Mining Artificial Intelligence: Why Bitcoin Miners Are Changing Their Business Model
Mining artificial intelligence sounds like a metaphor, but for the largest Bitcoin miners, it has become a practical business strategy. Companies that have spent years building farms for cryptocurrency are increasingly repurposing data centers for artificial intelligence, machine learning, and high-performance computing. We explore why mining has ceased to be the only source of income, what is happening with the hash rate, and how dangerous this restructuring is for the Bitcoin network.
What Does Mining Artificial Intelligence Mean
Mining artificial intelligence is not about extracting a separate AI coin or trying to "mine a neural network." In practice, it involves renting computational power for artificial intelligence tasks: training and running models, processing data, machine learning, and high-performance computing. The operator provides the space, electricity, cooling, network, engineers, and equipment, while the client uses these resources for calculations.
In simple terms, the process looks like this:
- The operator has a data center or mining site with access to electricity, cooling, and network.
- Suitable equipment is set up for AI tasks, most often GPUs rather than ASICs for Bitcoin.
- The client uploads computational tasks: training models, processing data, or running neural networks.
- The operator receives payment for access to the power, rather than a block reward as in Bitcoin mining.
How AI Computations Differ from Cryptocurrency Mining
The difference lies primarily in the purpose, equipment, and type of calculations.
- Bitcoin mining is needed for finding blocks and securing the network; AI computations are needed for training and operating artificial intelligence models.
- In Bitcoin mining, the main work is done by ASIC devices under SHA-256; for AI, GPU systems and server infrastructure are often required.
- ASICs excel at performing one narrow task, while GPUs are better suited for parallel computations needed in machine learning.
AI can assist miners not as a way to "find Bitcoin faster," but as a farm management tool: distributing loads, monitoring cooling, planning maintenance, and assessing profitability considering electricity prices and hash rates. However, if the cost of mining exceeds the market price of Bitcoin, AI alone does not turn mining into a profitable business.
Can You Rent Miners for AI Tasks and How Much Can You Earn
A typical ASIC for Bitcoin is almost unsuitable for AI computations: it is designed for SHA-256 and performs one narrow task. For renting for neural networks, GPU servers, reliable power, cooling, network, and a team capable of maintaining the equipment are needed. Therefore, large deals are made not by owners of a single miner, but by operators of sites like IREN or HUT 8, who rent out the infrastructure as a whole.
Profitability depends on electricity prices, equipment efficiency, network difficulty, and Bitcoin prices. The benchmark for Bitcoin mining is visible from the hash price: about $32 per day for 1 PH/s by August 2026. Meanwhile, CoinShares estimated the average cost of one Bitcoin for public miners in the U.S. at $79,995, with a market price of around $70,000, meaning that for some companies, mining was operating at a loss.
In the AI direction, revenue looks different: payment is not made for a found block, but for renting infrastructure. Therefore, amounts depend on the contract. IREN's expectations in this direction reach up to $1.9 billion in annual revenue, but this is the level of a large operator, not a home farm.
To mine 1 Bitcoin, a miner receives a share of block rewards proportional to their power. The higher the overall hash rate of the network and the weaker the equipment, the longer the path to 1 Bitcoin; with a home miner's device against an industrial network, this period stretches significantly.
Miners Transition to AI Infrastructure
Large public mining companies, primarily from the USA, are gradually moving away from a model where nearly all business relies on cryptocurrency mining. Instead, they are redirecting their facilities towards data centers for artificial intelligence and high-performance computing.
Notable participants in this shift include:
- IREN (Iris Energy).
- HUT 8.
- TerraWulf.
- Riot Platforms.
- Bitdeer.
- Cipher Digital.
- Core Scientific.
These companies are being closely monitored, as the demand for computing power is growing not only in the crypto industry.
Artificial intelligence requires vast amounts of energy, cooling, space, network infrastructure, and engineering expertise. Industrial miners already possess all of this. Therefore, for companies working with neural networks, cloud computing, and machine learning, former or partially repurposed mining sites have become a quick way to access power without building data centers from scratch.
Why Bitcoin Mining Has Become Less Profitable
Any mining operation lives by a simple logic: costs must be lower than revenues. After the halving in 2024 and the drop in Bitcoin's price at the end of 2025, this formula has become less effective for many major players. Cryptocurrency mining has continued, the Bitcoin blockchain has not stopped, but the economics of industrial farms have changed significantly.
According to CoinShares, in the last quarter of 2025, Bitcoin mining for public American mining companies became unprofitable on average. The cost of mining one Bitcoin was estimated at $79,995, while the market price hovered around $70,000. At one point, TerraWulf's costs for mining a single coin reached $385,000. With such figures, revenue no longer compensates for expenses.
Another telling indicator is the hash price. This index reflects the expected daily income of a miner with a power of one petahash per second. By August 2026, the figure was around historical lows—approximately $32. This is nearly half the level of a year ago, more than ten times lower than the peak in 2021, and over a hundred times lower than the maximum in 2017. Simply put, the same computing power brings miners less and less money with each cycle.
These figures do not mean that Bitcoin mining has become unprofitable everywhere and for everyone. Miners have different electricity prices, different conditions for purchasing ASIC equipment, taxes, rents, staff salaries, and maintenance costs. Moreover, technology quickly becomes obsolete, necessitating regular replacements. Mining becomes a problem primarily for those companies that cannot effectively control costs while simultaneously financing development.
Why Capital Is Flowing into AI Data Centers
It is important to understand: former Bitcoin miners are not turning into neural network developers. They are not creating a competitor to OpenAI or becoming operators of artificial intelligence models. Their role is different—to provide facilities, electrical infrastructure, cooling, and operational teams to those who need computing power.
The transition of miners to AI infrastructure is explained not by a trend in neural networks, but by simple economics: the power, energy, and cooling are already built, and the demand for computing from technology companies is growing faster than the profitability of Bitcoin mining.
The market is willing to pay amounts for such infrastructure that were previously difficult for mining companies to obtain solely from cryptocurrency extraction. HUT 8 has signed a $7 billion deal to lease capacity, including the River Bend farm in Louisiana. Technical partners include Anthropic, the creator of Claude, and the cloud platform Fluidstack. The financing is linked to Google, J.P. Morgan, and Goldman Sachs.
Key parameters of the IREN deal are as follows:
- Company: IREN.
- Partner: Microsoft.
- Deal amount: $9.7 billion.
- Equipment and service: access to AI systems based on Nvidia GB-300 chips in Texas.
- Additional purchase: graphics processors from Dell Technologies for $5.8 billion.
- Expected revenue: about $1.9 billion per year.
For tech giants, this approach also seems rational. Building a data center from scratch is time-consuming and expensive. Miners already have sites, connections to power grids, experience in operating equipment, and an understanding of how to manage high-density energy consumption facilities. Therefore, it is often more profitable for large capital to lease or develop infrastructure together with specialized operators.
-- Price
Why simply replacing the sign "Mining" with "AI" is not enough
The transition to artificial intelligence is not just about changing the sign at the entrance of a building with mining rigs. Industrial cryptocurrency mining primarily relies on ASICs—specialized equipment designed for specific computational tasks. An application-specific integrated circuit (ASIC) is designed to perform one function as efficiently as possible without wasting resources on unnecessary tasks.
ASIC miners differ depending on the hashing algorithm. Bitcoin uses SHA-256, Litecoin uses Scrypt, and other coins have their own algorithms. Therefore, devices are often incompatible with each other.
For example, Bitmain produces several models for different algorithms:
- Antminer S21: SHA-256.
- Antminer L9: Scrypt.
- Antminer D9: X11.
Such equipment is excellent for mining specific coins but poorly suited for tasks requiring flexibility.
Artificial intelligence and machine learning operate differently. They require different computing systems, most often built on GPUs. A graphics processor handles parallel computations well, but efficiency depends on the manufacturer, model, and architecture. This is not a case where simply replacing ASICs with standard Intel Core processors will yield a data center for neural networks.
A room that previously mined coins does not automatically become an AI cluster. To service computations for artificial intelligence, it is necessary to change the equipment, rebuild part of the engineering infrastructure, and manage the load differently.
At the same time, not every mining site fully transforms into an AI center. Some are building new specialized campuses, while some of the capacity remains for mining, and part goes to high-performance computing. Among the examples of such reorientation is the Bitdeer campus in the Norwegian municipality of Tydal.
What happened to Bitcoin's hash rate
The pivot of large miners towards artificial intelligence has indeed reflected on the Bitcoin network. From October 2025 to January 2026, the seven-day moving average hash rate decreased by 28%—from 1.15 to 0.83 zettahashes per second. This is a noticeable shift, but it does not seem unexpected: American mining companies account for up to 40% of the total Bitcoin hash rate.
By August 2026, the hash rate had begun to recover, fluctuating around 0.9 zettahashes per second. According to CoinShares' forecast, by the end of 2026, this figure could exceed 1.8 zettahashes per second, and by the end of 2027, it could reach 2 zettahashes per second.
The most important point: the transition to AI infrastructure does not mean a complete abandonment of Bitcoin mining by miners. In many cases, companies are not shutting down mining entirely but are simply reducing their investment pace in expanding the mining sector. Against the backdrop of large contracts with the AI sector, this seems logical: money goes where the expected margin is higher.
How Many Bitcoins Miners Could Have Missed Out On
It is almost impossible to accurately calculate the "missed" Bitcoins. Firstly, if miners had not engaged in AI infrastructure, they would not have received large contracts. This money cannot be automatically attributed to potential mining expansion. The same applies to borrowed capital: it is attracted under a specific business model, not just for abstract farm expansion.
Secondly, a decrease in production by one company does not always indicate a systemic turnaround. Riot Platforms mined 1,473 Bitcoins in the first quarter of 2026. This is only 57 coins less than in the same period of 2025, meaning a decrease of 3.7%. Such a difference resembles a normal fluctuation rather than a sharp abandonment of mining.
There is also a network effect. If IREN, TerraWulf, or HUT 8 mined fewer coins, it means that fewer Bitcoins were credited to their balances. However, the coins themselves do not disappear: the block reward remains set by the protocol, and after the difficulty recalculation, it becomes slightly easier for other participants to mine blocks. On a global scale, the "unmined" Bitcoins by some miners are received by others.
How Revenue Structure is Changing
There are no exact figures for future revenue yet, but forecasts indicate a direction. CoinShares expects that the share of revenues from AI-related activities for mining companies could grow to 70%. At the beginning of the year, it accounted for about 30%. If this scenario materializes, mining will cease to be the main business for some major players and will become one of the directions alongside AI infrastructure.
This does not equate to shutting down farms. Rather, companies are reassessing their priorities. As long as Bitcoin mining provides an acceptable margin, it will be maintained. However, new investments are increasingly going where contracts are larger, cash flow is more predictable, and demand from the tech sector continues to grow.
Why Miner Stocks Have Risen, but Risks Remain
The stock market has generally reacted positively to the transition of major cryptocurrency miners to AI infrastructure.
Stock dynamics since the beginning of the year:
- IREN: +5.69%.
- TerraWulf: +45.92%.
- MARA Holdings: +9.49%.
- Riot Platforms: +57.72%.
- Cipher Digital: +14.76%.
- Bitdeer: -4.73%.
Investors see the AI direction as a new source of revenue, but concerns have not disappeared. The main question is whether miners can actually earn enough to cover current and future expenses for upgrading facilities, purchasing equipment, and servicing contracts.
The second issue is debt. The transition from cryptocurrency mining to AI infrastructure requires enormous investments. TerraWulf raised about $5.7 billion in debt financing, while IREN raised about $3.7 billion. For shareholders, such a burden appears risky, even if the AI market promises high returns.
Conclusion
Major mining companies are changing their model not due to a rejection of Bitcoin as an idea, but because of economics. Mining has become more expensive and less predictable, while the demand for data centers for artificial intelligence is rapidly growing. Miners already have sites, energy capacities, cooling, and experience in operating complex infrastructure, so they naturally find themselves among the beneficiaries of the AI boom.
For the Bitcoin network, this turnaround does not yet seem like a critical blow. The hash rate has dropped but has begun to recover, and the difficulty mechanism redistributes opportunities among participants. For the companies themselves, the consequences are deeper: mining for many of them is transforming from a sole business into part of a broader infrastructural strategy, where artificial intelligence becomes the main driver of revenue.
This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.
You may also like

EIP-8363 Quantitative Review: What Does Ethereum Want to Regain by Cutting Staking 'Subsidies'?

AI Agent's 'Coming of Age': What Step Is Missing from Simulation Training to Real Trading?

CFTC Considers the Viability of Perpetual Futures in GPU Computing

When Computing Power Becomes a Trading Asset - From Computing Power Futures to Computing Power Dollars

Security Experiment for ML-DSA Institutional Transfers Begins on NEAR Testnet

EY Han Young and Upstage Sign Agreement to Support Sovereign AI Adoption

NoOnes crypto platform enters withdrawal-only mode, users urged to withdraw funds

Four years ago, a cryptocurrency mogul disappeared; now his successor has vanished

Will Crypto-Trump Cause a Pump?

Ika-Based Wallet Expands for AI Agents

Trump defends his political endorsements despite ten defeats in Republican primaries

What is a public key in crypto? Keys, addresses, and signatures explained

What is Etherscan and how do you use it?

Bitcoin Ripped 20% To $80K — Is $100K Next, or Is a Brutal Retest Coming?
Bitcoin tore through the mid-$60,000s, cleared $70,000, blew past $75,000, and briefly touched $80,000 before easing back to around $79,000 — a 20%+ move in a single week. The mood across crypto flipped almost overnight. But a violent rally is only the opening act. The more explosive the move, the more it calls for a clear, calm mind.

Illegal Mining in Russia: Why Home Farms Are Easily Detected

Cryptocurrency Exchange in Russia: Top 3 Services for 2026

Hyperliquid's Path to U.S. Compliance: From No License to Licensed HIP-3

Cold Storage of Bitcoins in 2026: Why Hardware Wallets No Longer Seem Invulnerable

$215 Billion Surge in Altcoin Market Value; Has Altseason Begun?

Hacker Who Leaked GTA 6 Created a Coin, Soared 20 Times in a Day

CVM Brings Tokenization Agenda to NYSE, Nasdaq, and SEC

Two U.S. KC-135 Tankers Depart from Bulgarian Base

The Definitive Difference Between Option Analysis and Technical Analysis | Why Can We Target Market Peaks and Bottoms?

Waymo Releases Partial Response to Child Collision Investigation

US Expands Treasury Buyback, Causing Market Turmoil and Potentially Repeating Japan's Currency Depreciation

Can the Lido EarnUSD Vault Boost Aave's Liquidity?

Binance continues EU onboarding despite missing MiCA licensing deadline: report

SEC’s latest crypto rules only open a few of Wall Street’s ‘million doors’ – Bitwise CIO Matt Hougan

If an Iranian Exchange Goes Bankrupt, Is the User the Owner of the Assets or a Creditor?









