Miners lose over $19k for each BTC mined: Supply dynamics reshaped by cost inversion

Measuring miners’ true costs is not easy. It involves at least three layers: cash costs (power, operations and maintenance, labor), depreciation costs (amortization of mining equipment), and all-in costs (including financing expenses and capital expenditures). Under different accounting frameworks, the gap can reach tens of thousands of dollars.

According to CoinShares’ Q1 2026 mining report, the weighted average cash cost to produce one Bitcoin for publicly traded mining companies rose to approximately $79,995 in Q4 2025. If depreciation, management expenses, and capital expenditures are included, some estimates push all-in costs into the $83,000 to $90,000 range.

As of April 8, 2026, according to Gate market data, the Bitcoin spot price is trading around $71,000, implying that the theoretical cash loss per Bitcoin mined by mining companies is between $12,000 and $19k. About 15% to 20% of miners’ machines are already operating at a loss.

Behind the surge in costs is the combined effect of multiple forces: the Bitcoin halving in April 2024 cut block rewards from 6.25 BTC to 3.125 BTC, slashing miners’ core revenue by half; global energy prices have continued to rise; the iteration of next-generation ASIC miners has pushed up capital expenditures; and network difficulty hit a record high of 156 T by the end of 2025. Together, these factors have transformed mining from a “money-printing machine” into a “gold-devouring beast.”

How big is the selloff by mining companies?

Facing cash losses of nearly $20k per coin, mining companies’ balance sheets are under unprecedented pressure. In Q1 2026, publicly traded miners staged a collective wave of selling.

Riot Platforms sold 3,778 Bitcoins in the quarter, at an average sale price of $76,626 per coin, generating about $289.5 million in proceeds. This figure implies its selling volume was about 2.6x its quarterly production (1,473 BTC). Its Bitcoin holdings fell from about 19,223 BTC to 15,680 BTC, down about 18% year over year.

MARA Holdings’ selloff was even larger. Between March 4 and March 25, the company sold 15,133 Bitcoins in phases, raising about $1.1 billion in cash, primarily to repurchase convertible notes. This reduction far exceeds the scope of “newly mined production in the month of sale.” In 2025 as a whole, MARA mined only 8,799 BTC, while the amount sold in March alone was already close to 1.7x its full-year production.

The selloff wave did not stop with the top miners. Bitfarms announced it would liquidate all of its Bitcoin reserves and redeploy the funds into building AI data centers. Core Scientific explicitly said it plans to sell all of its roughly 2,500 BTC holdings in Q1 2026. Several publicly traded mining companies together sold more than 19k BTC in Q1 2026.

The selling pressure from miners has also spilled over to the network layer. Bitcoin hashprice fell from about $63 per PH/s per day in July 2025 to about $28 to $30 per PH/s per day in early March 2026, reaching the lowest level since the halving.

What is happening to post-halving supply economics?

Behind the selloff wave is a structural reshaping of the post-halving supply landscape.

The halving in April 2024 reduced block rewards from 6.25 BTC to 3.125 BTC, and cut newly produced Bitcoins per day from about 900 to about 450. Under the traditional logic of the halving cycle, supply contraction should boost prices. But reality from 2025 to 2026 has diverged from this narrative: the Bitcoin price fell more than 45% from its roughly $126k peak in October 2025 to the $65k range.

The core reason is that the supply shock from the halving has been largely diluted by ETF flows and institutional capital inflows. In past halvings, the daily marginal selling pressure dropped by hundreds of BTC, but now the size of institutional holdings is far greater than miners’ daily output. Miners’ selling behavior no longer dominates price, though its structural impact is deeper.

From the computing power side, after Bitcoin network hash rate reached a peak of about 1,160 EH/s at the end of 2025, Q1 2026 saw the first quarterly decline in six years. Year-to-date it is down about 4% to 10%, and has recently stabilized between 900 and 1,020 EH/s. At the same time, Bitcoin mining difficulty saw a significant 7.76% decrease in late March, the largest drop in more than a year.

These indicators point to a signal: inefficient miners are being systematically squeezed out of the network. As institutions such as CICC and GF Futures have noted, the steepening of the miners’ cost curve means the industry is undergoing an “capacity clearing,” and the exit of weaker participants will pave the way for optimization of the subsequent supply structure. The supply-contraction effect of the halving has not disappeared—it has been masked by the price-down cycle. But once prices rebound, the hash rate that has exited will be difficult to restore quickly, and supply elasticity may be lower than the historical average.

What three response paths are miners facing?

With nearly $20k of cash loss per coin, miners’ options are extremely limited. From industry developments in Q1 2026, the paths mainly cluster into three directions.

Path one: sell reserves to obtain cash. This is currently the most common choice. Riot sold 3,778 BTC for about $289.5 million, mainly to cover costs for building and operating the Corsicana AI data center. MARA sold 15,133 BTC for about $1.1 billion, reducing its outstanding debt from $3.3 billion to $2.3 billion, a drop of roughly 30%, and saving about $88.1 million in interest expenses in one go. The cost of this path is giving up potential gains from a future Bitcoin uptrend, but it buys survival space in the present.

Path two: transition to AI and high-performance computing (HPC). Several publicly traded mining companies have announced cumulative AI and HPC contracts exceeding $70 billion. Hut 8 has signed a 15-year, $7 billion leasing agreement with Google. Riot is repurposing two-thirds of its Corsicana mining site facilities for AI workloads. AI hosting provides predictable, higher-margin revenue—typically an operating margin of 80% to 90%—through multi-year fixed-dollar contracts, sharply contrasting with the highly volatile Bitcoin block rewards. For some leading companies, AI-related revenue currently accounts for about 30% of total revenue and is expected to rise to as high as 70% by the end of 2026.

Path three: shut down inefficient mining rigs. This is the most direct but also the most destructive choice. Approximately 252 EH/s of marginal hash rate is currently offline, mainly older hardware with efficiency below the 25 J/TH standard. This path reduces hash-rate competition, improving profitability prospects for surviving miners, but it also means those miners completely exit the market.

The three paths are not mutually exclusive. Riot and MARA take a “sell + transition” combination strategy; Core Scientific and Bitfarms are more inclined toward “liquidate everything + fully transition.” Overall, AI transition is currently the most strategically valuable direction—it turns miners from passive acceptors of price volatility into proactive service providers for hash-rate infrastructure.

Why hodling is systematically breaking down in 2026

Bitcoin miners’ “hodling” behavior—mining and holding for the long term rather than selling immediately—has historically held a near-faith-like status in the industry. Its core logic is: miners’ production costs are denominated in fiat, but their output is denominated in Bitcoin. If you believe Bitcoin will rise long term, delaying sales can lock in higher fiat gains. This strategy was validated at scale during the 2020 to 2021 bull-market cycle.

But in 2026, this belief is being systematically dismantled. From the data perspective, publicly traded Bitcoin mining companies have already sold more than 15,000 BTC between October 2025 and February 2026, making this the most intense “treasury liquidation” wave in the current cycle. At the beginning of 2026, MARA revised its Bitcoin inventory reserve policy, expanding from allowing sales only of newly mined output to allowing discretionary sales of all reserves held on its balance sheet. Core Scientific has clearly stated that it plans to monetize the vast majority of its Bitcoin holdings.

The breakdown of hodling has deep underlying logic. First, miners’ marginal profit margins have turned from positive to negative—continuing to hold newly mined BTC means absorbing roughly $19k of cash loss per coin, which is difficult for any balance sheet to sustain long term. Second, miners’ financing structures have changed: the pressure from convertible note maturities and the capital expenditures required for AI transition force management to treat Bitcoin as a liquid asset that can be deployed, rather than a long-term reserve. Furthermore, the emergence of Bitcoin ETFs has changed the market structure—institutional investors can gain Bitcoin exposure through ETFs, so Bitcoin held by mining companies no longer commands a scarcity premium.

It’s worth noting that miners’ selling is not a full, industry-wide retreat. The market shows structural differentiation: Strategy bought 44,377 BTC in March alone; Japan-listed Metaplanet increased its holdings by 5,075 BTC in Q1. Bitcoin demand has not disappeared—it has concentrated among participants with stronger financial capacity.

Can AI transition become miners’ long-term way out?

Transitioning to AI is the most significant structural change in the 2026 Bitcoin mining industry. But is this trend truly a way out, or just a temporary safe harbor?

From the revenue structure perspective, the logic of AI transition holds. Bitcoin mining revenue is highly dependent on coin prices and network transaction fees, which are extremely volatile. In contrast, AI hosting contracts are priced in fixed dollars over multi-year terms, providing predictable cash flows. CoinShares estimates that the operating profit margin for AI hosting can reach 80% to 90%, far above the current marginal profit of Bitcoin mining. By the end of 2026, for some leading miners, the share of revenue from AI could be as high as 70%, effectively transforming them into data center operators.

From capital allocation, AI transition is already accelerating. Bitdeer disclosed data from its GPU cloud business: it has deployed 2,096 GPUs with a utilization rate of 64%, generating about $21 million in annual recurring revenue. Core Scientific obtained a $500 million credit facility from Morgan Stanley to fund expansion into the HPC hosting space.

But AI transition is not without boundaries. The transition costs are extremely high—turning a Bitcoin mining site into an AI mining site can cost as much as $1 million to $1.5 million per megawatt. This helps explain why only well-capitalized players can carry out the transition. In addition, AI data centers have requirements for power stability, network latency, and cooling systems that far exceed those of Bitcoin mining, meaning not all mining sites meet the conditions for conversion.

The deeper question is: if a large number of miners move into AI infrastructure, will the AI compute market face oversupply? Currently, even for mega-scale cloud providers such as Google, demand for compute is still expanding rapidly. Miners transitioning to AI looks more like filling a shortage of a market that is not meeting demand. As long as this gap exists, AI transition can provide miners with a sustainable way out.

Summary

The Bitcoin mining industry is experiencing its most severe profitability crisis since the 2024 halving. Publicly traded miners’ weighted average cash cost is about $79,995, while the BTC spot price has been trading in the $68,000 to $70,000 range. The theoretical loss per Bitcoin mined is about $19k. In Q1 2026, Riot sold 3,778 BTC (about 2.6x quarterly production) and MARA sold 15,133 BTC (about $1.1 billion), and multiple publicly traded mining companies together sold more than 19,000 BTC. Bitcoin hash rate saw its first quarterly decline in six years, and approximately 252 EH/s of marginal hash rate is offline.

Miners face three response paths: selling reserves for cash, transitioning to AI and HPC, and shutting down inefficient mining rigs. Among these, AI transition is currently the most strategically valuable direction. Publicly traded mining companies have announced cumulative AI and HPC contracts exceeding $70 billion, and it is expected that by the end of 2026, for some companies the share of revenue from AI could reach 70%. Hodling is being systematically dismantled, and the market is showing structural divergence—miners sell due to operating pressure, while institutions such as Strategy are still continuously increasing their positions. The post-halving supply contraction effect has not disappeared, but it has been masked by the price-down cycle. Once prices rebound, the hash rate that has exited will be difficult to restore quickly, and supply elasticity may be below the historical average.

FAQ

Q: Why has Bitcoin mining cost exceeded $80k?

Mining costs are mainly made up of power expenses, mining rig depreciation, operating and maintenance expenses, and financing costs. After the April 2024 Bitcoin halving, block rewards fell from 6.25 BTC to 3.125 BTC, and miners’ unit earnings were directly cut in half. At the same time, global energy prices rose, the iteration of next-generation ASIC miners increased capital expenditures, and network difficulty reached a historical high of 156 T by the end of 2025, together pushing up all-in costs. According to CoinShares’ Q1 2026 mining report, the weighted average cash cost for publicly traded miners has risen to about $79,995.

Q: What impact does miners’ selling of Bitcoin have on market supply?

Miners are one of the most stable sources of supply in the Bitcoin market. In Q1 2026, Riot sold 3,778 BTC and MARA sold 15,133 BTC, and multiple publicly traded mining companies together sold more than 19,000 BTC. These selloffs increase short-term supply pressure on the market, but they also represent a limited upper bound on future sell pressure, because the remaining miners are mostly participants with low-cost energy and advanced ASIC infrastructure. From the hash rate perspective, once inefficient miners are squeezed out of the network, the industry’s supply structure will become more optimized.

Q: Does hash-rate decline mean Bitcoin network security is threatened?

Bitcoin’s network has a difficulty adjustment mechanism that can automatically adapt to changes in hash rate. The current hash-rate decline is mainly driven by inefficient miners exiting due to profitability pressure, not network security problems. Global hash rate is still stable in the $900 to $1,020 EH/s range, far above historical levels. Network difficulty was reduced by about 7.76% in late March, improving profitability for surviving miners and demonstrating the self-regulating ability of the Bitcoin protocol.

Q: Is AI transition a feasible path for all miners?

AI transition is not something every miner can participate in. Transition costs are high: converting a Bitcoin mining site into an AI data center can cost as much as $1 million to $1.5 million per megawatt, and it also has higher requirements for power stability, network latency, and cooling systems. Only well-capitalized publicly traded mining companies have the ability to transition. For small and mid-sized miners, shutting down or selling mining rigs remains a realistic option.