In November, experts said the sector is in its most severe profit crisis yet. The cost of efficient production has climbed sharply owing to advanced cooling systems and next-generation ASICs.
In this environment, only miners with ample financial buffers can stay afloat—and, in effect, capture hashrate from weaker rivals. Even they are partially or fully diverting capacity to AI infrastructure.
The 2028 halving will make survival tougher still and heighten dependence on bitcoin’s price. If the community fails to revive the on-chain economy to lift fees’ share of miner revenues, the network risks morphing into a set of a few centralised conglomerates and governments with the stamina to endure a “marathon of losses”.
A perfect storm
For the first time in bitcoin’s history, miners face a systemic crisis. Experts at TheMinerMag flagged the unprecedented severity of current conditions in their weekly industry review.
Whereas the industry entered the third quarter with a hashprice of $55 per PH/s per day, November’s market slide pushed the figure down to $34.21 per PH/s per day.
More troubling still, even industry leaders with efficient fleets and competitive power tariffs are hovering near break-even. Analysts estimate their median “hashcost” in the third quarter at roughly $44 per PH/s per day.
What happens after the next reward cut? The next halving is expected around April 2028 and will reduce miners’ base income from 3.125 BTC to 1.5625 BTC per block.
By Satoshi Nakamoto’s design, transaction fees should balance miner incentives. Reality looks different. In November, the share of miner revenue from fees fell below 0.7%, versus historic spikes of 5–15%.
The halving block in April 2024 brought a record $2.4m in fees from speculation with the Runes protocol, but fees soon fell to multi‑month lows. That raises longer-term concerns about maintaining network security amid weak incentives for its “defenders”.
The paradox is a rising hashrate alongside falling profitability.
According to Hashrate Index, on October 16 the hashrate hit an all-time high of 1.16 EH/s. Mining difficulty also set a new peak, nearing 156 T.
As production costs rise, profits fall. Competition intensifies, and expectations shift to faith in bitcoin’s long-term success and a desire to grab market share at any cost.
Only a few will make it through to profits—those with free capital over the next two years, access to cheap power and the most efficient hardware.
According to BlockEden.xyz, large miners with electricity costs below $0.05/kWh sustain margins of 30–75%. Among them:
- MARA — producing 1 BTC costs roughly $39,235 with total production costs of $26,000–28,000;
- — Riot Platforms — electricity costs of $0.025–0.03 per kWh (Texas);
- — CleanSpark — production cost of about $35,000 per 1 BTC.
At today’s bitcoin price, the break-even electricity range is $0.05–0.07/kWh for the newest machines, making home mining (typically $0.12–0.15/kWh) uneconomic. Smaller operators running older Bitmain Antminer S19s are nearing unprofitability; the S21 dominates with a 20–40% efficiency edge.
In the view of MARA CEO Fred Thiel, the market will self-regulate as miners hit the profitability wall. By 2028, he expects operators will either generate their own power, become part of energy companies, or partner with them.
“The days of miners simply plugged into the grid are numbered”, — said Thiel.
He also argues that rising hashrate and difficulty reflect players turning on their own fleets.
“Equipment manufacturers are launching their own farms because customers are buying fewer machines. The global hashrate continues to grow, which means the profitability of other participants falls”, — explained the head of MARA.
Bitmain, Canaan and Bitdeer have put their unsold ASICs to work.
In October 2024, Chinese car firm Cango unexpectedly acquired 32 EH/s of hashrate. Bitmain likely transferred $256m worth of unsold Antminer S19 XP units to it via subsidiary Antalpha.
A side hustle in AI
In September alone, the profitability of mining the oldest cryptocurrency fell by more than 7%. Miners turned to diversification.
Many experts say the hashrate’s rise reflects managements’ desire to maximise cash flows to build GPU farms for AI infrastructure.
Over the year, total industry debt rose sixfold—from $2.1bn to $12.7bn. VanEck analysts link this to the need to invest in hardware while fulfilling artificial-intelligence orders.
In July 2024, Bernstein analysts noted that miners had secured large power allocations—about 6 GW at the time—which they expect to double by 2027.
Digital-asset miners have an edge in plugging into large power grids and can help partners secure capacity faster.
“Bitcoin data centres are ideal for upgrades thanks to powerful servers, cooling infrastructure and their operational capabilities”, — the analysts wrote.
Forecasts suggest that by end-2027, 20% of bitcoin mining capacity will be redirected to AI, while the five largest US miners will control 25% of global hashrate.
Reasons include:
- AI workloads generate 2–5 times more revenue per kWh than mining bitcoin;
- tech giants such as Microsoft and OpenAI need GPU capacity;
- mining centres already have two key resources required for AI infrastructure—cheap power and industrial-scale cooling.
Companies have shifted energy capacity to AI and high-performance computing. The wave of 2025 takeovers and investments set a structural shift in motion. One of the largest deals was the acquisition in July by AI hyperscaler CoreWeave of mining data-centre operator Core Scientific for $9bn.
In August, Google increased its stake in TeraWulf to 14% by expanding financial guarantees to $3.2bn. In September, the latter announced a $3bn raise to build data centres.
Earlier, IREN invested $193m in AI computing.
Bitfarms also announced a gradual wind-down of bitcoin mining by 2027 and a pivot to AI infrastructure.
In April 2025, industry leader Riot Platforms abandoned the long HODL strategy it had followed since January 2024. It sold all 463 BTC mined that month plus an additional 12 BTC from reserves.
The sale generated about $38.8m in net proceeds to support operations.
“These sales reduce the need to raise capital by issuing new shares, decreasing the degree of dilution”, — said Riot CEO Jason Les.
In subsequent months the miner continued to sell its production: 465 BTC in September and 400 BTC in October.
By the third quarter, Riot Platforms’ net profit reached $104.5m on record total revenue of $180.2m.
During the period it mined 1,406 BTC versus 1,104 BTC a year earlier. As of September 30, Riot held 19,287 BTC (~$2.2bn) in reserves and had $170m in cash.
Riot vice-president Josh Kane noted the company is satisfied with its bitcoin mining performance but is focused first on extracting maximum value from its power access.
“As our strategy has evolved, so has our approach to bitcoin mining. We no longer view it as an end goal, but rather as a means to an end — to maximise the value of our megawatts”, — said Kane.
Since September 2024, the company has increased deployed capacity from 20.1 EH/s to 33.1 EH/s. Of this, 126 MW at the Corsicana site has already been reserved to build the “core and shell” of a data centre for AI services.
Mining infrastructure allows operators to change course quickly. When mining turns uneconomic, they can switch to AI workloads within minutes.
Survival of the fittest
Advances in ASIC miners have markedly improved efficiency in recent years. Liquid cooling has become the new standard for viable entry.
The 2024–2025 generation is a technological leap—but also shows a slowdown consistent with Moore’s law.
Over five years the efficiency of top models has improved about 65%, from 31 J/TH in 2020 to 11–13.5 J/TH today. As chips approach 3–5 nm, efficiency gains have slipped to 20–30%.
Upcoming models promise fiercer competition. The S23 Hydro, touted for Q1 2026, targets an unprecedented 9.5–9.7 J/TH at 580 TH/s.
Bitdeer is developing 3–4 nm SEALMINER chips aiming for 5 J/TH by 2026, while Block, Jack Dorsey’s firm, is collaborating with Core Scientific to deploy open-source 3 nm devices.
Per BlockEden, equipment pricing mirrors economics. Hydro-cooling alone adds $500–1,000 per unit, while immersion systems require $2,000–5,000 up front. The spend pays back via 20–40% higher returns as overclocking raises hashrate by 25–50%.
Cutting-edge cooling has shifted from nice-to-have to strategic necessity. Air-cooled miners are very noisy—75–76 dB—and need heavy ventilation. Immersion cooling solves this: machines are submerged in non-conductive liquid, noise disappears and hashrate rises by about 40%.
Leading hydro models such as the S21 XP Hyd and MicroBT M63S+ deliver water at 70–80°C, enabling heat reuse for agriculture, district heating or industrial processes.
This new technological ceiling has set a high barrier to entry and leaves little scope to improve decentralisation via small players.
Few options
The next two to three years could bring a full transformation of mining.
A likely bear phase would not only open opportunities to buy cheap “digital gold” during panics, but also free up hashrate for new large entrants.
If prices rise for an extended period, reallocation will continue—just less brutally. Despite higher revenues, keeping up with cheap power and “green” infrastructure will not be easy.
If large miners continue shifting capacity toward AI, the network risks:
- a slower hashrate growth;
- lower difficulty;
- pressure on decentralisation;
- greater vulnerability to hostile mining syndicates or state-linked entities.
Research into incentives for securing bitcoin offers various ways out:
from using “selfish” mining with game theory to tying rewards to the current hashrate.
In practice, however, only a few avenues exist to raise fees’ share of miner income:
- strengthening Bitcoin L2 with features such as programmable smart contracts and opcode activations for BTCFi;
- global roll-out of micropayments such as the Lightning Network;
- mass adoption of bitcoin as a means of payment.