Bitcoin's network hashrate remains 17–24% below its late-2025 peak of approximately 1.3 zettahashes per second (ZH/s), marking the longest sustained period without a new all-time high in the metric's history. Twenty One Capital CEO Rapha Zagury, speaking at Bitcoin Asia 2026 in Hong Kong on Augus...
"We're living, I think, through the first bear market in hash rate that we've ever seen in Bitcoin history." — Rapha Zagury, CEO, Twenty One Capital (Bitcoin Asia 2026, August 28, 2026)
Bitcoin's network hashrate remains 17–24% below its late-2025 peak of approximately 1.3 zettahashes per second (ZH/s), marking the longest sustained period without a new all-time high in the metric's history. Twenty One Capital CEO Rapha Zagury, speaking at Bitcoin Asia 2026 in Hong Kong on August 28, described the condition as Bitcoin's first "hashrate bear market."
The cause is structural, not cyclical. Unlike the 2021 hashrate collapse triggered by China's mining ban — where machines simply relocated and eventually came back online — the current decline reflects a permanent reallocation of capital, power capacity, and physical infrastructure from Bitcoin mining toward artificial intelligence and high-performance computing (HPC). Publicly listed miners have signed over $70 billion in aggregate AI and HPC contracts, and CoinShares projects that AI-derived revenue could constitute 70% of listed miners' total revenue by year-end 2026, up from approximately 30% at the start of the year.
With Bitcoin trading near $77,100 and the weighted average cash cost to produce one bitcoin among public miners at approximately $80,000 as of Q4 2025, mining economics are negative for a significant portion of the industry. The result is a sector-wide identity crisis: companies that built their businesses on SHA-256 computation are becoming data center landlords.
Bitcoin's network hashrate peaked near 1.3 ZH/s in late 2025. By Q2 2026, realized hashrate among publicly listed miners had fallen from 368.3 exahashes per second (EH/s) in Q4 2025 to 319 EH/s — a 13.4% decline, according to data compiled by Miner Weekly. The broader network-wide hashrate dropped to approximately 1,004 EH/s in Q2 2026, down 5.8% quarter-on-quarter, per CoinShares' Q1 2026 mining report.
The decline is not uniform. A weather-related event in late January 2026 — Winter Storm Fern — caused a temporary plunge from 1.1 ZH/s to 663 EH/s over a single weekend as Texas-based miners curtailed operations to relieve grid pressure. That event recovered within days. The structural decline since has been slower and steadier, driven by deliberate capital reallocation rather than forced shutdown.
CoinShares projects the hashrate will recover to 1.8 ZH/s by year-end 2026 and reach 2 ZH/s by end-Q1 2027, but the path depends on Bitcoin's price trajectory and the relative economics of mining versus AI hosting.
The most concrete evidence of the pivot comes from Q2 2026 earnings reports filed with the SEC:
TeraWulf (WULF): Generated $44.8 million in Q2 2026 revenue, of which $31.9 million (71%) came from HPC lease revenue. Bitcoin mining contributed $12.8 million. This follows a Q1 in which TeraWulf reported $21 million in AI/HPC hosting revenue, the first quarter in which AI surpassed mining as the company's largest revenue source. Bernstein reaffirmed an Outperform rating, citing the HPC revenue trajectory. The company operated 81 MW of revenue-generating critical IT capacity at its Lake Mariner facility as of June 30, expanding to 102 MW in early July, with an additional 336 MW under construction. TeraWulf ended the quarter with approximately $3.0 billion in cash and restricted cash.
Riot Platforms (RIOT): Reported $23.2 million in data center revenue alongside $113.7 million from mining in Q2 2026. Data center revenue represented approximately 17% of total revenue, up from near-zero a year earlier.
Cipher Digital (formerly Cipher Mining): Operating hashrate halved from 23.6 EH/s in Q4 2025 to 11.6 EH/s, with Q1 2026 revenue declining 41.7% sequentially to $34.8 million. Despite the revenue contraction, contracted HPC revenue reached $11.4 billion in total value, positioning the company for stable, long-term cashflows from investment-grade counterparties.
MARA Holdings: Q2 2026 revenue reached $174.9 million, a 27% decline from $238.5 million in the year-ago period. Net loss expanded to $611.3 million, compared with $808.2 million in net income a year earlier, driven partly by a $343 million fair-value impairment on digital asset holdings.
CleanSpark (CLSK): Reported $138.0 million in fiscal Q3 revenue, down 30.5% year-over-year. Net loss was $239.8 million ($0.89/share) versus net income of $257.4 million ($0.90/share) in the prior-year period. Adjusted EBITDA swung to a negative $113.0 million from positive $377.7 million.
Bitdeer (BTDR): Total revenue grew 47% year-over-year to $228.8 million in Q2 2026. AI Cloud revenue reached $14.0 million, up from $1.3 million a year earlier — a tenfold increase, though still a small share of total revenue.
HIVE Digital: An outlier that expanded Bitcoin hashrate from 6.5 EH/s to 25.1 EH/s, driving fiscal 2026 revenue up 158% to $297.8 million. HIVE announced plans for a 320 MW AI "Gigafactory" in the Greater Toronto Area, targeting $660 million in annualized recurring AI revenue by year-end 2028.
Nine comparable public miners collectively spent $5.11 billion on capital assets during H1 2026 while generating $341.2 million in directly reported AI/HPC revenue — a 15-to-1 capex-to-revenue ratio, according to Cointelegraph analysis. However, Q2 AI/HPC revenue reached $205.8 million across those nine miners, a 52% quarter-on-quarter increase, suggesting the revenue ramp is accelerating.
The aggregate contract value signed by publicly listed miners now exceeds $70 billion, per industry estimates. The largest single commitments:
The economics favor the pivot. According to CoinShares, AI hosting contracts can deliver 80–90% operating margins, compared with volatile and often negative margins in Bitcoin mining at current difficulty and price levels.
The financial pressure driving the pivot is quantifiable:
The post-halving block reward of 3.125 BTC, combined with elevated network difficulty, has compressed margins to the point where only operators with sub-$0.08/kWh power and latest-generation hardware (under 15 J/TH) maintain positive unit economics. Companies running S19-series machines (21–30 J/TH) are operating at a loss absent power subsidies or supplementary revenue streams.
The hashrate bear market raises legitimate questions about Bitcoin's security model. Publicly listed U.S. miners have accounted for over 40% of global hashrate. A sustained reduction in their contribution could, paradoxically, improve geographic decentralization of the network — or it could concentrate remaining hashpower among opaque, unregulated operators.
The January 2026 Winter Storm Fern event — which briefly reduced hashrate by 30–40% — exposed the geographic concentration risk inherent in U.S.-heavy mining. When Texas facilities curtailed simultaneously, the network experienced a visible drop in computational security, though Bitcoin's difficulty adjustment mechanism responded as designed.
However, absolute hashrate levels remain orders of magnitude above what would be required for a practical 51% attack. Even at the Q2 2026 trough of approximately 1,000 EH/s, the network's computational security exceeds levels from as recently as early 2024 by a wide margin. The security concern is more theoretical than immediate.
The industry is stratifying into three camps:
Full pivot: TeraWulf, Cipher Digital, and Core Scientific have committed the majority of their future capacity to AI/HPC. Bitcoin mining is becoming a legacy revenue line for these companies. TeraWulf's 71% HPC revenue share in Q2 2026 makes it, functionally, a data center company that happens to still mine bitcoin.
Dual-track: Hut 8, Bitdeer, and HIVE Digital are expanding both mining and AI simultaneously. HIVE's 285% hashrate expansion coexists with plans for a 320 MW AI facility. These companies are hedging — maintaining mining exposure while building AI optionality.
Mining-first: MARA, CleanSpark, and Riot Platforms continue to operate large-scale Bitcoin mining as their primary business, though all three have begun exploring or deploying AI/HPC capacity. Riot's $23.2 million Q2 data center revenue signals early diversification. These companies are betting that Bitcoin's price trajectory will eventually restore mining margins.
The divergence creates an unusual dynamic in public equities: investors now evaluate former mining companies using data center REIT metrics (contract value, MW capacity, lease terms) rather than traditional mining metrics (hashrate, BTC held, cost per coin).
The Bitcoin mining industry is undergoing its most significant structural transformation since the ASIC revolution of 2013–2014. The catalyst is not technological — the SHA-256 computation underlying Bitcoin mining is unchanged. The catalyst is economic: artificial intelligence has created a competing bid for the same scarce resources (power, land, cooling infrastructure) that mining requires, and AI's bid is higher.
The $70 billion in signed contracts, the 71% HPC revenue share at TeraWulf, and the 13.4% decline in public miner hashrate are not temporary dislocations. They reflect a permanent repricing of how computing infrastructure is valued and deployed. Companies that once defined themselves by exahashes now define themselves by megawatts.
For Bitcoin's network, the implications are ambiguous. Lower hashrate from public miners does not necessarily mean lower total hashrate — private and international miners may fill the gap. CoinShares' projection of 1.8 ZH/s by year-end 2026 implies a recovery, contingent on Bitcoin's price. But the structural pull of AI revenue ensures that any recovery will be slower and more contested than previous cycles.
The mining industry is not dying. It is splitting in two: a shrinking cohort of pure miners betting on Bitcoin's price, and a growing cohort of infrastructure companies that mine bitcoin as a byproduct of operating power-intensive facilities. The latter group is winning the capital markets argument. Whether they are winning the right argument depends entirely on where Bitcoin trades in 2027.