Mining is described as producing coins, which makes it sound like a factory. It is closer to selling electricity converted into a lottery ticket, in a competition where every improvement anybody makes is immediately taxed away from everybody, and where the price of the output is set by somebody else entirely.
A miner converts electricity into attempts at solving a problem, and is paid when one attempt succeeds. The revenue is the block subsidy plus the transaction fees in that block, and the probability of success is proportional to the share of the network's total computing power the miner controls. Over enough blocks, revenue is that share multiplied by total network revenue, which makes it a commodity business with unusually clean arithmetic.
Nothing about that business is under the operator's control except costs. The price of the output is set by a market, the difficulty of production is set by what competitors do, and the issuance schedule is set by the protocol. A miner is a price taker in every direction at once, which is a rare and demanding position for any business to occupy.
The consequence is that mining is not a bet on the asset in the way people assume. It is a bet on the spread between the price of computation and the price of the asset, which is a different quantity that can move against you while the asset rises. That distinction explains most of what is confusing about how mining companies perform.
A miner controls costs and nothing else. Price, difficulty and issuance are all decided elsewhere, and all three move.
Three inputs dominate and they behave differently. Electricity is the largest continuing cost and the one where competitive advantage lives, since a difference of a few percent in the rate compounds across every hour of operation. Miners locate where power is cheap, which is why the industry clusters around stranded generation, seasonal surplus, and grids with low marginal costs.
Hardware is the largest one-off cost and it depreciates in a specific and brutal way. Machines do not wear out so much as they become uncompetitive, because newer models convert electricity into attempts more efficiently, and difficulty rises to absorb the improvement. A machine's economic life ends when its electricity cost per attempt exceeds what the network pays, which can happen while the machine still works perfectly.
Hosting and infrastructure are the third: buildings, cooling, connection to the grid, and the staff to keep it running. This is where operational competence separates operators more than any technical choice, because the machines are the same machines everywhere and the difference is uptime and the cost of achieving it.
The energy question deserves one honest paragraph, because it is argued badly in both directions. A miner is an unusual customer for a grid: the load is large, constant, and can be switched off within seconds without damaging anything, which makes it valuable to operators managing supply and demand. That flexibility is why miners increasingly sit on generation that would otherwise be wasted, and it is also why the total consumption figure, quoted alone, describes the size of the industry rather than its effect on anybody's grid. The figure worth asking for instead is what fraction of a given operation's power comes from generation that had no other buyer.
The protocol adjusts how hard the problem is so that blocks continue to arrive at the target interval regardless of how much computing power is competing. If everybody doubles their capacity, difficulty doubles, and everybody's revenue per machine halves. The network produces the same number of coins whatever anybody does.
That mechanism is what makes mining structurally hard as a business. Any improvement in efficiency is competitive only until competitors adopt it, at which point difficulty rises and the improvement has been transferred to nobody. The industry works continuously to become more efficient, and the aggregate effect of that work is a higher difficulty rather than higher profits.
What survives is relative position. A miner with cheaper power than the marginal competitor earns a margin equal to that difference, and the margin persists as long as the difference does. The business is therefore about being on the cheap end of a cost curve rather than about being efficient in absolute terms, which is a distinction that determines who survives each downturn.
Hash price is the revenue a unit of computing power earns per day, and it is the single figure that summarises mining economics. It rises with the asset price and with transaction fees, and falls when difficulty rises. Every operator's viability is a comparison between hash price and their own cost per unit of computation.
It is published, it is comparable over time, and it collapses three variables into one that actually predicts behaviour. When hash price falls below a given operator's cost, that operator stops, and stopping reduces difficulty at the next adjustment, which raises hash price for everybody who remained. The industry has a self-correcting mechanism and it works by pushing out the least efficient.
For anybody analysing the sector, this replaces most of the commentary. A statement about miners being under pressure means hash price is below the costs of some identifiable fraction of the network, and that is checkable rather than atmospheric.
Miners earn in the asset and pay their costs in national currency, which makes them structural, continuous sellers regardless of any view they hold on the price. That flow is one of the few genuinely predictable sources of supply in the market, and its size can be estimated: total network revenue is public, and the fraction that must be sold is roughly the fraction of costs to revenue.
The behaviour that varies is what they do with the rest. Operators with strong balance sheets can hold, and their holdings become a source of supply later, at a moment of their choosing. Operators under pressure sell everything and sometimes sell reserves as well, which is why miner selling increases at exactly the times the price is already falling.
Both patterns are observable on the public ledger, since mining addresses are identifiable and transfers from them can be tracked. It is a slow and noisy signal, it is one of very few supply-side measurements that exist, and it is far more grounded than most of what circulates as analysis.
Miners earn in the asset and pay their bills in currency. That makes them sellers every day, whatever they think about the price.
New machine generations arrive irregularly and change the economics abruptly. A substantially more efficient model makes the previous generation uncompetitive at a given power price, and the resulting purchases raise network computing power, which raises difficulty, which reduces revenue per machine for everybody including the buyers.
The timing of these purchases is therefore a collective action problem with a familiar shape. Buying early captures a margin before difficulty absorbs it; buying late means paying for machines whose advantage has already been competed away. Operators who time this badly have bought equipment at the top of a cycle and run it into a difficulty level that never justified the price.
There is a second-order effect worth noticing. Because the hardware is specialised and worth little for anything else, a downturn produces a large supply of used machines at low prices, which allows operators with cheap power to expand cheaply. That is how capacity migrates toward the lowest-cost operators after every cycle, and it is the mechanism by which the industry becomes progressively more concentrated in specific locations.
Several mining operations are listed, and their shares are frequently treated as a way to hold exposure to the asset. They are not, or rather they are that plus several other things, and the other things dominate more often than the buyers expect.
A mining share is exposure to the asset price, multiplied by the operator's leverage, minus their cost structure, minus dilution from raising capital, minus the depreciation of their hardware, plus or minus whatever they hold on their balance sheet. Several of those terms are large and none of them track the asset. A rising asset price with rising difficulty and a share issuance can produce a falling share against a rising asset, and has.
The honest description is that a mining share is an operating business in a commodity industry with a volatile input price, and it should be analysed as one. Cost per unit of computation, hedging policy, balance sheet holdings, and dilution history are the terms that matter, and none of them appear on a price chart of the asset.
Three things, all public and none requiring a subscription. Hash price against estimated marginal costs tells you whether a meaningful fraction of the network is operating at a loss, which is the condition under which capacity leaves and forced selling increases. Difficulty adjustments, which are dated and mechanical, tell you what happened to capacity in the preceding period.
And transfers from mining addresses, which are visible, tell you whether miners are holding or selling. None of these predict a price, and together they describe one of the few supply-side dynamics in this market that can be measured rather than inferred, which makes them worth more than their share of the commentary suggests.
The block subsidy plus the transaction fees in blocks they produce, with probability proportional to their share of total network computing power. Over enough blocks it is that share multiplied by total network revenue, which makes it a commodity business with unusually clean arithmetic and no control over price.
Because the protocol adjusts difficulty to keep blocks arriving at the target interval regardless of how much computing power competes. If everybody doubles capacity, difficulty doubles and revenue per machine halves. The network issues the same number of coins whatever anybody does.
The revenue a unit of computing power earns per day. It rises with the asset price and with transaction fees and falls when difficulty rises, which collapses three variables into the one number that determines whether any given operator is viable.
Because they earn in the asset and pay electricity, hosting and staff in national currency. That makes them structural daily sellers regardless of any view on the price, and the size of the flow can be estimated from public network revenue and typical cost ratios.
No. A mining share is asset exposure multiplied by operating leverage, minus the cost structure, minus dilution from raising capital, minus hardware depreciation, plus whatever sits on the balance sheet. Those terms are large and do not track the asset, and shares have fallen while the asset rose.
Because its economic life ends when its electricity cost per attempt exceeds what the network pays, not when it breaks. Newer models convert power into attempts more efficiently, difficulty rises to absorb the improvement, and the older machine falls below the line while functioning perfectly.
Capacity migrates toward the lowest-cost operators. The hardware is specialised and worth little for anything else, so a downturn creates a large supply of used machines at low prices, which operators with cheap power buy. That is the mechanism by which the industry concentrates in specific locations.
Hash price against estimated marginal costs, which indicates whether a meaningful fraction of the network is operating at a loss. Difficulty adjustments, which are dated and mechanical and describe what happened to capacity. And transfers from mining addresses, which show whether miners are holding or selling.