Share
The story of Bitcoin was always, at its core, a story about incentives. Satoshi Nakamoto’s white paper described a system where rational economic actors, miners, would be rewarded for securing the network. Their greed, channeled correctly, would make the network stronger. For over a decade, that held. But something has changed, and it may be the most structurally important development in Bitcoin’s history.
The miners are leaving.
Not all at once, and not loudly. But the data is unambiguous, the corporate filings are explicit, and the directional shift is now irreversible unless something dramatic changes. Nearly every major publicly traded Bitcoin mining company has made formal commitments to pivot a dominant share of their infrastructure, capital, and future growth plans toward artificial intelligence.
This is a wild and concerning trend for Bitcoin.
This is far worse than I had realized.This is a list of all the major public Bitcoin miners.
ALL have made statements to pivot to AI.
ALL are targeting major shares of revenue from AI from here, not Bitcoin.On average current… pic.twitter.com/lJqlq0mAHt
— Charles Edwards (@caprioleio) April 17, 2026
Moreover, the financial logic driving this decision is so compelling it is almost impossible to argue against from a shareholder’s perspective, which is precisely why it is so alarming for Bitcoin itself.
To understand why this is happening, you need to understand what broke the old model.
Bitcoin mining was always a race, a relentless, capital-intensive arms race where the prize, newly minted Bitcoin, was split among whoever solved a cryptographic puzzle first. The network adjusts difficulty automatically to keep blocks arriving roughly every ten minutes, meaning that as more computing power joined, rewards per unit of hash rate, known as ‘hashprice,’ compressed. For years, rising Bitcoin prices masked this structural pressure. When BTC was climbing from $10,000 to $69,000, even with shrinking margins per unit of compute, the total dollar value of rewards kept operators flush.
Then came April 2024’s halving, the fourth in Bitcoin’s history, and the arithmetic became brutal. Every four years, the number of Bitcoins miners receive as a reward is halved, and competition on the network has grown exponentially. The 2024 halving cut block rewards from 6.25 BTC to 3.125 BTC per block, instantly slashing revenue for every miner on earth. At the same time, Bitcoin mining revenue is projected to plummet from around 85% of total revenue in early 2025 to less than 20% by the end of 2026 for companies that have secured AI contracts.
CoinShares’ Q1 2026 Bitcoin Mining Report highlights falling hash prices, to $29 per PH/s/day, and rising costs forcing public miners like MARA, RIOT, and CleanSpark to redirect rigs toward AI and high performance computing. At $29 per petahash per day, many operators are at or below breakeven, particularly those running older hardware or paying market rates for electricity.
The pressure is not just from halving. Revenue earned for validating transactions on the blockchain is being driven down by plummeting token prices and soaring energy costs, accelerating the pivot that began around three years ago to provide infrastructure that generates computing power for AI firms.
The pull of AI is not just that Bitcoin mining became less attractive. It is that AI infrastructure became dramatically, almost absurdly, more attractive, and Bitcoin miners happened to own exactly the assets that AI companies desperately needed.
The convergence is almost too neat. AI hyperscalers such as Amazon, Microsoft, and Google are spending trillions of dollars building out data centers for model training and inference. The single most constrained resource in that buildout is not chips or software or engineers. It is power. Specifically, access to large, secured, grid-connected power at scale, ideally with existing high voltage infrastructure already in place. Bitcoin miners had spent billions of dollars building exactly that.
HIVE estimates 10 MW of Nvidia H100 GPUs yield revenue similar to 100 MW of Bitcoin mining. The same physical footprint, the same land, the same power connections, the same buildings, can generate ten times more revenue when filled with AI GPUs than when filled with Bitcoin ASICs. From a pure capital allocation standpoint, the decision is not even close.
BIRD gets $50M in predatory financing to chase GPU business with zero infra or expertise. Stock +582%.
HIVE announces $75M, 0% convert to buy NVDA B200s & scale HPC toward $140M+ ARR, following record $93M rev qtr & actual Tier III data centers. Stock down.
— matthew sigel, recovering CFA (@matthew_sigel) April 16, 2026
In 2025 alone, public Bitcoin miners signed over $65 billion worth of AI and high performance computing contracts with hyperscalers like Amazon and Microsoft. The asset management company projects that AI infrastructure can generate three times the revenue per megawatt compared to Bitcoin mining.
Reports go even further, suggesting that AI workloads can deliver up to 25 times more revenue per kilowatt hour than mining, depending on the application. With numbers like these, it is no surprise that an estimated 70% of mining companies have now pivoted to include AI infrastructure in their portfolios.
The margin profile is equally seductive. Companies that have pivoted can generate 80% to 90% operating margins from these AI deals, compared to the thin margins typically associated with Bitcoin mining operations.
Regulators are pushing the same direction. Regulatory bodies, particularly in energy constrained regions like Texas and New York, have become increasingly skeptical of pure Bitcoin mining, viewing it as a drain on the grid with little local economic benefit. This regulatory pressure has accelerated the AI pivot. AI data centers are viewed by policymakers as essential infrastructure for the 21st century economy, making it easier for companies to secure power permits and tax incentives.
This is not a fringe trend among a few experimental companies. It is the dominant strategy of virtually every major publicly traded Bitcoin miner, announced in earnings calls, SEC filings, and press releases. The pivot is not hypothetical. Much of it is already operational.
Core Scientific, once one of the largest crypto miners in the United States, the company that famously filed for bankruptcy in 2022, has undergone the most dramatic transformation. It has restructured and emerged as a bare metal AI infrastructure provider, offering H100 cluster rentals to AI startups and research labs. Its deal with CoreWeave is the clearest example of how complete this transformation can be: around 350 MW had already been energized under its CoreWeave contract by end of 2025, and the company remains on track to deliver around 590 MW by early 2027.
Riot Platforms, historically one of the most Bitcoin committed miners in the world, has turned Corsicana, Texas into a staging ground for AI. In early 2025, Riot engaged financial advisors to explore opportunities leveraging 600 MW spare capacity at its Corsicana, Texas site, one of the largest US data centers near Dallas, with management confirming this AI and HPC focus in its 2024 results.
The company’s commitment became so complete that Chief Data Center Officer Jonathan Gibbs resigned in April 2026, forfeiting 1.1 million unvested shares worth approximately $18.7 million, just ten months after his high profile recruitment, a sign of how rapidly and disruptively the company’s identity is shifting.
MARA Holdings, which holds the largest Bitcoin treasury among all public miners at around 50,000 BTC, has pursued a more hybrid model. MARA’s partnership with Starwood describes campuses designed to operate both Bitcoin mining and AI compute, with the ability to toggle workloads depending on pricing and customer demand. This ‘toggle’ model sounds elegant, maximizing optionality, but it also means Bitcoin is no longer the unconditional priority. It becomes the fallback when AI contracts are not available.
The largest Bitcoin miner balance sheet is now a distressed seller after liquidating over $1B of its Bitcoin at ~$72,000 after making purchases 30-40% higher above $90,000 per BTC. This is just the beginning of a major unwind across the sector. https://t.co/2UBxZPrtin pic.twitter.com/1QEa3y37CB
— Quinn Thompson (@qthomp) March 26, 2026
IREN, formerly Iris Energy, has gone perhaps furthest in stock market terms. Projections show IREN reaching 660 MW of contracted critical IT load by 2026, representing about 250,000 GPUs and roughly $6 billion in annualized cloud services revenue, a number that dwarfs anything achievable from Bitcoin mining alone. Its Microsoft deal reshaped the company’s valuation almost overnight.
TeraWulf has announced plans to lease out a data center on its western New York campus to Fluidstack, an AI cloud platform with support from Google, while also partnering to develop an AI data center in Texas. Bitfarms has converted one of its Quebec facilities into an AI training center, installing AMD MI300X racks to support open source model development in partnership with university labs.
CleanSpark, long considered one of the most mining pure operators, has quietly shifted too. The company secured up to 890 megawatts of new utility grade power capacity and is advancing AI ready site portfolios across Texas and Georgia, with leadership stating the company is progressing toward AI tenancy at key sites.
Here is where the business story collides with something more fundamental. Bitcoin’s security model depends entirely on what is called proof-of-work, the continuous expenditure of enormous amounts of computational energy to validate transactions and prevent the blockchain from being rewritten. The more total computing power, hashrate, secures the network, the more expensive it becomes for any adversary to attempt a 51% attack, which would require controlling a majority of that power.
The mechanism is elegant and self reinforcing when it works. Miners are paid in Bitcoin to protect Bitcoin. Their economic interest aligns with network security. But that alignment is breaking down.
The Bitcoin network confronted its most severe difficulty adjustment in 2026 as major mining operations redirected resources toward artificial intelligence infrastructure. The 7.7% mining difficulty decline marked one of the sharpest drops on record, directly correlating with mining operators’ strategic pivot to AI data centers.
For the first time in six years, Bitcoin’s hashrate posted a Q1 decline, dropping around 4% year to date to hover near 1 zettahash per second, breaking its streak of quarterly gains amid profitability squeezes.
The revenue consequences are cascading. Bitcoin mining economics have tightened significantly: hashprice has dropped to near all time lows, hovering around $28 to $35 per PH/s/day in Q1 2026, near or below breakeven for many operators.
Companies are not just pivoting their future capital, they are liquidating their present Bitcoin holdings to fund the transition. Major operators including MARA Holdings sold 13,210 BTC, Riot Platforms divested 4,026 BTC, and Core Scientific liquidated 1,992 BTC to fund operational transitions in Q1 2026 alone.
As mining capacity migrates to AI applications, the remaining hash rate becomes concentrated among fewer participants, potentially increasing centralization risks. This is subtle but important: it is not just that there is less total hash rate, it is that whatever remains may be concentrated in the hands of fewer, larger operators who themselves have diversified incentives. The distributed, adversarially aligned security model Satoshi envisioned becomes something more fragile.
Bitcoin miners are the system’s bodyguards, protecting the nearly 20 million circulating coins. Every major US miner has already started their migration to AI as mining became deeply unprofitable following the last halving event. Many miners are not even planning to reinvest in Bitcoin specific hardware.
Rather than purchasing the latest generation ASICs, operators are running existing equipment to end of life and redirecting all new capital into GPU clusters for AI. Analysts project that 20% of global Bitcoin mining power will shift to AI workloads by the end of 2027.
Against this backdrop, the emerging quantum computing threat to Bitcoin’s cryptographic foundations has moved from theoretical to increasingly urgent. The timing is deeply problematic, Bitcoin is experiencing its largest ever drain on its security workforce precisely as its underlying cryptography faces a credible timetable for compromise.
Bitcoin’s security relies on two cryptographic systems. Proof-of-work via SHA-256 protects the history of the blockchain, the record of all transactions. The Elliptic Curve Digital Signature Algorithm protects individual wallets, the mechanism by which you prove you own your Bitcoin and authorize spending. These face very different quantum threats.
Bitcoin mining relies on SHA-256, which is highly resistant to quantum algorithms. A quantum computer would merely act as a faster miner, not a threat to network consensus. The more critical vulnerability is ECDSA, the math used to generate public and private keys, which is the mechanism that proves ownership and authorizes transactions.
Recent research has dramatically shortened the projected timeline for when this becomes a real problem. In March 2026, landmark research fundamentally shifted the landscape. These papers demonstrated that breaking Bitcoin’s ECDSA encryption requires far fewer resources than previously estimated. Instead of needing millions of physical qubits, researchers found that advanced architectures could theoretically execute Shor’s Algorithm with under 500,000 qubits. With these optimized systems, recovering a private key from an exposed public key could take mere minutes rather than days.
On March 30, 2026, research published by major institutions showed that the computing power needed to break crypto security is about 20 times lower than previously estimated. While the actual threat is still years away, the deadline is moving closer much faster than expected.
Up to 60% of the network’s entire supply is at risk if a malicious actor gets their hands on a quantum computer capable of breaking Bitcoin’s encryption, with around $800 billion in Bitcoin potentially exposed according to industry estimates. Consensus among cybersecurity experts had placed the appearance of a cryptographically relevant quantum computer anywhere between two and eight years away, and that was before the March 2026 research breakthroughs.
The situation is nuanced rather than binary. Estimates suggest an attacker would need millions of qubits, orders of magnitude more than current systems, to crack a key within hours or days. Researchers estimate that even the most advanced quantum computers are 10 to 100,000 times too weak to pose a real world threat, pushing meaningful risk into the 2030s or later. The most dangerous addresses are legacy P2PK addresses, where the public key is exposed onchain, making offline cryptanalysis possible. Modern address types hide keys until coins are spent, providing some protection.
But here is the structural problem: addressing the quantum threat to Bitcoin requires a coordinated, community wide upgrade to post quantum cryptography. That process requires active engagement from miners, developers, and node operators. It requires people who are economically invested in Bitcoin’s long term security to be at the table and paying attention. Mitigating the quantum risk would require widespread consensus among miners and nodes, technically feasible, given the cryptocurrency’s history of incremental improvements, but requiring that alignment to actually exist.
Quantum compute isn't a threat, it's an opportunity.
Most people think that quantum compute is here to destroy everyone's lives.
That's not the case.It's going to change a lot of lives, just like Internet, #Crypto and AI has been doing this already, and given the early… pic.twitter.com/9Cb8V51GWW
— Michaël van de Poppe (@CryptoMichNL) April 21, 2026
For Bitcoin to be secure, especially long term, it needs to have a robust security system, meaning as many miners as possible online, offering up as much computational power as possible to render an attack futile. The opposite is happening.
What makes this moment genuinely unprecedented is that none of these pressures are freak accidents. Each is the natural outcome of how the system was designed.
The halving was always scheduled. The compression of mining rewards was always the plan, Satoshi designed Bitcoin to become increasingly scarce over time, with security eventually paid for by transaction fees rather than new coins. The problem is that transaction fee revenue has not scaled to replace block rewards. Bitcoin’s fee market remains episodic and unpredictable, incapable of reliably sustaining the security budget that proof of work requires.
The AI boom was not designed, but it was inevitable given the physical assets miners had accumulated. Power capacity at scale is the scarcest resource in the modern economy. Miners built the infrastructure. Of course they would be first in line when buyers came calling.
The quantum threat is the product of physics research that has been advancing for decades. Its timeline accelerating is a natural consequence of global investment in the technology.
The new metric of success in 2026 is power capacity, not hashrate, not Bitcoin treasury. Those who stayed focused solely on mining were eventually absorbed by those who built broader infrastructure, much like early internet service providers who stayed on dial up were swallowed by those who built broadband.
The market has delivered its verdict with unusual clarity. The AI contracts generate three times the revenue on a per megawatt basis compared to traditional mining operations. Over $70 billion in AI and HPC contracts have been announced, with listed miners potentially deriving 70% of revenues from AI by year end 2026, up from 30% earlier in the year.
For Bitcoin’s security model to remain robust through this transition, several things would need to happen that currently show little sign of materializing.
Transaction fees would need to meaningfully replace block rewards, but Bitcoin’s fee market remains volatile and inconsistent. For miners to stay committed, users must pay more for transactions, which requires significantly higher network usage – something that depends on improvements in scalability, user experience, and utility.
The post quantum upgrade would need to happen before Q-Day arrives. The Bitcoin community is actively researching post quantum solutions, with two algorithms gaining significant traction, Dilithium and Falcon, which are lattice based cryptographic solutions.
The current challenge lies in signature size: while a traditional Bitcoin signature is about 70 bytes, a PQC signature can balloon to several kilobytes. That is not a trivial problem, it would meaningfully increase block sizes, transaction costs, and node storage requirements. Achieving consensus in a community as decentralized and ideologically diverse as Bitcoin’s is slow and contentious even for minor changes. This is a major one.
New classes of miners would need to emerge who are committed to Bitcoin only operations even at thin margins. Some pure play operators like Canaan, notably the ASIC manufacturer that closed its AI chip division in July 2025, opting to concentrate exclusively on Bitcoin mining hardware, are making that bet. But they remain a shrinking minority of the overall landscape.
There is also a scenario where none of this ends badly, at least not immediately. Despite the pivot to AI, publicly traded mining companies still grew their Bitcoin mining operations during 2025, listed miners collectively added more mining computing power in the first nine months of 2025 than they did in the same period of 2024.
The absolute hashrate, in other words, was still growing even as the AI share of revenue grew. But Q1 2026 appears to mark the inflection point where that dynamic has reversed.
Bitcoin was built on a simple but powerful idea: align incentives so that independent actors, each acting in their own self-interest, collectively secure the network better than any central authority could. Miners weren’t supposed to be altruistic, they were supposed to be rational. Their profit motive was the system.
What’s changing now isn’t a betrayal of that idea, it’s a consequence of it. These same rational actors are seeing a more profitable use of their infrastructure in AI. For publicly listed companies, the shift isn’t ideological, it’s financial. They’re doing what they’re expected to do: follow the returns.
The problem is that Bitcoin’s incentive design never accounted for a competing industry that could generate significantly higher yields using the same hardware, energy, and capital. For over a decade, Bitcoin dominated as the primary outlet for this kind of compute. That era is starting to shift, not because the network failed, but because something more lucrative emerged alongside it.
So this isn’t about Bitcoin breaking. It’s about its incentives being tested in a new environment. What happens next, whether fees rise to compensate, new miners step in, or the system evolves in other ways, will determine whether this moment becomes a setback or a turning point.
Share
preferred on
