Buy a bitcoin miner and you are not betting on cryptography. You are betting on cheap electricity, purpose-built silicon and a coin price you do not control, wrapped inside an energy business that halves its own revenue on a four-year clock.
Key takeaways
- It is an energy business, not a tech bet. A miner’s profit is bitcoin produced multiplied by spot price, minus the power bill. With an all-in cash cost to mine one coin of roughly USD 44,600 at Riot and USD 48,600 at MARA, the margin depends on two prices you do not set: the coin and the kilowatt-hour.
- The halving is the defining risk. On 20 April 2024 the block reward fell from 6.25 to 3.125 BTC, halving per-block issuance overnight. Analysts estimated 3 to 7 per cent of hashrate could go offline, up to 16 per cent if the price fell.
- Miner equities are a high-beta proxy on bitcoin. When the price fell in early 2026, MARA sold 15,133 BTC and Riot sold 3,778 BTC to cover costs. The equities swing harder than the coin in both directions.
- The AI pivot is the new frontier, and a new risk. Core Scientific went from bankruptcy to a USD 6.7bn partnership in eight months by leasing its power to AI data centres, then its shareholders rejected a USD 9bn buyout in the same cycle.
- The routes reach beyond one exchange. A pure-play miner ETF such as WGMI charges 0.75 per cent on Nasdaq; UK and European investors can reach the same trade through London-listed miners like Argo Blockchain, Frankfurt and Singapore names, or a physically backed bitcoin ETP on the London Stock Exchange. Most of these vehicles hold no spot bitcoin, so the exposure is to the business, not the coin.
The 60-second version
Bitcoin mining began as a hobbyist’s side project and is now an industrial energy business. The activity is simple to state and hard to do well. Purpose-built computers race to solve a cryptographic puzzle, the winner adds the next block to the Bitcoin ledger, and the network pays them newly issued bitcoin plus the transaction fees inside that block. Underneath that sentence sits a global fleet of machines drawing roughly the electricity of a mid-sized country, housed in warehouses next to power plants, wind farms and stranded gas wells. When you invest in bitcoin mining, you are betting on the economics of cheap electricity, purpose-built silicon, and a coin price you do not control.
That reframing matters because the numbers have grown up. The global miner market was worth USD 11.19bn in 2024 and is projected to reach USD 94.14bn by 2033, a compound annual growth rate of 26.7 per cent, on Data Bridge‘s figures. The largest operators now describe themselves less as coin-hunters and more as power companies, and the chairman of the biggest listed miner has said so directly. Mining is an infrastructure story wearing a cryptocurrency costume, and the two halves behave very differently as investments.
It also runs on a hard clock. Every four years the network halves the reward paid per block. That cuts the industry’s core revenue overnight while the cost of competing keeps climbing. Deflationary supply meeting an ever-harder puzzle is what separates the operators who survive from the ones who quietly switch their machines off.
I. What it is
Bitcoin mining is the process of validating transactions and minting new bitcoin by expending electricity on cryptographic computation. It is the mechanism that keeps the Bitcoin network honest without anyone in charge. Machines called ASICs (application-specific integrated circuits, chips built to do one calculation and nothing else) guess trillions of numbers per second, searching for one that satisfies the network’s difficulty target. Find it, and you win the right to write the next block and collect the reward.
The reward has two parts. The first is the block subsidy, freshly created bitcoin issued to the winning miner, currently 3.125 BTC per block since the April 2024 halving. The second is transaction fees, what users pay to have their transactions included. Fees are usually small, but during unusual periods they spike hard. Some blocks around the April 2024 halving carried fees exceeding 10 BTC, well above the subsidy itself, per Bitcoin Magazine. That is not sustainable, but it hints at how the industry’s revenue model is meant to evolve as the subsidy shrinks toward zero over the coming decades.
The collective computing power pointed at the network is called hashrate, the total number of guesses per second the whole industry makes, measured in exahashes (EH/s, a quintillion hashes per second). By 2025 the network stabilised near 1,004 EH/s on a 30-day average, up from roughly 600 EH/s earlier in the year, on Coinlaw‘s data. That figure is the security budget of the entire system. The more hashrate honest miners contribute, the more prohibitively expensive it becomes for an attacker to rewrite history. Strip away the jargon and a miner is a business that buys electricity and hardware, converts them into hashes, and sells the resulting bitcoin. Everything else follows from that.
For the investor, the crucial distinction is that you rarely buy the machines yourself. You buy the companies that own them, or a fund that owns those companies. That means your return depends not only on bitcoin’s price but on management’s skill at securing cheap power, deploying the newest chips before rivals, and surviving the halving that arrives on schedule whether they are ready or not.
II. Market history and growth
Mining’s arc runs from a laptop CPU in 2009 to warehouse-scale industry in 2026. The economics forced that industrialisation. As more machines competed, difficulty rose, and only operators with the cheapest power and densest hardware could stay profitable. What began as something anyone could do on a home computer became something only industrial operators with megawatt power contracts could afford, the same concentration curve that turned wildcat oil drilling into a business dominated by a handful of majors. The market data now reflects a genuine capital-goods sector. The global miner market reached USD 11.19bn in 2024, is projected at USD 14.17bn for 2025, and is forecast to hit USD 94.14bn by 2033, a 26.7 per cent CAGR, on Data Bridge‘s numbers.
Underneath the whole-market figure sits the hardware itself. The bitcoin mining hardware sub-market was USD 645.62m in 2025, rising to USD 731.36m in 2026 and reaching USD 2.25bn by 2035 at a 13.28 per cent CAGR, on Global Growth Insights‘s estimate. ASICs are the pick-axes of this gold rush, and whoever controls their supply controls a large share of who can compete.
Revenue tells the operating story. Miner total revenue was roughly USD 11.2bn in 2025, up 7.1 per cent year-on-year, with mining contributing around 66 per cent of the wider crypto-sector mining revenue of about USD 13.5bn, on Coinlaw‘s data. Capital keeps arriving too. Global investment in mining hardware is set to reach around USD 4.5bn in 2025, up 12.5 per cent year-on-year.
The single most important event in this timeline is the halving. The fourth halving occurred on 20 April 2024 at block height 840,000, cutting the subsidy from 6.25 to 3.125 BTC, per The Block. Overnight, every operator’s core revenue was halved while their costs stayed put. History says the price often rises afterward and eventually more than compensates, but there is no guarantee it does so on any given cycle, and the gap between the cut and any recovery is where weak operators die. That four-year clock is the metronome the entire industry marches to.
III. Demand drivers
The demand driving this industry is not really demand for bitcoin. It is demand for a particular kind of energy consumer, and that reframing is where the infrastructure thesis lives.
Start with the scale of consumption. Bitcoin draws roughly 211.58 TWh annualised, about 0.83 per cent of global electricity, on the Cambridge Centre for Alternative Finance’s index, with methodologies ranging from about 155 to 212 TWh, per Coinlaw. That power draw rivals whole nations, comparable to Poland or Argentina and exceeding Ukraine’s roughly 160 TWh, on EZ Blockchain‘s comparison. A consumer that large is not a footnote to a grid. It is a force that reshapes it.
Critics tend to stop at the consumption figure, but the mix is what matters for the investment case. Bitcoin’s energy is now 52.4 per cent non-fossil, split across hydropower 23.4 per cent, wind 15.4 per cent, nuclear 9.8 per cent, solar 3.2 per cent and other 0.5 per cent, on Coinlaw‘s breakdown. Miners chase the cheapest electricity on earth, and the cheapest electricity is frequently power that would otherwise be wasted: hydro that spills in the wet season, wind curtailed at night, gas flared at the wellhead. A machine that can switch on wherever surplus power appears, and switch off the instant that power is needed elsewhere, is a genuinely useful thing to a grid operator.
That flexibility has a price, and grids are paying it. In August 2023, ERCOT paid Riot Platforms USD 31.7m to curtail in a single month, split between USD 24.2m in power-curtailment credits and USD 7.4m in demand-response, per ABC13 and San Antonio Current. Those curtailment credits ran around 15 per cent of Riot’s electricity cost in 2024. The grid paid a miner tens of millions to stop mining when the state needed the electricity, a revenue line that has nothing to do with bitcoin’s price and everything to do with being useful infrastructure. The same logic is playing out beyond Texas. In Ethiopia, the state power utility is selling surplus hydropower from the Grand Ethiopian Renaissance Dam to miners rather than letting it sit idle, and in Paraguay miners buy Itaipú Dam surplus at power costs as low as USD 0.033 per kWh.
The US Energy Information Administration frames miners as exactly this kind of resource, in the regulator’s own language:
“a distinctive type of flexible demand-side resource that can help balance supply and demand and improve reliability by quickly curtailing energy consumption.”
When a national energy regulator describes your industry as a grid-balancing asset rather than a nuisance, the demand driver stops being whether people want bitcoin and becomes whether grids keep paying for flexible load. Increasingly, they do.
IV. The players
Mining is now dominated by a handful of large, listed operators, plus the people who run them. Naming them matters, because their strategies diverge sharply and the divergence is the investment decision. The best-known names carry a US listing, but the field is genuinely global once you look past the biggest tickers.
MARA Holdings (NASDAQ: MARA) is the largest listed pure-play by hashrate, run by chairman and chief executive Fred Thiel. Thiel has spent two years re-labelling the company as an energy and compute business rather than a coin miner, and his framing is the clearest statement of where the industry’s leadership thinks it is heading. MARA mined 2,011 BTC in Q4 2025 and grew energised hashrate to 66.4 EH/s, up 25 per cent year-on-year from 53.2 EH/s, per its Q4 results.
Riot Platforms (NASDAQ: RIOT) is the Texas-centric operator whose demand-response strategy earned those ERCOT payments. It represents the “own the power relationship” school of mining.
CleanSpark (NASDAQ: CLSK) scaled aggressively on hashrate, and, as the cautionary case study below shows, learned that scale and profit are not the same thing.
Core Scientific (NASDAQ: CORZ) is the AI-pivot story: a company that went bankrupt as a pure miner and re-emerged by leasing its power infrastructure to high-performance computing.
Outside the United States, the roster is real. Bitdeer (NASDAQ: BTDR) is headquartered in Singapore and is the vertically integrated outlier, because it makes its own chips. Argo Blockchain is a London-headquartered miner dual-listed on the London Stock Exchange (ARB) and Nasdaq (ARBK). Northern Data (Frankfurt: NB2) was Europe’s largest crypto miner before it sold its mining division for up to USD 200m in late 2025 to become a pure AI and high-performance computing operator, per Coinspeaker. The Canadian names have drifted south: Hut 8 now runs from Miami and Bitfarms redomiciled to the US and rebranded as Keel Infrastructure, trading under KEEL on both Nasdaq and the TSX from April 2026. Cipher (NASDAQ: CIFR) and Bitfufu (NASDAQ: FUFU) round out the mid-cap field.
The sell side has begun to differentiate these names rather than treat them as one bitcoin-beta trade. In its 2026 re-rating, JPMorgan upgraded Cipher and CleanSpark while trimming targets on MARA and Riot, per The Block, pricing AI-pivot optionality separately from raw hashrate. An investor sizing them up should read them the same way.
V. Geography
Mining follows cheap power around the planet, and the map has redrawn itself repeatedly as governments have welcomed or expelled the industry. The concentration is real: in the first quarter of 2026, the United States, Russia and China together controlled about 68 per cent of recorded global hashrate, at roughly 38, 17 and 13 per cent respectively, per Finbold.
North America is the centre of gravity. The United States commands roughly 35 to 40 per cent of global hashrate, on Coinlaw‘s data, driven by renewable access, deregulated power markets, state-level policy and deep institutional capital. Texas in particular has become the industry’s showcase, precisely because its grid pays flexible loads to curtail. The ERCOT payments to Riot are the clearest example of a jurisdiction actively designing miners into its energy system.
Russia retains a large share, around 17 per cent, anchored by cheap natural gas, Siberian hydropower and a cold climate that cuts cooling costs.
Central Asia and the Gulf. Kazakhstan absorbed a wave of miners after China’s 2021 ban, briefly reaching roughly 18 per cent of hashrate before energy caps and regulatory pressure pushed it down to about 2.1 per cent. Gulf states with cheap energy and sovereign ambition have moved in the other direction, building state-backed capacity.
Latin America and Africa attract miners to stranded hydro and flared gas. Paraguay accounts for roughly 4.3 per cent of global hashrate off Itaipú Dam surplus, on Hashrate Index‘s figures, a striking figure for a country of about seven million people. Ethiopia, powered by the Grand Ethiopian Renaissance Dam, has climbed to roughly 2.7 per cent of the network and the world’s eighth-largest mining base, per BitKE. This is the “monetise wasted energy” thesis in its purest form.
The Nordics and Canada offer cold climates that cut cooling costs and abundant hydro and geothermal power, making them structurally low-cost and low-carbon locations, though as noted the largest Canadian operators have redomiciled to the United States.
The geographic lesson for an investor is that a miner’s single most important asset is not its machines. Those depreciate and are replaceable. It is its power contracts and the jurisdiction hosting them. A fleet of the newest ASICs plugged into expensive or politically hostile electricity is worth less than an older fleet plugged into cheap, stable, welcoming power. When you analyse a mining company, you are really analysing an energy portfolio.
That portfolio is also a policy bet. The reason the United States now holds 35 to 40 per cent of global hashrate is not geological luck. It is that its states competed to welcome the load while other jurisdictions expelled it. China’s 2021 ban did not shrink the industry. It relocated it, mostly to North America, Russia and Central Asia, and the network’s climb back toward 1,004 EH/s is the proof that miners simply move to wherever power is cheapest and policy is friendliest. For an investor, that mobility cuts both ways. An operator concentrated in one jurisdiction carries the risk that a single election, tariff or grid-stress event rewrites its economics overnight. Geographic and power-source diversification inside a mining company is worth as much as a diversified fund is across companies.
VI. How to actually invest
You will not, in practice, run machines yourself unless you have industrial power access and technical staff. The realistic routes are listed equities and funds, and they differ in an important way. Most of these vehicles do not hold any bitcoin. They hold mining businesses. That distinction defines the exposure. Where you can buy them also depends on where you sit, so the routes below split by geography.
The cleanest US fund route is the CoinShares Bitcoin Mining ETF (NASDAQ: WGMI), actively managed, holding pure-play miners defined as companies deriving at least 50 per cent of income from mining, hardware or services. It does not hold spot bitcoin or futures, per ETF.com. It holds the miners. The expense ratio is 0.75 per cent, its top-10 holdings make up 83 per cent of the fund, and it ran roughly USD 336m in assets in mid-2026. It was renamed from the Valkyrie Bitcoin Miners ETF around 15 July 2025, per the SEC filing.
The US direct route is single stocks. The large-cap pure-plays trade under MARA, RIOT, CLSK, CORZ, BTDR, CIFR and FUFU. Both ETFs and single stocks carry no investment minimum beyond the price of one share.
UK and European investors have their own routes and do not need a US brokerage. On the equity side, Argo Blockchain trades on the London Stock Exchange as ARB alongside its Nasdaq line, Bitdeer is Singapore-based on Nasdaq, and Northern Data lists in Frankfurt as NB2 for those who want the AI-infrastructure end of the same trade. On the fund side, there is no pure-play miner ETF domiciled in Europe, because UCITS diversification rules block single-asset crypto funds, and no European equivalent of WGMI yet exists. What is available since the FCA lifted its retail ban on crypto exchange-traded notes on 8 October 2025 is physically backed spot-bitcoin exposure on the London Stock Exchange: the iShares Bitcoin ETP (IB1T) and the 21Shares Core Bitcoin ETP (CBTC) among them. That is exposure to the coin rather than the miners, which is a different bet, and these notes sit outside the Financial Services Compensation Scheme, so treat them as the higher-risk products the regulator labels them.
| Vehicle | Ticker | Where it trades | What you own | Holds spot BTC? | Fee / cost |
|---|---|---|---|---|---|
| Pure-play miner ETF | WGMI | Nasdaq (US) | Basket of listed miners | No | 0.75% p.a. |
| Large-cap miner (leader) | MARA | Nasdaq (US) | Single operating company | No | Brokerage commission only |
| UK-listed miner | ARB / ARBK | LSE + Nasdaq | Single operating company | No | Brokerage commission only |
| Vertically integrated (own chips) | BTDR | Nasdaq (Singapore-based) | Single operating company | No | Brokerage commission only |
| Physically backed BTC ETP | IB1T | LSE (UK/EU) | The coin, not the miners | Yes | c. 0.25% p.a. |
The choice between the fund and the single stock is the choice between diversifying away single-company risk and expressing a specific view. If you cannot distinguish an AI-pivot miner from a pure-hashrate miner, the basket is the more honest position. If you can, and you have a view on which strategy wins, the single stock is where that view pays. The miner routes in the table buy you the operating business; the European ETP buys you the coin, so the two are not substitutes.
VII. Unit economics
The business reduces to one equation. Revenue equals bitcoin produced multiplied by spot price. Profit is that revenue minus the power bill and the depreciation on the machines. A worked example from the most recently reported quarter shows how the maths lands in practice.
Start with Riot Platforms, Q1 2026. Its average all-in cost to mine one bitcoin, excluding depreciation, was USD 44,629, up from USD 43,808 in Q1 2025, per its 8-K. That cost rose year-on-year, because network difficulty climbed as rivals added hashrate. Riot produced 1,473 BTC and booked USD 111.9m in mining revenue, down from USD 142.9m a year earlier on a lower coin price and higher network hashrate.
If it costs roughly USD 44,600 in cash to produce a coin, then every dollar the spot price sits above that line is gross margin, and every dollar below it is a loss on production. At a bitcoin price near USD 90,000, or about GBP 71,000, the cash margin per coin is around USD 45,000 and the business prints money. At USD 50,000, the margin collapses to a few thousand dollars before depreciation, and after depreciation the operation may be underwater. The operator does not control which world it wakes up in.
The same picture emerges at MARA, Q4 2025. Its purchased-energy cost per bitcoin at owned sites was USD 48,611, at USD 0.05 per kWh, per its Q4 results. It mined 2,011 BTC on 66.4 EH/s of energised hashrate and reported USD 202.3m of revenue, down 6 per cent year-on-year. Two of the largest, best-capitalised operators on earth both sit in the USD 44,600 to 48,600 range to produce a single coin, which is the industry cost curve in miniature. The margin is a function of coin price and power price, and the low-cost operator survives conditions that bankrupt the marginal one.
The power price is the variable that moves everything. At USD 0.05 per kWh MARA produced a coin for USD 48,611; every one-cent move in the power price, and every halving, compresses that margin further. This is exactly why the stranded-energy jurisdictions matter. A Paraguayan operator buying Itaipú surplus at USD 0.033 per kWh sits well below MARA’s owned-site cost, which is the whole point of chasing wasted power. When the puzzle got harder after the 2024 halving, analysts estimated 3 to 7 per cent of hashrate could go offline, rising to as much as 16 per cent if the price fell, per Bitcoin Magazine. The machines drawing the most expensive electricity are always the first to switch off.
The number that anchors everything: it costs the biggest listed miners roughly USD 44,600 to USD 48,600, or about GBP 35,000 to 38,500, in cash to mine a single bitcoin. The spot price above that cost is margin, and anything below it is a loss on production. An investor can see a miner’s break-even clearly but has no control over the price it sells into.
VIII. Macro sensitivity
Because a miner’s revenue is coin-price times output minus power cost, the equities behave as a leveraged proxy on bitcoin, swinging harder than the coin in both directions. How they behave across different regimes is worth understanding before owning them.
| Regime | What happens to the coin | What happens to miners | Evidence |
|---|---|---|---|
| Rising price, cheap power | Bitcoin climbs, kWh stable | Margins expand faster than spot; equities outperform the coin | Cash margin per coin balloons above the ~USD 44,600 cost line |
| Falling price, spiking power | Bitcoin drops, energy costs rise | Margins vanish; miners sell held coin to fund operations | MARA sold 15,133 BTC (~USD 1.1bn) Mar 2026 (247WallSt); Riot sold 3,778 BTC Q1 2026 (source) |
| Post-halving squeeze | Price flat, subsidy halved | Marginal operators forced offline; survivors gain share | 3 to 7% of hashrate offline, up to 16% if price falls (Bitcoin Mag) |
| Structural diversification | Any | Revenue shifts toward power infrastructure & AI hosting | Leaders reallocate to owned power and HPC (Spark) |
The second row is the one investors tend to underweight. When the price fell in early 2026 amid an energy-price spike, both leaders were forced sellers of their own treasuries. MARA sold 15,133 BTC for roughly USD 1.1bn in March 2026 and Riot sold 3,778 BTC in Q1 2026 to cover costs. A business that must sell its product into a falling market to pay its electricity bill carries high operating leverage. In good regimes that leverage works in the shareholder’s favour. In bad ones it turns a coin drawdown into a much sharper equity fall.
The fourth row is the industry’s attempted answer. The regime shift from 2024 to 2026 has been described by Spark as miners becoming an “energy company that happens to mine bitcoin”, reallocating toward owned power and AI hosting to diversify away from pure block-subsidy revenue. Whether that diversification actually dampens the beta, or simply adds a second cyclical exposure on top of the first, remains unresolved.
Transaction fees add a volatile secondary revenue line. Around the April 2024 halving and the Runes launch, some blocks carried fees above 10 BTC, far above the subsidy, but such spikes are not sustainable and should never anchor a valuation.
IX. Tax
This section is jurisdiction-neutral and illustrated with one rule set. It is not tax advice. The mechanics vary by country and you should confirm your own position with a qualified adviser before acting.
The general principle that holds across most tax systems is that mining is treated as an income-generating activity, and it typically produces two separate taxable events rather than one.
The first event is receipt. Mined coins are ordinarily taxed as income at fair market value on the date they are received, regardless of whether you sell or hold them. That receipt-date value then becomes your cost basis. The US framing follows the IRS treatment of digital assets; the UK applies a broadly similar logic through HMRC’s crypto-asset rules, and most jurisdictions land in the same place, income on receipt.
The second event is disposal. When you later sell, you realise a capital gain or loss equal to the sale price minus the cost basis, that receipt-date value. That is two layers taxed at potentially different rates, income on the way in and capital gains on the way out.
How the activity is structured changes the arithmetic. Run mining as a business and the income goes on a business schedule, subject to self-employment tax, but equipment, electricity and software become deductible. Run it as a hobby and the income is taxed with no business deductions available. The globally portable principle is that device, power and depreciation deductibility hinges on the business-versus-hobby classification.
The double-taxable-event structure carries a trap. Because coins are taxed as income at fair-market value on the day they are received, a miner who holds through a falling market can owe income tax on a value the coins no longer have. The tax bill is struck on the receipt-date price, while the coins in the wallet are worth less by the time the bill is due. That mismatch is exactly the squeeze that pushed operators into forced coin sales when the price fell in early 2026. For the investor buying miner equities rather than mining directly, none of this applies to the mining activity. You are taxed on the shares like any other equity, which is simpler, but it means the tax-advantaged deductibility of power and hardware accrues to the company, not to you. The upside of the equity route is one clean layer of tax on a security. The downside is you inherit none of the deductions that make the underlying business work.
X. Case studies
Case 1, Bitdeer (BTDR): the vertically integrated operator. Bitdeer, headquartered in Singapore, manufactures its own SEALMINER ASICs rather than buying from a rival, which means it controls the single scarcest input in the industry. The results show what that control is worth: Q3 2025 revenue of USD 169.7m, up 173.6 per cent year-on-year, with self-mining hashrate scaled to 55.2 EH/s by December 2025 and 636 BTC mined in December alone, up 339 per cent year-on-year, per its 6-K. When everyone else is queuing for the same hardware, the company that makes its own can deploy at its own pace.
Case 2, Core Scientific (CORZ): the AI-pivot recovery. Core Scientific filed for Chapter 11 bankruptcy in December 2022 as a pure miner. It re-listed in January 2024, up 140 per cent on relisting, by pivoting its power infrastructure to AI and high-performance computing. It signed 12-year CoreWeave leases worth around USD 10bn in revenue and moved from “bankruptcy to a USD 6.7bn partnership in eight months”, per NBC, later seeking a USD 3.3bn bond sale in April 2026 to fund the build-out. What that shows is that a miner’s real asset is its power and its buildings, and those can be repointed at a more valuable buyer than bitcoin. Northern Data in Frankfurt reached the same conclusion more bluntly, selling its mining arm outright to become a pure AI operator.
Case 3 (cautionary), CleanSpark (CLSK): scale on its own does not produce profit. CleanSpark did everything the growth playbook demands. It ran a healthy 47.3 EH/s average hashrate in March and held 13,561 BTC. It still went backwards, with Q1 2026 revenue falling 25 per cent year-on-year to USD 136.4m and the net loss widening to USD 378.3m, from a USD 138.8m loss the prior year, on Bailey‘s balance-sheet analysis. Impairments on the value of held coin and dilution from raising capital can sink even a large, high-hashrate operator. Scale bought market share without buying profit.
The AI pivot in Case 2 is not a one-way street either. Core Scientific’s shareholders rejected CoreWeave’s USD 9bn all-stock buyout, and the merger was terminated on 30 October 2025. The miner-to-AI trade is real, but so is the deal risk inside it, and an announced partnership is not the same as a closed one.
XI. The core constraint
For a bitcoin miner, the one variable that decides survival is not the coin price, not the hardware, not even management. It is the cost of electricity relative to the halving-shrunk reward.
The obvious alternatives fail the test. Coin price is a market you take, not a lever you pull, and every miner faces the same price. Hardware is broadly available, and the newest chips reach everyone eventually. What differs between a survivor and a casualty is the price they pay per kilowatt-hour, because that number is fixed into long-term power contracts and cannot be renegotiated when conditions turn. When the halving cut the subsidy from 6.25 to 3.125 BTC, it did not touch anyone’s power bill. It simply halved the revenue that bill had to be paid from. The operator on cheap, contracted power absorbed it. The operator on expensive spot power was pushed toward the 3 to 7 per cent of hashrate that goes offline.
This is why the whole piece keeps returning to the power contract. When you evaluate a mining investment, the constraint you are really underwriting is whether this operator can keep producing coins below the market’s cost curve after the next halving cuts the reward again. If yes, it survives every regime and gains share when rivals fail. If no, it is a call option on the coin price with a running electricity bill attached, and the running cost bleeds the position over time.
The constraint is sharper still because the network is self-correcting in a way that punishes the marginal operator automatically. As rivals add machines, difficulty rises, which is precisely why Riot’s cost to mine a coin climbed from USD 43,808 to USD 44,629 year-on-year even as it ran the same machines. Every efficient operator that switches on makes life harder for the inefficient one, so the constraint is not static but tightens over time. An operator merely at the industry-average power price today is drifting toward the wrong side of the cost curve tomorrow, unless it keeps securing cheaper power or newer chips. The binding constraint is therefore not cheap power on its own but power that stays cheap relative to a curve that is always moving against it, which is a harder test than it first appears and a large part of why so few miners survive multiple cycles.
XII. Inside the asset
Step inside a modern mining operation and the infrastructure framing stops being a metaphor. What you are looking at is a data centre built to convert the maximum electricity into hashes at the lowest possible cost per hash.
The physical asset is a warehouse, increasingly sited next to a power source rather than a population centre, filled with racks of ASICs. These chips run hot and constant, which is why cold climates, immersion cooling and cheap power co-location dominate the industry’s geography. The machines depreciate fast, both physically and competitively, because a chip that is best-in-class today is uneconomic within a couple of hardware generations, which is why hardware investment keeps running around USD 4.5bn a year. The fleet is a depreciating asset on an accelerating treadmill: the hardware market compounding at 13.28 per cent reflects operators being forced to keep buying just to hold their position.
But the enduring asset underneath the machines is the power interconnect and the site. That is what Core Scientific discovered when it repointed its infrastructure at AI workloads: the substation, the cooling, the grid connection and the building outlast any generation of chips and can be repurposed for a higher-value tenant. It is also what makes miners valuable to grids, a large flexible demand-side resource that can curtail quickly. The security this all produces, a network humming at roughly 1,004 EH/s, is the product miners actually sell to the Bitcoin network, and the electricity is the raw material. Seen from inside the facility, the coin is almost incidental to the real business, which is the power.
XIII. The central dilemma
The dilemma at the heart of this asset class is whether you are buying a bitcoin proxy or an energy infrastructure business, because the leaders are trying to be both, and the two theses want different things.
The pure-miner thesis is a leveraged bet on bitcoin’s price. It wins spectacularly in bull regimes and is brutalised in bear ones, as MARA’s forced sale of 15,133 BTC showed. The infrastructure thesis is a bet on cheap power and flexible load, steadier, less spectacular, and increasingly monetised through grid payments and AI hosting rather than block rewards. The industry’s own leadership has picked a side. Here is Fred Thiel, chairman and chief executive of MARA Holdings, on the reframing:
“We’re not just a bitcoin miner. We’re becoming a digital energy and infrastructure company.”
Thiel’s framing captures the dilemma directly. If he is right, the sector deserves an infrastructure multiple and its beta to bitcoin should fall over time. If he is early, and the AI leases and grid payments never scale enough to offset the halving-shrunk subsidy, then the “energy company” language is aspiration and an investor is still holding a leveraged coin proxy that happens to own some real estate. The evidence runs both ways. Core Scientific’s USD 10bn of CoreWeave leases support Thiel, while its rejected buyout undercuts him. Buying a miner today means buying that unresolved question.
XIV. The next frontier
The frontier for this asset class is the collision between mining and artificial intelligence, and it is reshaping the investment case in real time.
The logic is straightforward. An AI data centre and a mining facility want the same three things: abundant power, a grid interconnect and cooling. Miners spent a decade building exactly that infrastructure in exactly the cheap-power locations AI now needs, so the incumbent miners are repurposing their sites for high-performance computing tenants who pay far more per megawatt than the block reward ever did. Core Scientific’s 12-year, roughly USD 10bn CoreWeave leases and its USD 3.3bn bond raise to fund the AI build-out show the pattern, and it is not only a US pattern: Northern Data in Frankfurt exited mining entirely to chase AI compute, and the redomiciled Canadian names rebranded around the same pivot. The market is already pricing this optionality separately, with JPMorgan’s 2026 re-rating upgrading the names with AI leverage and trimming the pure-hashrate names.
The same frontier can hurt as easily as help, as the buyout collapse between Core Scientific and CoreWeave showed: the AI pivot rescued a bankrupt miner and then a related deal fell apart mid-process. The question the next few years will answer is whether the best mining companies are quietly becoming the cheapest way to own AI-adjacent power infrastructure, or whether they are stacking a second speculative cycle on top of the first. The infrastructure of power, cooling and interconnect is genuinely valuable. Whether the miners capture that value or hand it to better-capitalised data-centre operators is still open.
XV. Lessons from history
Mining rhymes with older resource industries, and their histories carry usable warnings.
The gold rush and the pick-axe. The durable lesson of every rush is that the reliable money is often in the tools, not the digging. Bitdeer’s 173.6 per cent revenue jump from controlling its own SEALMINER chip supply is the modern pick-axe. When the hardware market is compounding at 13.28 per cent, owning the supply chain is a structurally different, and often steadier, business than racing to dig.
Commodity producers and the cost curve. Miners behave like any commodity producer. The price is set by the market, and survival is decided by where you sit on the cost curve. When bitcoin fell in early 2026, the operators forced to sell their own coin to fund operations were the high-cost producers doing what high-cost oil and copper producers always do at the bottom of a cycle. The USD 44,600 to 48,600 cost line is the same analytical tool a mining analyst would apply to a copper mine.
Scale is not a moat. CleanSpark’s USD 378.3m net loss on a healthy 47.3 EH/s restates the oldest lesson in capital-intensive industry, that production capacity you cannot fund profitably is a liability rather than an asset. Growth financed by dilution and haunted by impairments has bankrupted producers in every commodity cycle, and mining runs on the same arithmetic.
XVI. The case for it
The strongest version of the bull case treats a miner as cheap, flexible power infrastructure that happens to carry a call option on bitcoin. The numbers behind it are real.
The sector is growing into a real market. A 26.7 per cent CAGR toward USD 94.14bn by 2033 is not a fad’s trajectory. It is a capital-goods industry maturing.
The infrastructure earns money independent of the coin. ERCOT paying Riot USD 31.7m in a single month to curtail, roughly 15 per cent of its 2024 electricity cost, is revenue with no bitcoin exposure at all, and the same stranded-energy model is now earning hard currency for Ethiopia. When a national energy regulator classifies your industry as a grid-balancing resource, the business has an infrastructure dimension that sits apart from the coin trade.
The AI optionality is real and being paid for. Core Scientific’s USD 10bn of AI leases show the power and buildings can be repointed at the highest-value tenant of the decade.
The leverage cuts upward too. Because a miner’s margin sits above a visible ~USD 44,600 cost line, a rising coin price expands profit faster than it lifts the coin, which is the reason investors accept the downside beta in the first place.
XVII. The risks
Every one of these risks has already cost investors money in the last two years, which is why they belong at the front of the analysis, not the footnotes.
The halving is a scheduled revenue cut. The April 2024 halving cut the subsidy in half overnight, and another arrives on schedule roughly every four years. Any investment whose core revenue is designed to halve is structurally different from one whose revenue can grow.
The operating leverage is brutal in reverse. When the coin fell in early 2026, MARA sold 15,133 BTC and Riot sold 3,778 BTC to keep the lights on. Forced selling into weakness is the signature of a fragile business model.
Scale does not guarantee survival. CleanSpark’s USD 378.3m net loss proves a large, high-hashrate operator can still destroy shareholder value through impairments and dilution.
Deal and pivot risk is live. The collapse of Core Scientific’s USD 9bn CoreWeave buyout shows the AI escape hatch can jam shut.
Jurisdiction risk is real and moves fast. Kazakhstan went from roughly 18 per cent of global hashrate to about 2.1 per cent once energy caps and regulation bit, and Paraguay and Ethiopia both face proposed tariff increases that could reshape their economics. A miner concentrated in one country carries that risk in full.
Rising costs, thinner margins. Riot’s all-in cost per coin rose year-on-year to USD 44,629 as network difficulty climbed. The cost of competing keeps rising even as the reward keeps halving. Underneath everything sits the coin’s own volatility and the regulatory and energy-policy risk of an industry consuming roughly 0.83 per cent of global electricity. None of these risks is hypothetical. Each one has already played out in the last two years.
XVIII. The Alternative Fortune verdict
Bitcoin mining is a genuine infrastructure and energy business wrapped around a volatile commodity. The infrastructure side is real, in the form of power contracts, grid interconnects, cooling and buildings that outlast the machines and can be repointed at AI tenants worth billions. The volatility is just as real, in a revenue line designed to halve, a cost line that keeps rising, and operators forced to sell their product into falling markets to pay the electricity bill. What you are actually buying is high-beta exposure to bitcoin with a maturing infrastructure business underneath, and the mix varies enormously from one operator to the next.
Where the edge actually sits. The edge is not in owning a miner in general but in distinguishing the low-cost, well-contracted operator from the marginal one, because the USD 44,600 to 48,600 cost curve decides who survives each halving and who switches off. The second edge is the infrastructure-versus-coin question. The operators genuinely becoming energy and compute businesses may deserve a different multiple than the pure coin proxies, and the market is only beginning to price that split. An investor who cannot tell those apart is usually better served by the diversified basket than by a single name. And note where you sit: US investors get a pure-play miner ETF; UK and European investors reach the trade through individual London, Frankfurt and Singapore-listed names, or buy the coin itself through a London-listed bitcoin ETP, which is a different exposure again.
Questions to ask, by vehicle:
- If you are buying a pure-play miner ETF (for example WGMI): What is the expense ratio, how concentrated is the top-10 holding, and am I comfortable that it holds miners rather than spot bitcoin, so I am buying the operating businesses, not the coin?
- If you are buying a single large-cap miner: What is its all-in cash cost per coin and its power price per kWh, and where does that put it on the industry cost curve after the next halving?
- If you are a UK or European investor: Am I buying a listed miner (Argo in London, Bitdeer in Singapore, Northern Data in Frankfurt), or a physically backed bitcoin ETP on the London Stock Exchange, and do I understand that the first buys the business while the second buys the coin, with no FSCS protection?
- If you are buying the AI-pivot thesis: Is the AI lease signed and funded, or announced and still exposed to deal collapse?
This is a category-level analysis of bitcoin mining as an asset class, not a recommendation of any named company. Bitcoin mining sits within the wider world of digital assets alongside other alternatives. Do your own due diligence and seek advice suited to your own circumstances before committing capital.
