Bitcoin's network hashrate has declined approximately 12% from its late-2025 peak above one zettahash per second to roughly 868 exahashes per second as of late July 2026. Mining difficulty has fallen 19.9% from its all-time high — the third-deepest drawdown since ASICs became standard hardware. T...
"Execution, not signing, becomes the next premium." — Matthew Sigel, Head of Digital Asset Research, VanEck
Bitcoin's network hashrate has declined approximately 12% from its late-2025 peak above one zettahash per second to roughly 868 exahashes per second as of late July 2026. Mining difficulty has fallen 19.9% from its all-time high — the third-deepest drawdown since ASICs became standard hardware. The cause is not a government ban or a market panic. It is a rational economic reallocation: publicly traded miners have signed more than $70 billion in AI and high-performance computing contracts, and the most aggressive operators now derive 27–39% of revenue from non-mining activities.
The weighted average cash cost to produce one bitcoin climbed to nearly $80,000 as of CoinShares' Q4 2025 report, while bitcoin traded near $63,000 in mid-August 2026 — a gap that has forced 15–20% of the global mining fleet to operate at a loss. Hash price collapsed to roughly $32 per petahash per second per day, below breakeven for many facilities. Miners responded by selling a record 32,000 BTC in Q1 2026 alone, exceeding their combined sales for all of 2025.
This is not a temporary curtailment. Multi-year enterprise AI leases permanently lock computing power and electrical capacity away from mining. The industry that once defined itself by converting electricity into bitcoin is converting itself into something else entirely.
The Bitcoin network reached a milestone of 1 zettahash per second in late 2025. By late July 2026, the 7-day moving average had retreated to 868 EH/s. Bitcoin Magazine Pro tracked 287 consecutive days of downward trend through early August.
Mining difficulty peaked at approximately 156 trillion in November 2025 and stood at 126.23 trillion as of July 25, 2026 — a 19.9% decline. According to CoinMarketCap, only two prior episodes produced deeper drawdowns: China's 2021 mining ban and the 2018 bear market contraction.
Hash price — the daily revenue a miner earns per petahash per second of deployed capacity — fell to roughly $32/PH/s/day by late July. CoinShares estimated in March 2026 that this figure sat below the breakeven threshold for 15–20% of the global fleet. Hardware with efficiency above 25 joules per terahash has been largely phased out.
In January 2026, Winter Storm Fern caused a temporary 30–40% hashrate plunge, dropping the network from 1.1 ZH/s to as low as 663 EH/s over several days. While weather-driven, the episode exposed the concentration of mining infrastructure in the United States, which — together with China and Russia — controls approximately 68% of global hashrate, according to KuCoin's network analysis.
The April 2024 halving reduced the block subsidy from 6.25 BTC to 3.125 BTC. The full economic impact took months to materialize, but by early 2026, the arithmetic became unavoidable.
CoinShares' Q4 2025 mining report, released in March 2026, estimated the weighted average cash cost to produce one bitcoin at nearly $80,000. Electricity costs alone accounted for $52,463 per BTC, up from $44,679 in Q2 2025, reflecting increased difficulty and energy prices. Bitcoin traded at approximately $63,000 as of August 14, 2026 — roughly 49% below its October 2025 all-time high of $126,080, according to Fortune.
The Cambridge Bitcoin Electricity Consumption Index (CBECI) estimated on August 1, 2026 that producing one bitcoin requires approximately 858,000 kilowatt-hours. At a bitcoin price near $63,000, that implies a breakeven electricity rate of roughly 7.3 cents per kilowatt-hour — a threshold that excludes large portions of the global fleet.
Publicly traded miners sold more than 32,000 BTC in Q1 2026, according to CoinDesk. That figure exceeds their combined sales for all of 2025 and surpasses the 20,000 BTC liquidated during the 2022 Terra-Luna collapse. The selling reflects an industry operating above its revenue line.
More than $70 billion in cumulative AI and high-performance computing contracts have been announced across the public mining sector, according to KuCoin. The deals share a common structure: miners contribute electrical infrastructure and site permits; technology companies supply GPU hardware and long-term lease commitments.
The economics are straightforward. A megawatt of capacity allocated to AI hosting generates materially higher and more predictable revenue than a megawatt allocated to bitcoin mining. AI tenants sign multi-year leases at fixed rates; bitcoin mining revenue fluctuates with price, difficulty, and hash price. For an industry sitting on gigawatts of permitted electrical capacity, the pivot requires less new construction than it requires new customers.
VanEck analysts Patrick Bush and Matthew Sigel wrote in a May 2026 report: "This is not a temporary curtailment but a full pivot away from Bitcoin mining." CoinShares projected that some publicly listed miners could derive up to 70% of total revenue from AI hosting by end of 2026, up from approximately 30% in mid-year.
Hut 8 (HUT): Signed a 15-year, $9.8 billion lease for a 352-megawatt Texas facility built to NVIDIA's reference architecture. Total contracted AI portfolio value reached $26.6 billion. CEO Asher Genoot described the company's model as "integrated power and compute," telling CNBC the lease validated the pivot from merchant mining to infrastructure services. Stock up 72% year-to-date through late July.
Core Scientific (CORZ): Signed a 15-year deal with AMD on July 28, 2026, covering up to 2.5 GW of energy capacity, with potential lifetime revenue exceeding $14 billion. AI colocation revenue already accounts for 39% of total revenue. The company sought to raise $3.3 billion through a high-yield bond offering to finance conversion of mining facilities. Stock up 31% year-to-date.
IREN (IREN): Secured a $9.7 billion contract with Microsoft for 76,000 NVIDIA GPUs. Projects AI cloud business will generate more than $4 billion in annual recurring revenue by end of 2026. AI revenue currently at 9% of total but scaling rapidly.
TeraWulf (WULF): HPC leasing now its primary revenue driver, with AI-adjacent revenue at 27% and rising.
Riot Platforms (RIOT): 10-year data center lease with AMD became operational in January 2026. Portfolio includes nearly 2 GW of data center infrastructure capacity. Stock up 83% year-to-date.
Marathon Digital (MARA): The largest publicly traded miner by hash rate has been more cautious. Acquired AI-relevant facilities and announced exploratory partnerships but maintained heavier mining exposure. Stock down 38% year-to-date — the widest divergence from AI-pivot peers.
CleanSpark (CLSK): Emphasized continued bitcoin mining focus with limited AI-adjacent investment. Strategic direction remains more pure-mining than peers.
Bitfarms: CEO publicly stated: "We are no longer a Bitcoin company."
VanEck's June 16, 2026 report introduced skepticism into the AI-pivot narrative. The firm estimated miners face a roughly $50 billion near-term funding gap and as much as $221 billion in long-term capital needs to deliver on contracted AI capacity.
The central finding: the industry has delivered only about 25% of the AI and HPC capacity it has leased to customers. The remainder requires construction of purpose-built data centers, installation of cooling infrastructure, and procurement of GPU clusters — none of which are trivial at the scale promised.
VanEck expects valuations to hinge on energized power capacity and tenant quality. Companies with investment-grade hyperscaler clients (Microsoft, Google-backed entities, AMD) occupy a different risk tier than those with smaller or less creditworthy counterparties. Missing construction milestones, VanEck warned, could result in lasting valuation damage.
The report reframes the market narrative: signing contracts is table stakes; building and energizing the facilities is the hard part. As Sigel put it: "Execution, not signing, becomes the next premium."
A 12% hashrate decline reduces the economic cost of a theoretical 51% attack. However, several mitigating factors limit systemic risk.
The absolute hashrate of 868 EH/s remains orders of magnitude above any historical attack threshold. The cost of assembling sufficient hardware and electricity to sustain a 51% attack at current scale is estimated to exceed the economic value of any plausible double-spend or chain disruption. Geographic distribution across the US, Russia, China, and smaller jurisdictions means no single actor controls a dominant share — though the concentration of 68% in three countries represents a structural vulnerability.
No chain disruptions, reorganizations, or credible attack attempts have been observed during the 2026 drawdown. The difficulty adjustment mechanism, which recalibrates every 2,016 blocks, has functioned as designed — reducing difficulty to match available hashrate and preserving the 10-minute block target.
The deeper concern is structural: if AI contracts permanently redirect electrical capacity away from mining, future hashrate recovery cannot be assumed. Previous drawdowns (China ban, bear markets) were followed by hashrate rebounds as new capacity came online or relocated miners resumed operations. Multi-year enterprise leases do not reverse on a 6-month cycle.
As of August 1, 2026, the CBECI estimated Bitcoin's network power demand at 16.09 GW, with annualized consumption of 141.02 TWh. The modeled range spans 8.61 to 28.19 GW.
The recorded network hashrate of 878 EH/s on July 31, 2026 implied fleet-wide efficiency of approximately 18.3 joules per terahash — a figure consistent with widespread deployment of current-generation ASICs and the retirement of older hardware above the 25 J/TH threshold.
The energy footprint has contracted alongside hashrate. At its peak near 1 ZH/s, the network consumed proportionally more power. The decline represents a natural response to economic incentives: unprofitable machines are switched off, and the remaining fleet operates at higher average efficiency.
An underexamined dimension: megawatts diverted from mining to AI hosting do not disappear from the grid. They shift from proof-of-work computation to inference and training workloads. The environmental calculus changes — the same facilities consume roughly similar power but produce different economic outputs.
The market has drawn a sharp line between miners with AI contracts and those without.
Year-to-date through late July 2026: Riot Platforms up 83%, Hut 8 up 72%, Bitfarms up 50%, Core Scientific up 31%. Marathon Digital, the largest miner by hash rate but the slowest AI adopter, was down 38%.
Citizens initiated coverage in mid-2026 with Outperform ratings on MARA ($24 target), CLSK ($27 target), and BTDR ($35 target), citing the AI compute shift as the primary thesis, with implied upside of 59–106%.
The divergence reflects a market re-rating: investors are pricing mining stocks not on hash rate or bitcoin production but on energized power capacity, tenant creditworthiness, and construction execution timelines. The firms that signed deals with Microsoft, Google-backed entities, and AMD trade at premiums; those with weaker counterparties or no AI contracts trade at discounts.
The Bitcoin mining industry is undergoing a structural transformation. The term "pivot" understates what is happening: major operators are not temporarily reallocating capacity during a price downturn. They are signing 10- and 15-year leases with hyperscalers, issuing billions in debt to finance data center construction, and publicly redefining their corporate identities.
The economic logic is sound. A megawatt of permitted capacity is worth more to an AI tenant paying a fixed, multi-year rate than to a bitcoin miner exposed to volatile hash price, difficulty, and BTC price. The question is whether the industry can execute at the scale it has promised — and VanEck's $50 billion funding gap suggests the answer is not yet clear.
For the Bitcoin network, the implications extend beyond the mining cycle. Previous hashrate drawdowns reversed because the incentive structure eventually attracted new capacity. If AI contracts permanently capture the electrical infrastructure that mining once occupied, the network's security budget depends increasingly on transaction fees and BTC price appreciation — a dependency the protocol was not designed to rely on this early in its emission schedule.
The data is clear: the industry that built itself on converting electricity into bitcoin is now converting its electricity into something with higher margins and longer contracts. Whether that leaves Bitcoin's network adequately secured is a question the market has not yet priced.