Bitcoin's network hashrate sits at approximately 934 EH/s as of September 22, 2026 — roughly 50% below its expected growth trend, according to CoinShares. The gap represents the longest sustained deviation from trend since China's 2021 mining ban. The cause is not regulatory prohibition. It is ec...
"A new constraint has emerged in the US...it is becoming increasingly difficult to build new data centres." — Luke Nolan, Senior Research Associate, CoinShares
Bitcoin's network hashrate sits at approximately 934 EH/s as of September 22, 2026 — roughly 50% below its expected growth trend, according to CoinShares. The gap represents the longest sustained deviation from trend since China's 2021 mining ban. The cause is not regulatory prohibition. It is economic: AI infrastructure generates approximately $1.5 million in annual profit per megawatt, versus $500,000 for Bitcoin mining. That 3-to-1 margin disparity has triggered a structural exodus from hash power.
CoinShares' Q2 2026 Bitcoin Mining Report, authored by Luke Nolan and published September 15, found that the weighted average cash cost to produce one bitcoin among listed miners reached $75,500 in Q2 — while BTC ended the quarter at $58,400. For the first time since the 2024 halving, the listed mining sector fell below aggregate cash breakeven. Monthly hashprice hit an all-time low of $27.7 per PH/s per day in June.
Bitcoin has since rallied to $86,000 as of September 21, pushing hashprice back to approximately $39.63/PH/s/day — a 43% recovery from the June trough. The relief may be temporary. At least four major publicly listed miners — Keel (formerly Bitfarms), IREN, Cipher Digital, and TeraWulf — have committed to exiting Bitcoin mining entirely, redirecting their power and infrastructure toward AI and high-performance computing. CoinShares calls this pivot "structurally irreversible."
Bitcoin's hashrate breached 1 zettahash per second (ZH/s) for the first time in January 2026, peaking at 1.1 ZH/s in October 2025 when BTC traded near its all-time high of $126,500. Since then, the trajectory has reversed. As of September 22, CoinWarz reports network hashrate at approximately 863 EH/s, with Luxor's Hashrate Index listing a 7-day average near 934 EH/s. The discrepancy reflects measurement methodology differences, but both figures sit well below the 1 ZH/s milestone.
Difficulty stands at 132.76 trillion as of block 968,086. The network has logged 18 distinct difficulty adjustments since January 2026, including a 15% single-epoch increase in February — the largest since China's 2021 mining ban — and a 19% decline later in the year as miners curtailed operations.
Hashprice, the metric that captures miner revenue per petahash per second per day, tells the profitability story in one number:
| Period | Hashprice ($/PH/s/day) | BTC Price | Context | |--------|----------------------|-----------|---------| | Oct 2025 | ~$62 | $126,500 | All-time BTC high | | Jan 2026 | ~$48 | $98,000 | Post-peak decline | | June 2026 | $27.70 | ~$55,000 | All-time hashprice low | | Sept 7, 2026 | $39.63 | $79,700 | Recovery | | Sept 22, 2026 | ~$44 (est.) | $86,000 | Current rally |
The June low of $27.7/PH/s/day was comparable to post-COVID crash levels in 2020. Transaction fees contributed just 0.43% of total miner revenue in the most recent 24-hour period, underscoring the sector's near-total dependence on the block subsidy.
CoinShares' Q2 2026 mining report documents the most punishing quarter for listed miners since the April 2024 halving cut the block reward from 6.25 to 3.125 BTC.
The weighted average ex-tax cash cost to produce one bitcoin reached $75,500 among the listed mining cohort. Bitcoin's average Q2 price was approximately $58,400. The math is unforgiving: listed miners, in aggregate, spent more to produce bitcoin than the asset was worth.
Some operators faced production costs exceeding $100,000 per BTC when Bitcoin traded below $60,000. The economics forced extreme responses. Core Scientific paid $41.9 million to cancel delivery of 15 EH/s worth of next-generation 3nm Proto mining chips — hardware it had already ordered — rather than deploy it for mining.
The all-time-low monthly hashprice of $27.7/PH/s/day in June reflected the combined weight of compressed BTC prices, elevated difficulty, and the halved block subsidy. For operators without sub-$0.04/kWh power and latest-generation ASICs rated below 15 J/TH, the quarter was existential.
The migration of Bitcoin mining infrastructure toward AI and high-performance computing (HPC) has accelerated from corporate narrative to contractual reality.
CoinShares estimates that AI data center operations generate approximately $1.5 million in annual profit per megawatt. Bitcoin mining generates approximately $500,000. The gap is structural, not cyclical. AI workloads come with long-duration contracts — leases stretching 15 to 20 years — from investment-grade counterparties. Bitcoin mining offers variable rewards denominated in a volatile asset with no contractual floor.
Market assessments put signed AI and HPC contracts by current and former Bitcoin miners above $70 billion as of early 2026. Analysts at Bernstein report that every major U.S.-listed Bitcoin miner has moved toward AI infrastructure in some form. Sector-wide data center capital expenditure increased 400% between March 2025 and February 2026.
The CoinShares Q2 report characterizes the pivot as "structurally irreversible." Companies that have signed 15-year data center leases and cancelled mining hardware orders cannot reverse course based on a Bitcoin price recovery. The leases are binding. The hardware is cancelled. The sites are being retrofitted.
TeraWulf signed a 20-year data center lease with Anthropic projected to generate upwards of $19 billion in revenue over the contract's life. That single contract dwarfs what any Bitcoin mining operation has generated cumulatively.
The exit list is specific and growing:
Exiting Bitcoin Mining:
At least 35 EH/s is scheduled to leave the publicly listed mining cohort as these exits complete. That represents approximately 3.7% of current network hashrate.
Remaining as Miners:
The surviving miners share common traits: sub-$0.04/kWh power costs, latest-generation ASIC fleets rated at or below 15 J/TH, and scale sufficient to negotiate favorable power purchase agreements.
The departure of listed miners from hash power production creates what Luxor Mining has termed a "hashrate vacuum." If 35 EH/s leaves the public cohort and is not replaced by private miners or new entrants, the network difficulty will adjust downward, improving economics for remaining participants.
Bitcoin's difficulty adjustment algorithm recalibrates every 2,016 blocks (approximately two weeks), ensuring that blocks are produced at a roughly 10-minute interval regardless of hashrate changes. The mechanism is self-correcting: as hashrate falls, difficulty drops, making mining more profitable for survivors, which eventually attracts new capacity.
The security question is whether total hashrate declines enough to reduce the cost of a 51% attack to economically plausible levels. At 863–934 EH/s, the network requires an attacker to control approximately 432–467 EH/s — a capital expenditure measured in tens of billions of dollars for hardware alone, plus ongoing electricity costs. The security margin remains wide by any practical measure, even in a reduced-hashrate environment.
The longer-term concern is geographic and operator concentration. If surviving miners are predominantly large U.S.-listed companies with similar power sources and regulatory exposure, the network's censorship resistance narrows. Approximately 37–38% of global hashrate originates in the United States, Russia follows at 17%, and China at 12%.
According to the Cambridge Centre for Alternative Finance's April 2025 survey (the most recent comprehensive dataset), 52.4% of Bitcoin mining electricity comes from zero-emission sources, up from 37.6% in 2022. The breakdown:
| Source | Share | |--------|-------| | Natural gas | 38.2% | | Hydropower | 23.4% | | Wind | 15.4% | | Nuclear | 9.8% | | Coal | 8.9% | | Solar | 3.2% | | Other | 1.1% |
Coal's share has dropped from 36.6% in 2022 to 8.9%. Natural gas is now the single largest individual fuel source. Within the natural gas segment, stranded gas operations supply 507 MW of mining capacity — 3.3% of total — and achieve up to 99.9% methane destruction efficiency, compared with 91–98% for open flaring.
With global hashrate near 1 ZH/s and average fleet efficiency of approximately 16 J/TH, the network consumes an estimated 350–420 GWh per day, or roughly 128–153 TWh annually. The departure of some miners to AI workloads does not necessarily reduce total energy consumption at those sites — it redirects the same megawatts toward different computational tasks.
Bitcoin's Q3 rally — up 44%, its best quarter since Q4 2024 — has dramatically improved the picture for miners who stayed. With BTC averaging well above the $75,500 average cash cost that characterized Q2, surviving operators are posting substantial margins:
| Company | Q3 Revenue | Net Income | BTC Mined | Avg. Cost/BTC | |---------|-----------|------------|-----------|---------------| | Marathon Digital | $252M | $123M | 2,144 | <$40,000 | | CleanSpark | $181.7M | N/A | 1,957 | ~$46,000 (est.) | | Riot Platforms | $180M | $104M | N/A | N/A |
Marathon's sub-$40,000 production cost against BTC at $86,000 implies a gross margin exceeding 53%. The disparity between Q2 losses and Q3 profits illustrates the violent cyclicality of Bitcoin mining economics. The miners who survived Q2's cash-breakeven crisis are now capturing the upside of a 44% quarterly BTC price increase with reduced competition.
The Bitcoin mining industry in September 2026 is undergoing a structural separation. On one side: a shrinking cohort of low-cost, scale operators posting 50%+ margins on a $86,000 bitcoin. On the other: former miners repurposing their power infrastructure for AI workloads that pay three times as much per megawatt with contractual certainty.
The network's self-correcting difficulty adjustment ensures that blocks will continue to be produced on schedule. The security margin remains wide. What changes is the industry's composition — fewer operators, higher concentration, and a permanently altered relationship between Bitcoin mining and the broader energy-compute market.
The hashrate vacuum left by departing miners will eventually be filled — by surviving operators expanding, by private miners in low-cost jurisdictions, or by new entrants attracted by improved unit economics after difficulty adjusts downward. The question is not whether Bitcoin mining continues. It is whether the mining industry, as a distinct business category, survives its collision with a better-paying customer for the same scarce resource: power.