Bitcoin's seven-day average hashrate recrossed 1 zettahash per second (ZH/s) on October 22 for the first time since March, even as the public mining sector accelerates a structural exit from the business that built it. Difficulty rose 5.92% on October 16 to 140.58 trillion — the third-largest hik...
"We are no longer a Bitcoin company." — Ben Gagnon, CEO of Keel Infrastructure (formerly Bitfarms), May 2026
Bitcoin's seven-day average hashrate recrossed 1 zettahash per second (ZH/s) on October 22 for the first time since March, even as the public mining sector accelerates a structural exit from the business that built it. Difficulty rose 5.92% on October 16 to 140.58 trillion — the third-largest hike of 2026 — and hashprice climbed back above $40/PH/s/day, giving miners their strongest revenue month since January at an estimated $1.12 billion.
The paradox is that the companies most associated with Bitcoin mining are no longer principally mining Bitcoin. Nine publicly listed miners have announced AI and high-performance computing pivots since late 2025. Disclosed contract values across these firms now exceed $70 billion. CoinShares projects that listed miners could derive 70% of revenue from AI by December 2026, up from roughly 30% at the start of the year. BlocksBridge Consulting data shows public miners' realized hashrate fell 15% in the first half of 2026, with the gap being filled by private operators and overseas pools running newer-generation ASICs.
The net effect is a bifurcation of the mining industry into two distinct businesses occupying the same physical infrastructure: a shrinking but still-profitable Bitcoin mining operation subsidized by rock-bottom power contracts, and a rapidly expanding AI colocation business commanding multibillion-dollar, decade-plus leases.
Bitcoin's 14-day average hashrate reached approximately 1,060 EH/s by October 27, according to CoinWarz, recovering from a trough of 898 EH/s in early August. The milestone is notable: 1 ZH/s (1,000 EH/s) had not been sustained since March 3, when the seven-day average last held above the threshold.
The recovery followed a turbulent summer. Mining difficulty fell 14% from its 2026 high by August 1, according to CoinDesk, as operators shut down older hardware or converted facilities to AI workloads. By mid-October, difficulty had rebounded to 140.58 trillion after a 5.92% upward adjustment at block 971,712 — the third-largest single increase this year. That figure remains roughly 10% below the all-time record of 155.97 trillion set in November 2025.
October revenue provided relief: miners earned an estimated $1.12 billion, the highest monthly figure since January, with spot hashprice rising above $40/PH/s/day according to Hashrate Index. The rebound reflects both a recovery in Bitcoin's price and the exit of marginal operators whose departure reduced difficulty earlier in the year.
The shift from mining to AI infrastructure is no longer anecdotal. According to BlocksBridge Consulting, disclosed AI and high-performance computing contracts across publicly listed miners now exceed $70 billion in total value. CoinShares, in its Q3 2026 mining report published October 28, projects that listed miners could derive roughly 70% of revenue from AI by year-end 2026.
The thesis is straightforward: Bitcoin miners own large-scale power infrastructure, often in regions with favorable energy costs, already connected to electrical grids. AI compute demand has outstripped available U.S. data center capacity, and miners happen to be sitting on the one resource hyperscalers need most — permitted megawatts.
Nine public miners have formally announced pivots: Bitfarms (rebranded as Keel Infrastructure), Core Scientific, Riot Platforms, IREN, TeraWulf, CleanSpark, HIVE Digital, Cipher Mining, and Cango. The scale of individual contracts is striking:
CoinShares data from late 2025 and early 2026 shows the AI revenue mix climbing rapidly at the company level: Core Scientific's AI/HPC data centers accounted for 39% of Q4 2025 revenue; TeraWulf's HPC business reached 27%; IREN's AI cloud business stood at 9%; HIVE's HPC operations at 5%.
Core Scientific (CORZ) provides the clearest case study. The company reported Q2 2026 revenue of $164.2 million, more than double the $78.6 million recorded a year earlier. Colocation revenue reached $136.7 million in the quarter — up from $10.6 million a year prior — and now constitutes approximately 83% of total revenue. The stock rose 75.8% in the first half of 2026, though it has since pulled back to around $16.45 amid broader AI-bubble concerns. The average analyst price target from 23 firms stands at $30.86.
Riot Platforms has taken a more measured approach, maintaining Bitcoin mining operations while layering AI contracts on top. Shares traded near $20.88 in late August, up approximately 58% year-to-date at that point. The AMD lease alone carries initial revenue of $311 million.
Marathon Digital chose the most aggressive treasury liquidation path, selling 15,133 BTC for roughly $1.1 billion in March 2026 to fund its infrastructure conversion.
Collectively, Bitcoin miner stocks surged approximately 90% in 2026 even as Bitcoin's own price fell 44% from its cycle high, according to BlockWind data. The market is pricing these companies as AI infrastructure plays, not Bitcoin proxies.
Public miners have sold approximately 28,000 BTC worth $1.78 billion since January 2026, reducing combined holdings from 127,000 BTC to roughly 99,000 BTC, according to CoinDesk data.
Marathon's 15,133 BTC sale was the largest single drawdown. Other notable liquidations include Core Scientific's approximately $175 million BTC sale to accelerate AI campus construction and broader quarterly sell-offs across Riot, CleanSpark, and Cipher Mining to cover power costs and facility conversion expenses.
The selling pressure is material but has been offset by a different category of buyer. Corporate treasury companies — led by Strategy (formerly MicroStrategy) — acquired approximately 220,000 BTC in 2026, absorbing 165% of all Bitcoin mined during the period. The net effect: miners produce and sell, treasuries accumulate, and network security is maintained by a shifting cast of operators.
If public miners' realized hashrate fell 15% in H1 2026 while network hashrate recovered to 1 ZH/s, someone else is plugging the gap. Several factors explain the substitution:
Private miners and overseas pools. Operators in Central Asia, the Middle East, and parts of Africa with subsidized electricity costs below $0.03/kWh have expanded capacity, particularly using newer sub-15 J/TH ASICs. These entities are not publicly traded and do not file earnings, making their aggregate capacity difficult to track precisely.
Next-generation hardware. ASIC efficiency improvements mean that fewer machines can produce the same hashrate at lower power draw. The most profitable ASICs in 2026 operate below 20 J/TH, according to mining profitability trackers. Miners upgrading from 30+ J/TH units to sub-20 J/TH machines effectively increase network hashrate per watt consumed.
Sticky power contracts. ViaBTC CEO Haipo Yang noted that AI demand is pushing mining out of the most expensive data center locations, but miners with long-term fixed-price power purchase agreements at $0.03–$0.05/kWh can still operate profitably at current hashprices. These operators have no incentive to pivot — their economics work.
The result is a less publicly visible but more geographically distributed mining network. Whether this improves or worsens Bitcoin's decentralization properties depends on the concentration of that private hashrate, which remains an open question.
The April 2024 halving cut block rewards from 6.25 to 3.125 BTC, compressing miner revenue per terahash from approximately $0.12 in early 2024 to roughly $0.06 by mid-2026. Revenue per petahash per day fluctuated between $36 and $55 throughout 2026, according to CoinShares' Q1 mining report.
CoinShares also found that hash price hit a five-year low during Q2, with approximately 20% of legacy miners operating at a loss. Profitability in October 2026 requires sub-15 J/TH hardware at power costs below $0.08/kWh. Less efficient equipment requires power below $0.04/kWh to clear a margin.
The industry's cost structure has polarized. CoinShares estimates that the most efficient operators — those running next-generation ASICs at sites with fixed-price power below $0.04/kWh — earn margins above 50%. Operators at the other end of the cost curve lose roughly $19,000 per coin produced, according to analysis published by BlockWind.
This polarization explains the AI pivot. For miners whose power costs sit between $0.05 and $0.10/kWh, converting to AI colocation generates revenue per megawatt that exceeds Bitcoin mining revenue by a factor of 5–10x at current hashprices. The economic logic is difficult to argue with.
The Bitcoin mining industry is undergoing a structural transformation that decouples the identity of its largest corporate participants from their original business. Public miners are becoming AI infrastructure companies that happen to mine Bitcoin on the side — or, increasingly, not at all. The hashrate itself has proven resilient, recovering to 1 ZH/s through a combination of private operators, next-generation hardware, and miners in low-cost jurisdictions who face no economic pressure to pivot.
For Bitcoin's security model, the immediate question is not whether hashrate holds up — it has — but whether the geographic and organizational distribution of that hashrate improves or deteriorates as public miners exit. The longer-term question is whether the AI revenue stream will ultimately bring more capital and infrastructure capacity into energy procurement than Bitcoin mining alone ever could, making the next generation of mining facilities larger and more efficient than anything the industry has built to date. The data so far suggests the answer to that second question is yes, though the timeline remains uncertain.