Bitcoin Firms Bet Big on AI As Mining Margins Shrink

It’s been close to two decades now since Bitcoin (BTC ) was first launched by its pseudonymous creator, Satoshi Nakamoto.

In those early years, mining Bitcoin was extremely easy and cheap. One could do that on their home computer and get their hands on hundreds of Bitcoin easily. But now, even companies with advanced machines and massive investments are struggling to mine BTC profitably.

Bitcoin mining has clearly come a long way since its early days, and is now entering a more complex phase, shaped by industrial-scale operations, rising competition, and tightening margins. 

While the price of BTC has surged from almost nothing to around $80,000 as of writing, the average cost of mining has also risen sharply and is now estimated to be close to the price of BTC itself.

As a result, large-scale miners are facing mounting pressures, from increasing hash rates and shrinking rewards to elevated energy and infrastructure costs. These forces, in turn, are driving a strategic pivot toward artificial intelligence (AI) workloads.

Large-Scale Miners Transitioning to AI for Higher ROI

Bitcoin is a decentralized digital currency that uses blockchain technology to enable secure peer-to-peer transactions without needing a central authority like a bank or government.

Unlike traditional financial services like bank transfers, Venmo, and PayPal (PYPL ), Bitcoin allows anyone from anywhere in the world to transfer value across borders. This is made possible by miners, which are specialized computers (ASICs) that validate and record transactions, release new BTC into circulation, and secure the Bitcoin network.

Bitcoin miners compete to solve complex math problems, and the first one to find the solution gets the chance to confirm the next block of transactions and receives fees plus BTC rewards.

In 2009, miners earned a hefty 50 BTC per block. But due to halving, which occurs every 4 years, rewards for each block have been significantly reduced. So far, Bitcoin has experienced four of these events, with the most recent occurring in April 2024, which reduced the reward to 3.125 BTC. The next halving in 2028 will further reduce it to just 1.5625 BTC.

Not only has miner revenue diminished considerably, but mining profitability has also declined.

At the time of writing, Bitcoin mining profitability is $0.0359 per day for 1 THash/s, according to data from Bitinfocharts. For most of 2025, profitability was above $0.05, while between 2022 and 2024 it was around $0.1.

Miner profitability depends on hardware efficiency, electricity costs, and Bitcoin’s price.

BTC price is currently down 36.5% from its peak of $126,000 hit in Oct. 2025. Meanwhile, energy costs have been fluctuating due to the Iran-U.S. conflict, which led to the closure of the Strait of Hormuz and higher oil and energy prices. At the same time, the capital expenditure required for maintaining competitive ASIC hardware continues to rise.

In addition to it all, Bitcoin’s hash rate, which is the total computing power used by miners to secure the network and serves as a performance indicator for mining machines, has been climbing to record levels. Bitcoin hash rate hit a new peak at 1.285 ZH/s in Sep. 2025.

Higher hash rates mean greater security for Bitcoin, but they also mean intense competition, making it challenging for miners to earn rewards.

These tightening business conditions in the mining industry led publicly traded Bitcoin mining companies, including Core Scientific (CORZ ), Riot Platforms (RIOT ), MARA Holdings (MARA ), CleanSpark (CLSK ), Bitdeer Technologies (BTDR ), and Cango, to sell more than 32,000 BTC in Q1 2026, more than the amount sold in all four quarters of 2025, according to TheEnergyMag.

This sale even exceeded the 20,000 BTC sold during Q2 of the 2022 bear market.

Bitcoin selling by miners came as hash price, a measure of miner profitability, hit a record low of $27.66 per petahash/second per day (PH/s) on Feb. 24, according to data from Hashrate Index. This has since increased to almost $39 (PH/s).

But according to asset manager CoinShares, about 15% of the global Bitcoin mining fleet remains unprofitable, especially those running older hardware. For them to be cash-profitable, they need access to power at sub-$0.05 per kWh.

Selling by miners isn’t anything out of the ordinary, though; they do that from time to time to cover their operating expenses. But rising energy costs and lower prices have forced far more offloading than miners would otherwise have held in their corporate treasuries.

Bitcoin held by miners has been declining over the long term. Most recently, the total number was sitting at about 1.8 million, while it was above 1.86 million BTC at the end of 2023.

The downturn isn’t just cyclical, according to CoinShares. The mining squeeze, driven by lower prices, rising network difficulty, and weak transaction fees, is increasingly limiting viable operators to those with advantages, such as more efficient fleets or access to low-cost power.

“We expect further capitulation among higher-cost operators in H1 2026 unless BTC’s price recovers materially.”

– CoinShares noted in its Q1 2026 Bitcoin Mining report

It’s pretty clear that Bitcoin miners are currently facing a structural squeeze. And in response to that, many large-scale mining firms have turned to AI and high-performance computing (HPC) as alternative or complementary revenue streams. 

From a revenue stability perspective, the pivot to AI is encouraging. Unlike Bitcoin mining, which relies on specialized ASICs with a single use case, AI workloads can run on GPUs and generate more predictable, contract-based income. AI data centers benefit from long-term demand driven by machine learning, cloud computing, and enterprise AI adoption, offering miners a potential hedge against Bitcoin’s cyclical volatility.

Due to the high demand for computing power as AI adoption continues to grow, AI hosting contracts provide steady cash flow to miners, which is particularly attractive during Bitcoin bear markets.

Mining Firms Reinvent Themselves as AI Infra Providers

The transition to AI data centers first gained momentum in 2023 and then accelerated through the last couple of years as publicly listed miners, which account for over 40% of the global hashrate, started repurposing part of their infrastructure for AI and HPC hosting.

Facing profit pressures and unsustainable mining costs, companies are either selling their BTC holdings or using debt to fund this shift.

For instance, Core Scientific has sold most of the 2,537 BTC it held at year-end 2025. In its annual report, the miner said that it “currently expects to monetize substantially all” of its bitcoin reserves this year to boost liquidity and fund its high-density colocation strategy.

The company has already signed multi-year hosting agreements with AI-focused clients, signaling a shift toward hybrid operations. This change has helped the company’s stock price jump over 69% year-to-date (YTD) and  176.74% in the past year.

TeraWulf (WULF ) is also keeping its holding strategy minimal to just about 15 BTC. It is currently focused on raising $900 million to fund the AI data center expansion. The company has priced 47.4 million shares at $19 each as part of the stock sale. As of writing, WULF is trading at $25.50, up 124% YTD and 744% in the past year.

Meanwhile, Hut 8 (HUT ), which holds 16,331 BTC, merged with US Bitcoin Corp and emphasized diversified compute infrastructure, including AI workloads. HUT is enjoying a 137.14% upside in its stock price this year and a 760.5% upside in the past year.

Just this week, the company signed a 15-year lease worth nearly $10 billion for its Beacon Point data center campus in Texas. The agreement covers 352 MW ​capacity in the first phase of the project, which is part of a planned 1 GW campus. Including annual rent increases, the contract could actually be worth more than $25 billion if renewal options are exercised.

Hive (HIVE ), originally focused on GPU mining before Ethereum’s move away from proof-of-work, is also leaning further into AI cloud services. However, HIVE’s stock price has seen a relatively small uptrend, up only 14% YTD. Its one-year returns meanwhile amount to almost 70%.

Then there’s Iren (IREN ), formerly Iris Energy, which has signed a multi-year cloud services contract with Microsoft (MSFT ). As a result, its stock price has risen 61.45% YTD and over 828% in the past year. The Bitcoin miner recently announced plans to acquire Mirantis, a cloud infrastructure and enterprise support services provider, for $625 million in stock.

Cipher Mining (CIFR ) has also entered into an agreement with Amazon to provide 300 MW of power capacity for AWS. Its stock price has increased by 48.44% year-to-date and 625.50% over the past year.

Last year, the company sold off its 49% stake in three Bitcoin mining joint ventures for about $40mln in stock. With that, Odessa remains its sole operating Bitcoin mining site, which CEO Tyler Page said is among the lowest-cost BTC producers with a fixed-price power purchase agreement (PPA) at $0.028 per kilowatt hour positions. The company currently holds 1,500 BTC.

This year, Cipher Mining signed its third data center campus lease, bringing its portfolio to roughly $11.4 billion in contracted revenue. 

The likes of Bitdeer Technologies, Bitfarms (BITF ), MARA Holdings, and CleanSpark are all treating Bitcoin as productive capital, using it to double down on AI infrastructure.

This makes sense given that Bitcoin miners are generating most of their revenue from AI, which CoinShares estimates will account for 70% of their combined revenue by the end of the year, up from 30% now.

“The long—term economics of HPC and AI data centers should trump Bitcoin mining. Just from a business operations standpoint, you get more visibility, better margin, and stronger cash flows out of the data center business.”

– Brian Dobson, a managing director at Clear Street, told Bloomberg. 

Bitcoin miners have already signed more than $70 billion in AI and HPC contracts that allow them to convert infrastructure into data centers.

But that’s not all. Several tech giants have now announced ongoing increases in AI investment, which may benefit Bitcoin mining companies that have transitioned to AI infrastructure.

Per Wall Street analysts’ estimates, total AI capital expenditures could climb to between $800 and $900 in 2026 and then surpass $1 trillion in 2027. “Cap-ex continues to soar as demand outpaces supply and pricing increases,” said analysts for Jefferies in a note to investors.

These projections are the result of Amazon (AMZN ), Microsoft, Google parent Alphabet (GOOG ), and Meta (META ) increasing their spending forecast for the year to $200 billion, $190 billion, $185 billion, and $135 billion, respectively.

The growing appeal of AI infrastructure amid deteriorating mining economics reflects a broader realization across the industry that traditional Bitcoin mining may no longer offer the margins or stability needed for long-term growth. And what began as a supplementary revenue strategy is fast evolving into a fundamental restructuring of the Bitcoin mining business model.

“This transition may prove to be the end of an era for some large US miners, not necessarily in terms of survival, but in how they operate, as they adapt away from models built for a different capital and energy market environment,” said Matthew Kimmell, an investment strategist at CoinShares. “Margins are just getting really thin, hash price kicking bottoms. It’s brutal out there.”

Will This Negatively Affect Bitcoin or Will This Increase Decentralization?

The migration of mining capacity from Bitcoin mining to AI data centers does have some effect on the largest cryptocurrency network, but it doesn’t inherently threaten its security or functionality. 

Bitcoin’s protocol is designed to adapt to changes through its difficulty adjustment mechanism, which recalibrates every about two weeks based on the total network hash rate. If a significant number of miners were to exit, the difficulty would decrease, making it easier and more profitable for the remaining participants to mine.

For instance, the global Bitcoin network hash rate dropped by 4% in Q1 of 2026. 

This first drop in six years reduced Bitcoin security, resulting in a downward adjustment in mining difficulty, the largest in over a year. As the difficulty to mine Bitcoin reduced to 125.8 T in mid-February, it became easier to mine BTC. This helped the Bitcoin hash rate recover to 997.04 EH/s, up from the 698 EH/s low in late January this year.

This self-correcting system ensures that Bitcoin continues to work as intended. So, a reduced hash rate may temporarily lower network security margins, but unless the decline is extreme and sustained, the system remains robust. 

Moreover, lower difficulty improves profitability for smaller or more efficient miners by reducing the computational power and energy required to solve blocks, thus inviting new entrants or allowing sidelined participants to return.

The impact of miners diversifying into AI is minimal, but only if a portion of large miners are engaged in this endeavor and they do so while maintaining core Bitcoin operations. If the exodus occurs on a broader scale, it could truly destabilize the network, at least in the short term.

This diversification isn’t all bad either. Many argue that this AI shift can improve the Bitcoin network’s decentralization by reducing the share of major miners and redistributing it among new entrants, potentially even individual ones.

For instance, a total of 30 public miners are currently contributing 416.26 EH/s to the Bitcoin network, up from 309.60 EH/s a year ago, which puts their share at 43.26%.

But while a case can be made for lowering the concentration among a handful of dominant firms, barriers to entry remain high. Even with reduced difficulty, mining still requires access to cheap electricity, efficient hardware, and operational expertise. 

While it’s possible for smaller participants to regain some footing, a full return to the early-era model is extremely unlikely under current technological and economic conditions.

There’s another question this AI craze among Bitcoin raises: whether this rotation is simply trend-chasing or a carefully evaluated strategic pivot. While the current surge in AI demand is fueled by generative models and enterprise adoption, businesses have yet to generate a profit. In fact, the majority of generative AI pilots at companies are failing.

Take Microsoft-backed OpenAI, which is behind the popular AI-powered conversational chatbot ChatGPT, which boasts over 900 million weekly active users, for instance. It reported a $5 billion loss in 2024 on about $3.4 billion in revenue and burned through $8 billion in 2025 on a $13.2 billion in revenue.

The report suggests the company is set to incur $14 billion in losses in 2026 and will continue to incur large losses totaling $44 billion until 2029.

The company is targeting approximately $600 billion in total compute spend by 2030 and has been busy making multibillion-dollar infrastructure deals. Its CEO, Sam Altman, touted $1.4 trillion in infrastructure commitments.

Anthropic, another major player, lost $5.3 billion in 2024 on revenues of $918 million. 

The field is very competitive and extremely capital-intensive. For Bitcoin miners, transitioning to AI requires significant upfront investment in GPUs, data center upgrades, and specialized talent. Meanwhile, their margins depend heavily on securing long-term contracts and maintaining high utilization rates. If AI demand growth slows or the supply of compute infrastructure exceeds demand, their returns could compress.

Additionally, companies that overallocate to AI may find themselves under-positioned for a rally in Bitcoin price, which has jumped more than 35% since falling under $60K in February.

So, while AI offers diversification and potentially steadier income, it is not a guaranteed superior alternative. It’s all about good execution and timely investments.

Conclusion

Bitcoin mining is no longer the simple, accessible activity it was in its early days. It has since evolved into a capital-heavy industry that is dominated by large-scale operators, who are now struggling with high production costs, lower spot prices, thin margins, and cyclical pressures. 

The recent shift toward AI is a response to this declining mining profitability and a bet on the future of compute demand.

For Bitcoin mining, this transition has meaningful implications. It could rebalance mining participation, introduce new revenue models, and reshape how mining companies operate, but it also introduces risks of operational complexity and potential overexposure to AI market cycles.

Looking ahead, the most resilient players will be those that take a balanced approach: leveraging AI to enhance returns while maintaining strategic exposure to Bitcoin’s long-term value proposition. This way, AI can help miners handle crypto’s high volatility without abandoning the core economics of the Bitcoin network.