that erode margins every two weeks. AI hosting revenue comes from 10–15-year contracts with investment-grade counterparties like Google, Microsoft, and Amazon. For a CFO managing a publicly traded company, the choice between Bitcoin's stochastic cash flows and Google's contractual guarantee is not a close call.
JPMorgan projects that miners with more than 500 MW of capacity could derive 30–50% of their revenue from AI by end of 2026 as hyperscalers scramble for power outside the congested U.S. grid.
The pivot is being driven by a simple infrastructure arbitrage. Bitcoin miners spent the last decade building exactly what AI companies now desperately need: large-scale power infrastructure in remote locations with grid interconnection, cooling systems, and security.
TeraWulf set a new benchmark by signing two decade-long hosting agreements with Fluidstack (backed by Google), totaling roughly 200 MW at headline revenue of approximately $1.85 million per MW per year. This became the pricing anchor for the entire industry.
The scale of capital flowing into these conversions is staggering:
| Company | Deal Size | Partner | Capacity | |---------|-----------|---------|----------| | IREN (fka Iris Energy) | $9.7B (5-year) | Microsoft | Multi-GW | | Core Scientific | Rejected $9B buyout | Independent | 1.2+ GW | | Hut 8 | $7.0B (15-year) | Google/Anthropic | Up to 2,295 MW | | Riot Platforms | ~$1.0B (10-year) | AMD | New facility, Rockdale TX | | TeraWulf | Undisclosed | Fluidstack/Google | 200 MW |
As of January 2026, AI-pivoted mining firms trade at nearly double the valuation per megawatt of power compared to their Bitcoin-heavy peers, creating an additional financial incentive for management teams to accelerate the transition.
Bitfarms → Keel Infrastructure (Full Exit) The most dramatic departure. On February 6, 2026, Bitfarms announced redomiciliation from Canada to Delaware, a rebrand to Keel Infrastructure (ticker: KEEL), and a strategic pivot to become an "infrastructure-first" HPC/AI developer. Shareholder vote is scheduled for March 20, with the transition expected to close by April 1. The company explicitly stated it is "no longer a Bitcoin company."
Hut 8 (Hybrid, AI-Dominant) Signed a $7 billion, 15-year lease to develop AI data center capacity for Anthropic, with Google providing the financial backstop. The initial 245 MW facility is scheduled for completion in Q2 2027, with capacity scalable to 2,295 MW. Hut 8 stock surged 25% on the announcement. CEO described the deal as "the first domino to fall."
IREN / Iris Energy (Hybrid, AI-Dominant) Secured a $9.7 billion, five-year AI cloud agreement with Microsoft in late 2025. By early 2026, IREN has become the poster child for the "Great Compute Pivot," trading at nearly double the valuation per megawatt of Bitcoin-only peers.
Core Scientific (AI Colocation Leader) Emerged from bankruptcy in 2024 as the dominant HPC colocation provider, rejecting a $9 billion buyout to maintain independence. Now operates as a critical infrastructure node for the AI economy.
Marathon Digital / MARA Holdings (Struggling Pivot) Rebranded as MARA Holdings but remains heavily reliant on its Bitcoin treasury. Acquired European AI firm Exaion in 2025 but faces high debt levels and has not matched the execution speed of IREN or Hut 8.
Riot Platforms (Late Mover) Signed a 10-year lease with AMD for a new data center at its Rockdale, Texas facility, potentially generating up to $1 billion in revenue. However, the facility is not expected to be operational until 2027, leaving Riot classified as a "construction play" rather than an active AI provider.
Not every miner is abandoning Bitcoin entirely. An emerging operational paradigm, dubbed the "Mullet" strategy by industry analysts, uses Bitcoin mining as a flexible, interruptible load on the back end while running high-margin AI contracts on the front end.
The logic is elegant. Bitcoin mining can be curtailed instantly — ASICs can be powered down in seconds with no data loss or contract penalties. This makes mining the perfect demand-response buffer for electricity grids under stress. During the February winter storm, many large-scale mining operations participated in demand-response programs, shutting down rigs to sell power back to the grid or earning energy credits.
Meanwhile, AI workloads require 99.99% uptime guarantees, making them the priority load that receives uninterrupted power. The combination allows a single facility to maximize both grid participation revenue and AI hosting revenue while using Bitcoin as the "shock absorber" for energy volatility.
An estimated 70% of mining companies have now incorporated some form of AI infrastructure into their portfolios, according to industry data. The pure-play Bitcoin miner — once the archetypal crypto business — is becoming an endangered species.
The exodus raises a question that Bitcoin's core developers and community have long debated in theory but never confronted in practice: what happens to network security when the most capitalized miners find something better to do?
Bitcoin's security model relies on miners investing real resources (electricity, hardware) to validate transactions. The 15% difficulty swing in February demonstrated the network's self-correcting mechanism — difficulty drops when hashrate falls, making mining easier and theoretically attracting miners back. But if the economic opportunity cost of mining exceeds the reward, the self-correction breaks down.
At current prices, Bitcoin produces roughly $47 million per day in block rewards and fees. The AI infrastructure market that former miners are entering is projected to exceed $500 billion by 2030. The gravitational pull of AI capital is orders of magnitude stronger than Bitcoin's block reward incentive.
The counterargument is that Bitcoin's price will eventually rise enough to restore mining profitability, and that the remaining miners — those with the cheapest power — will secure the network adequately. This has been true historically. But the current cycle is different in one crucial respect: for the first time, miners have a competing use case for their core physical asset (power infrastructure) that does not require Bitcoin's price to cooperate.
The Bitcoin mining industry is not pivoting. It is evacuating. When a company founded to mine Bitcoin publicly declares it is "no longer a Bitcoin company," when Google and Microsoft are the new counterparties replacing block rewards, when the average miner loses money on every Bitcoin they produce — this is not a strategic rotation. It is a structural break.
The economic logic is irrefutable. A megawatt of power pointed at AI workloads generates more revenue, more predictably, with more creditworthy counterparties than a megawatt pointed at SHA-256 hashing. The miners who built the infrastructure to secure Bitcoin's network have discovered that their real asset was never the hashrate — it was the power.
Bitcoin's network will survive this transition. Difficulty adjustments will recalibrate, the most efficient miners will persist, and if prices recover, hashrate will follow. But the era of the "Bitcoin mining company" as a standalone category is ending. What remains will be energy infrastructure companies that happen to mine Bitcoin when prices justify it — and shut off the ASICs the moment something better comes along. That something is already here, and it has a $65 billion head start.