CoinShares 2026 Report: Bitcoin Miners Are Facing Their Toughest Time

Author: James Butterfill | Compiled by: Wu Shuo Blockchain

TL;DR: Key Takeaways from the 2026 Q1 Bitcoin Mining Report

• Profitability under extreme pressure: 2025 Q4 became the toughest quarter since the halving, driven by a pullback in BTC price combined with high hash rate. Hashprice at one point fell below $30/PH/day to a five-year low. Around 15–20% of the network’s older mining rigs fell into losses.

• AI transition accelerates: Public mining companies have cumulatively announced AI/HPC contracts totaling more than $70 billion. The capital markets are giving the AI narrative a very high premium (valuation multiples up to 12.3x). The industry is accelerating its split into “infrastructure providers” and “pure-play mining companies.”

• Temporary hash rate retreat: Due to a combination of margin compression, winter power curtailments, and regulatory crackdowns, Q4 network-wide hash rate fell about 10% from its peak. However, industry forecasts still point to resilience: network-wide hash rate is expected to rebound and rise to 1.8 ZH/s by the end of 2026.

• Reshaping costs and debt: AI construction caused some hybrid miners (e.g., CIFR, WULF) to see a sharp increase in their all-in cost per BTC on paper while carrying massive debt. By contrast, low-leverage miners such as CLSK and HIVE have demonstrated strong financial discipline and advantages in pure mining costs.

• Core conclusion: The mining industry is undergoing a deep structural reshuffle. If BTC price fails to rebound to above $100,000 in 2026, high-cost miners will accelerate capitulation. Meanwhile, operators with ultra-low energy costs or those that successfully pivot into AI will dominate the future capital markets.

I. Executive Summary

The fourth quarter of 2025 was the most difficult quarter Bitcoin miners have faced since the April 2024 halving. A sharp pullback in the Bitcoin price (from an all-time high of about $124,500 in early October to about $86,000 by late December, a drawdown of about 31%), combined with network hash rate nearing its all-time high, compressed hashprice to a five-year low.

In the fourth quarter of 2025, the weighted average cash cost per BTC mined by publicly listed mining companies rose to approximately $79,995.

This quarter highlights three core themes:

Pressure on profitability: Hash price fell to about $36–38/PH/day, which is close to or even at breakeven for many miners. Three consecutive mining difficulty adjustments downward (the first consecutive cut since July 2022) marked “miner capitulation.” Entering the first quarter, hash price fell further to $29/PH/day, implying miners must endure even more pain.

AI/HPC pivot accelerates: The split between pure-play mining companies and infrastructure firms turning toward AI has intensified. At present, the entire publicly listed mining sector has cumulatively announced AI/HPC (high-performance computing) contracts worth more than $70 billion. WULF, CORZ, CIFR, and HUT are essentially evolving into data center operators that also mine Bitcoin.

Capital structure reshapes: Multiple miners are carrying large amounts of debt to raise funding for AI infrastructure buildouts. IREN currently has $3.7 billion in convertible notes; WULF’s total debt is $5.7 billion; CIFR issued $1.7 billion in senior secured notes. The industry’s total leverage ratio has fundamentally changed its risk profile.

II. AI and Bitcoin Mining Compete for Rack Space

AI is continuously competing for rack space in many data centers. In the long run, this could push Bitcoin mining toward more intermittent and cheaper power sources.

Bitcoin miners’ migration toward AI and high-performance computing (HPC) is accelerating rapidly. According to recent company announcements, by the end of this year, as much as 70% of the revenue of publicly listed mining companies could come from AI, up from about 30% currently. What started as a marginal diversification strategy is increasingly becoming the core business.

During 2025 and the early part of 2026, Bitcoin miners signed multiple GPU co-location and cloud service agreements with hyperscalers worth more than $70 billion in total. Although most agreements are planned around building new data centers, it is still highly likely that these deals will cannibalize existing mining facilities and lead to shutdowns. As the contracted capacity gradually ramps up over 2026, Bitcoin mining’s share of these operators’ revenue will decline significantly.

This shift is largely driven by economics. Hash price remains near cyclical lows, compressing mining profit potential, while AI infrastructure structurally offers higher and more stable returns. Against this backdrop, reallocating electricity and capital toward high-performance computing (HPC) looks very reasonable—especially for operators with scalable energy access and existing data center capabilities.

That said, this pivot is not uniform. Some miners, such as IREN and Bitfarms, are actively repositioning themselves as HPC providers—effectively using mining as a bridge into AI infrastructure. Other miners, such as CleanSpark, continue to prioritize mining in the near term, monetizing capacity that has recently been developed while gradually expanding into AI.

A third group remains committed to Bitcoin mining, but is evolving its operating model. These operators are no longer chasing ultra-large facilities. Instead, they focus on the lowest-cost energy sources, which are often intermittent—for example, stranded renewables or flare gas. For example, Marathon has deployed a smaller, approximately 10 MW localized containerized site at the edge of an energy network. Such configurations work well for mining operations that can tolerate power interruptions, but they are incompatible with AI workloads that require nearly continuous, uninterrupted operation.

Load balancing will likely remain a persistent niche within mining. By providing demand-side flexibility to grids such as ERCOT, miners can obtain more favorable electricity pricing. The importance of this role may increase over time, though it may become more attractive to smaller, more specialized operators.

A key unanswered question is how durable this AI-driven transition is. While today’s economics strongly favor AI, mining remains highly sensitive to BTC prices. If mining profitability sees a material recovery, some operators may reassess capital allocation between these two lines of business. In that sense, the current trend may not be a permanent pivot, but rather a result of relative return dynamics.

Looking further ahead, this could mean the pure-play mining cohort shrinks, while hybrid infrastructure companies spanning both mining and AI become more common. At the same time, new entrants may emerge to develop the niches vacated by established firms—especially in markets with constrained energy supply or highly flexible conditions.

There is a massive cost gap between Bitcoin mining infrastructure (about $0.7 million to $1.0 million per MW) and AI infrastructure (about $8.0 million to $15.0 million per MW). This conversion opportunity is being realized at scale:

CORZ: Approximately 350 MW of HPC has been powered on, and about 200 MW is being billed. A 12-year contract with CoreWeave expands to $10.2 billion. The goal is full commissioning of 590 MW by early 2027.

WULF: The Lake Mariner site has 39 MW of critical IT capacity online. Signed total HPC revenue is $12.8 billion. Other facilities are on track to proceed before Q4 2026. The platform will expand to five locations, with total capacity of about 2.9 GW.

CIFR: Working with Fortress Credit Advisors to develop the 300 MW Barber Lake site. A multibillion-dollar Fluidstack agreement has been reached (backed by Google). No revenue has been generated yet.

IREN: Scale expanded to more than 10,900 NVIDIA GPUs. The Childress Horizon phases 1–4 expansion (up to 200 MW liquid-cooled GPUs). Fourth-quarter AI cloud service revenue reached $17.3 million.

HUT: At its River Bend campus in Louisiana, HUT signed a 15-year, $7.0 billion, 245 MW leasing agreement with Fluidstack. The first data hall is planned to come online in early 2027.

The failed merger between CORZ and CoreWeave (shareholder vote rejected on October 30, 2025) highlights the tension between infrastructure value and equity value. Because CORZ improperly capitalized assets committed to be demolished during the HPC transition, it later restated its financials, showing the complexity of its accounting treatment as well.

Revenue contribution remains in the early stage but is growing: CORZ’s hosted AI/HPC data centers accounted for 39% of its Q4 revenue; WULF’s HPC business accounted for 27%; IREN’s AI cloud business accounted for 9%; HIVE’s HPC business accounted for 5%. Although mining still dominates, it is clear that AI’s revenue contribution will continue to rise broadly.

III. Network-Wide Hash Rate

In late August 2025, the Bitcoin network reached an important milestone—hash rate first surpassed 1 ZH/s. In early October, network-wide hash rate peaked at about 1,160 EH/s.

However, there was a significant reversal in Q4. Network-wide hash rate fell by about 10% from the October peak, landing at about 1,045 EH/s by late December (and then dipped further to 850 EH/s in early February before recovering somewhat). This was accompanied by three consecutive downward mining difficulty adjustments, the first consecutive cut since July 2022. The main drivers were:

BTC price pullback causing older S19-era miners to fall below breakeven points (S19 XP breakeven electricity price fell from about $0.12/kWh in December 2024 to about $0.077/kWh in December 2025).

Higher winter energy costs and ERCOT curtailment measures, which caused mining hours that were not economical to surge sharply between November and December.

Regulatory enforcement restarted in China’s Xinjiang region (the December 2025 inspection campaign restricted mining operations, even though those capacities were not permanently moved).

Despite the short-term decline, the Bitcoin network still added about 300 EH/s of hash rate throughout 2025. As of the time of writing, network-wide hash rate is basically maintained at the end-2025 level, around 1,020 EH/s.

Although the recent hash rate pullback may look concerning, observing it on a logarithmic scale shows that its severity is far less than China’s 2021 mining ban. This is more the result of cyclical factors combined with weather effects, rather than a sign that the industry will face a more severe crisis. The subsequent strong rebound in hash rate further indicates that many miners still view mining as an economically viable business activity.

Based on the segmented forecasting model (piecewise prediction model) we discussed in detail earlier, we currently expect network-wide hash rate to reach 1.8 Zetahash (ZH/s) by the end of 2026 and 2 ZH/s by the end of March 2027—pushed back by one month compared with our previous prediction.

Geographic shifts in hash rate: The top three countries (the United States, China, and Russia) control about 68% of global hash rate. The U.S. market share increased by about 2 percentage points quarter-over-quarter. Driven by miners such as HIVE (a 300 MW project in Paraguay) and BTDR (a 40 MW project in Ethiopia), emerging markets including Paraguay, Ethiopia, and Oman have successfully moved into the global top ten.

IV. Hashprice Dynamics

Hashprice (the metric that determines miners’ revenue per unit of hash rate) reached a peak of about $63/PH/day in July 2025, then continued falling throughout Q4. By November, it had dropped to about $35–37/PH/day, setting a five-year low at that time. It briefly rebounded to about $38–40/PH/day by late December and early January, but that proved short-lived. Hashprice crashed again after entering Q1 2026, falling to about $28–30/PH/day by early March, setting a new historical low since the halving.

This decline was caused by multiple factors stacking together: record-high mining difficulty (after being increased by 6.31% on October 29 to reach a peak of 155.97T), a weak BTC price (down about 31% from the October all-time high), and extremely low transaction fee revenue (consistently below 1% of total block rewards, with an average fee of about 0.018 BTC per block).

This created the harshest profit environment since the April 2024 halving. With average industrial electricity prices at $0.05/kWh (S19 XP at $0.077/kWh), miners running in-generation rigs (e.g., S19j Pro with roughly 29.5 J/TH) were already operating far below breakeven by year-end, and the situation worsened further after 2026.

Latest forecast: The deterioration in the hashprice environment has exceeded our prior expectations. In late February, it briefly touched about $28/PH/day. As of the time of writing, it has rebounded to about $30–35/PH/day. At current levels, miners running in-generation rigs need an electricity price below $0.05/kWh to remain cash-profitable. Meanwhile, the newest generation models (energy efficiency below 15 J/TH) can still maintain attractive profit margins under typical industrial electricity prices. For hashprice to sustain a recovery above $40, BTC would need to rebound to $100,000 by year-end, and the magnitude of the price increase must outpace the ongoing growth in network-wide hash rate.

Unless BTC sees a material rebound, we expect high-cost operators to face further “miner capitulation” in the first half of 2026. The current mining economics are insufficient to trigger a large-scale hardware refresh cycle. Hashprice must first fall further, forcing enough old capacity and operators to shut down and exit—thereby reducing network-wide hash rate and mining difficulty—to provide an entry window for new Bitcoin miners, or enough motivation for upgrades at existing operating nodes. However, despite relentlessly squeezed profit margins, network-wide hash rate still shows remarkable resilience. This may be supported by multiple factors: including state-backed mining activities oriented toward strategic goals rather than pure economic returns; operators that can access extremely cheap or stranded power; and ASIC manufacturers routing unsold inventory into their own facilities to maintain order commitments at wafer foundries such as TSMC and Samsung.

Mining industry pain has triggered large-scale miner sell-offs and capitulation. Publicly listed miners’ BTC treasury holdings have collectively decreased by more than 15,000 BTC from their peak. Core Scientific, for example, sold about 1,900 BTC in January alone (about $175 million) and plans to liquidate nearly all remaining holdings in Q1 2026. Bitdeer cleared its treasury to zero in February. Riot sold 1,818 BTC in December 2025 (about $162 million).

We believe it is not unrealistic to assume BTC could rebound to the $100,000 level. If it reaches that price, hashprice would recover to $37/PH/day. If the price remains below $80,000 for the rest of this year and assuming mining difficulty continues to rise, we expect hashprice will keep falling. But in that scenario, the real outcome could differ: as miners shut down unprofitable rigs, network-wide hash rate could fall further, making hashprice more likely to stabilize. If we see price begin to test the historical high of $126,000, hashprice could surge to $59/PH/day.

The decline in hashprice has already far exceeded our prediction range. Although we believe it is a temporary phenomenon driven by recent weakness in coin prices, we expect it to gradually stabilize in the $30 to $40/PH/day range.

At current hashprice levels, running multiple mining rigs has become unprofitable. At $30/PH/day, any miner with performance below S19 XP and facing electricity prices of $0.06/kWh (6 cents per kWh) or higher is operating at a loss— — we estimate that this equipment represents about 15% to 20% of the global active miner fleet.

V. Mining Cost Analysis

1. Overview

The table below shows the per-BTC cost breakdown for all mining companies included in our study scope for Q4 2025. All data is denominated as the U.S. dollar cost per BTC mined and uses the revenue-share methodology described in the appendix to allocate relevant costs to the self-mining business.

Key observations:

AI/HPC construction is distorting the all-in per-BTC cost metrics of hybrid operators. Debt arising from AI infrastructure buildouts, along with selling, general, and administrative expenses (SG&A), and depreciation and amortization (D&A), are being allocated across a shrinking BTC production base—thereby inflating the headline cost-per-BTC figures. For companies such as WULF, CORZ, and CIFR, their all-in cost increasingly reflects the economics of operating as a data center operator rather than the pure economics of Bitcoin mining.

Compared with Q2 2025, power costs across the industry increased materially and sharply. This reflects higher mining difficulty diluting BTC production, rising winter energy costs, and the drop in BTC price.

Depreciation and amortization (D&A) is the largest non-cash cost component, and it varies hugely across companies due to different depreciation policies. MARA’s $136,000/BTC and CIFR’s $88,000/BTC are outliers (MARA has a large mining fleet; CIFR uses a 3-year useful life depreciation assumption).

Share-based compensation (SBC, equity incentive costs) remains an important differentiating factor. HUT’s $48.5k/BTC (mainly a one-time reward to the CEO/CSO) and CORZ’s $35.5k/BTC are outliers. BTDR ($3,900/BTC) and CLSK ($6,700/BTC) show the strictest financial discipline.

Interest costs are currently a significant driver for several miners. WULF ($145k/BTC), CIFR ($56k/BTC), and BTDR ($16k/BTC) carry large debt burdens. By contrast, HIVE ($320/BTC) and CLSK ($830/BTC) have very low leverage and significant structural advantages.

2. Company-by-company Breakdown

MARA (MARA Holdings)

BTC output: 2,011

All-in cost: $153,040/BTC

Cash cost (before tax): $103,605/BTC

In Q4, MARA produced 2,011 BTC. Measured by output, it remains the largest publicly listed miner. As of late December, the company’s energized hashrate reached 53.2 EH/s (up 15% this quarter). However, due to rising network difficulty, average daily output fell to about 21.9 BTC—below prior quarters.

Its power cost was $64,703/BTC, in the middle among peers. This reflects its geographic diversity and heavy reliance on third-party hosting (in total power cost of $130.1 million, third-party hosting accounts for $79.4 million). Depreciation and amortization (D&A) was as high as $136,166/BTC, the highest among peers, reflecting its large mining rig scale (full-year D&A of $772.8 million).

The headline all-in cost is severely distorted by a $183.4 million income tax benefit, arising from fair value adjustments to its BTC holdings under ASU 2023–08 accounting guidance. Excluding this non-operating gain, all-in cost jumps to $240,407/BTC. In Q4, MARA maintained its “HODL” strategy and did not sell BTC, while leaving 7,377 BTC in a third-party lending arrangement. However, the company began softening this stance as early as Q3 2025, allowing sales of newly mined BTC to fund operations. In its 10-K filed on March 2, 2026, MARA further expanded the policy, authorizing sales from the reserves of all 53,822 BTC on its balance sheet. This change is partly due to pressure on its $350 million bitcoin-backed credit facility—as BTC fell toward $68,000 in early 2026, its loan-to-value (LTV) rose to about 87%. This marks a substantial departure from the comprehensive HODL strategy adopted in July 2024.

In addition, the company announced a partnership with Starwood Capital on AI and HPC data centers, and it acquired 64% equity in Exaion for $174.5 million in February 2026, indicating it is accelerating diversification beyond pure mining.

IREN (IREN Limited)

BTC output: 1,664

All-in cost: $140,441/BTC

Cash cost: $58,462/BTC

Thanks to favorable power agreements at its Childress facility in Texas and $1.8 million of demand response revenue in Q4, IREN achieved the lowest power cost per BTC, at $34,325. Its installed hashrate reached 46 EH/s, with fleet efficiency of about 15 W/T.

Its share-based compensation (SBC) was $31,717/BTC, second among peers. In Q4, SBC was $58.2 million, up 7.3x year-over-year, driven mainly by options with a $75 strike price and vesting of a large number of restricted stock units (RSUs). Payroll taxes associated with SBC increased actual cash costs by $6.8 million. Depreciation and amortization (D&A) nearly doubled year-over-year to $99.2 million, reflecting the expansion of the Childress project.

IREN is burdened with convertible notes across five series (2029–2033) totaling $3.7 billion. Measured at face value, this is the heaviest debt burden among peers. However, because the coupon rate is low (2.75%–3.50%), interest expense remains in a controllable range. The $111.8 million debt conversion inducement (non-cash) and the $182.5 million deferred income tax benefit are excluded from the cost analysis. Its AI cloud service revenue reached $17.3 million (9% of total revenue), while Horizon phases 1–4 GPU expansion projects (up to 200 MW) are under construction.

CLSK (CleanSpark)

BTC output: 1,821

All-in cost: $118,932/BTC

Cash cost (before tax): $71,188/BTC

CleanSpark has demonstrated excellent operating discipline. Its selling, general, and administrative expenses (SG&A) were $17,848/BTC, and its share-based compensation (SBC) was $6,662/BTC, the lowest among peers. A 100% allocation ratio (pure mining, no hosting/HPC revenue) simplifies the cost analysis.

Power cost was $52,463/BTC, up from $44,679/BTC in Q2, reflecting increased mining difficulty. With installed capacity around 50 EH/s, fleet efficiency of about 16 W/T remains industry-leading. Depreciation and amortization (D&A) was $58,381/BTC, roughly in line with peers. Interest expense was extremely low ($830/BTC), reflecting its low-leverage balance sheet.

New CEO Matt Schultz (taking over from Zach Bradford in August 2025) said that if market conditions allow, the company’s hash rate scale could rise to about 60 EH/s. The company is exploring diversification of equipment suppliers to reduce dependence on Bitmain. No explicit AI/HPC plan has been announced yet, but management hinted it may monetize data center assets located near major metropolitan areas (Georgia facilities). Note: CLSK’s fiscal year ends on September 30, meaning the current data belongs to its Q1 2026.

RIOT (Riot Platforms)

BTC output: 1,324

All-in cost: $170,366/BTC

Cash cost (before tax): $102,538/BTC

Riot produced 1,324 BTC with average deployed hashrate of 31.5 EH/s. ERCOT demand response credit limits of $9.9 million in Q4 (total $56.7 million for the full fiscal year) benefited its power cost significantly, effectively offsetting total power cost.

SG&A was $31,534/BTC, the highest among peers, reflecting corporate overhead expenses and development spending for its 1 GW Corsicana project. SBC reached $21,586/BTC, at a relatively high level. Depreciation and amortization (D&A) was $66,900/BTC, reflecting continued investment in mining rigs. As of December 31, the company held 17,722 BTC (valued at more than $1.5 billion based on end-of-period prices).

Riot’s strategic focus is centered on the Corsicana project, where 600 MW has been designated for AI workloads. Although this is a major long-term opportunity, the vast majority of revenue in Q4 was still driven by mining. The 1 GW point-of-presence capacity makes Riot one of the largest single-site facility operators in North America.

CORZ (Core Scientific)

BTC output: 421

All-in cost: $168,693/BTC

Cash cost: $110,282/BTC

Q4 was a milestone for CORZ’s transition to AI/HPC. Hosting revenue reached $31.3 million (39% of total revenue, up from $8.5 million in Q4 2024). Because capacity was intentionally shifted to the HPC segment, self-mining revenue fell year over year from $79.9 million to $42.2 million.

Lower BTC output (421 BTC) pushed up per-BTC metrics across the board. SG&A was $47,510/BTC and SBC was $35,506/BTC, both the highest among peers. This reflects corporate overhead expenses as well as costs resulting from the failed CoreWeave merger. Fleet efficiency of about 24.7 W/T lagged peers (15–18 W/T), driving power cost as high as $66,720/BTC.

The failed CoreWeave merger (October 30, 2025) created uncertainty, but execution work continues: about 350 MW has been powered on, and about 200 MW is being billed. The target is to fully commission 590 MW by early 2027 (contract amount totaling $10.2 billion over 12 years). Because assets committed to be demolished during the HPC conversion were improperly capitalized, CORZ made a major financial restatement for 2024–2025, leading auditors to switch to KPMG and with internal controls deemed to have failed. D&A was $17,701/BTC, the lowest among peers, which partly reflects impairment of the restated assets.

WULF (TeraWulf)

BTC output: 262

All-in cost: $471,841/BTC

Cash cost: $384,517/BTC

Important note: WULF’s all-in cost per BTC data is not comparable to that of pure-play mining peers.

The company has fundamentally transformed into an AI/HPC infrastructure company, while only maintaining a shrinking mining business. The 262 BTC mined this quarter was generated together with $9.7 million of HPC rental income.

QoQ mining revenue in Q4 fell 40% to $26.1 million. QoQ HPC rental revenue rose 35% to $9.7 million (27% of total Q4 revenue). Total revenue for fiscal 2025 was $168.5 million, of which the HPC segment contributed $16.9 million.

Extremely high all-in costs reflect the following: interest up to $144,974/BTC (total debt of $5.7 billion: including $2.5 billion convertible notes and $3.2 billion senior secured notes under WULF Compute); SG&A of $167,221/BTC (mainly from expanding headcount and milestone compensation); D&A of $77,217/BTC (newly built HPC infrastructure). By year-end 2025, cash reserves reached $370 million (up from $27.4 million), reflecting large-scale capital formation. The company has signed for 522 MW of capacity, involving $12.8 billion in long-term customer contracts.

CIFR (Cipher Digital)

BTC output: 591

All-in cost: $231,980/BTC

Cash cost: $103,516/BTC

CIFR’s all-in cost is the second highest (excluding WULF), driven mainly by D&A up to $87,768/BTC (D&A based on a 3-year useful life assumption adopted in 2024) and interest expense of $56,445/BTC.

The surge in interest is the decisive feature of CIFR’s Q4: in November 2025, it issued $1.733 billion of senior secured notes with an interest rate of 7.125%, causing Q4 interest expense to jump to $33.4 million, while total interest over the first nine months was only $3.2 million. Power cost was $41,047/BTC, highly competitive (the power purchase agreement at the Odessa site is roughly $0.028/kWh). SBC reached $40,695/BTC, at a relatively high level, and is classified under “compensation and benefits” rather than within SG&A (this presentation is not typical).

Major asset impairments in Q4 (Odessa miner impairment of $45.3 million, Black Pearl impairment of $96.1 million, and asset disposal losses of $29.4 million) are excluded from the cost analysis. The company changed its name to Cipher Digital Inc. on February 20, 2026. In terms of HPC, the 300 MW Barber Lake site (in partnership with Fortress) and the Fluidstack agreement (backed by Google) provide the foundation for CIFR’s diversification, though no revenue has been generated yet.

HUT (Hut 8 Corp.)

BTC output: 719

All-in cost: $160,402/BTC

Cash cost: $50,332/BTC

Hut 8’s headline all-in cost appears competitive, but multiple one-off items require caution in interpretation.

Its SBC is as high as $48,527/BTC, the highest among peers, mainly driven by equity awards issued to the CEO and CSO in November 2025 (2.3 million RSUs and PSU performance share units). SBC in Q4 was $39.7 million, while the total for the first nine months was only $18.1 million— — a 2.2x ratio. If SBC were normalized, all-in cost would drop significantly.

Because it received a $17.8 million Canada HST refund in December 2025, its general and administrative expenses (G&A, excluding SBC) of $7,413/BTC look artificially low. Normalized G&A should be close to about $30,000/BTC. Depreciation and amortization (D&A) of $48,621/BTC is presented at the consolidated reporting level; D&A related to mining is actually lower (because about 74% of property, plant, and equipment (PP&E) is mining-related). Interest expense of $6,840/BTC reflects total debt of about $411 million (including TZRC debt with an interest rate of 15.25%, Coinbase debt at 9%, and Coatue convertible notes at 8%).

Thanks to Bitmain mining rigs at the Vega site (hashrate of 14.86 EH/s), its BTC output rose from 578 in Q3 to 719. The company currently holds 15,679 BTC (valued at about $1.37 billion). Its complex business structure (including four business segments, an ABTC subsidiary, and intercompany eliminations) makes clear cost attribution extremely challenging. A $78.2 million income tax benefit in Q4 (a reversal of deferred income taxes) is excluded from the calculation.

BTDR (Bitdeer Technologies Group)

BTC output: 1,673

All-in cost: $118,188/BTC

Cash cost: $87,144/BTC

Bitdeer’s all-in cost is highly competitive among peers, though this partly reflects IFRS reporting conventions and multi-segment revenue (SEALMINER miner sales revenue of $23.4 million, and HPC/AI revenue of $2.3 million). Average power cost increased from $43/MWh in Q3 to $46/MWh.

The most notable issue is a change in depreciation policy in Q4: management shortened the useful life of mining rigs, causing depreciation and amortization (D&A) within cost of revenue from self-mining (CoR) to double QoQ (from $31.2 million to $63.9 million), even though its hashrate increased by about 60%. Gross margin for self-mining dropped from 27.7% in Q3 to 3.6%. This is purely the result of accounting treatment, not a deterioration in operational conditions.

Under IFRS presentation, D&A and SBC are bundled within cost of revenue (CoR), making comparisons with peers reporting under US GAAP more complex. Interest expense of $16,306/BTC reflects about $1.0 billion in convertible notes and related party borrowings. BTDR’s proprietary ASIC chip strategy (SEALMINER A2 with 16.5 W/T energy efficiency, and A3 expected to reach mass production with 9.7 W/T) is a significant competitive advantage; compared with buying Bitmain miners, it substantially reduces capex per TH.

HIVE (HIVE Digital Technologies)

BTC output: 884

All-in cost: $144,321/BTC

Cash cost: $75,274/BTC

In Q4 (its third fiscal quarter ended December 31), HIVE mined 884 BTC, driven by expansion in its Paraguay business, resulting in a significant increase in output. Fleet efficiency improved from 21 W/T to 18.5 W/T.

Power cost of $65,368/BTC is the highest among peers (excluding WULF). This is boosted by a forward-looking accounting change: HIVE capitalized $41.3 million of non-refundable Paraguay VAT into property and equipment (PP&E), and expensed $5.5 million of electricity-related VAT through operating expenses (opex). This accounting treatment increases both its D&A and power costs relative to peers.

Its SG&A of $9,054/BTC is among the lowest. SBC of $7,501/BTC is moderate (corresponding to RSUs issued at CAD 7.30 in October 2025). Interest expense is only $320/BTC, the lowest among peers—HIVE’s total debt is only $13.8 million, giving it a significant structural advantage. During the quarter, its 100 MW Valenzuela facility came online; HIVE currently has power purchase agreements totaling 300 MW in Paraguay through ANDE.

The company faces contingent value-added tax liabilities of about $79.2 million (stemming from assessments by the Swedish Tax Agency on its Bikupa subsidiary), which are currently in the appeal stage in court. It uses BTC as deposits with buyback options attached to the BTC—an unusual form of capital management.

BITF (Bitfarms)

Will be updated after Bitfarms releases its Q4 financial report.

VI. Stock Performance and Valuation of Miners

In Q4, the valuation premium for AI/HPC continued to expand. Currently, the EV/NTM sales multiple for miners with HPC contracts reached 12.3x, while pure-play mining companies are only at 5.9x. The drop in BTC price in Q4 (down 31% from the all-time high) created a double drag: not only reduced mining revenue, but also significantly shrank the value of BTC holdings in miners’ treasuries.

The valuation discount that CORZ experienced after the failed merger (possibly due to hedge funds closing positions) contrasts sharply with the valuation premiums enjoyed by WULF, CIFR, and HUT. The industry’s short interest is currently elevated. As of the time of writing, MARA’s short interest is about 30% of its float.

The industry has fundamentally bifurcated into “infrastructure companies” (such as WULF, CORZ, CIFR, HUT) and “mining companies” (such as MARA, CLSK, RIOT, HIVE). Whether these AI-oriented, high valuation multiples are justified ultimately depends on execution: not all announced agreements will translate into actual operational infrastructure, and the capital requirements behind them are still extremely large.

VII. Outlook for Q1 2026 and Beyond

1. Hashprice recovery depends on BTC price: When BTC is around $70,000 and hashprice is around $30/PH/day, many in-generation miner fleets are already at or below breakeven. If price stays below $70,000, it could trigger a larger scale of “miner capitulation,” but that would also benefit survivors by reducing mining difficulty and network-wide hash rate.

2. Deployment of next-generation hardware: Bitmain’s S23 series and SEALMINER A3 (with energy efficiency both below 10 J/TH) are expected to be deployed at scale in H1 2026. This will further widen efficiency gaps and accelerate the upgrade cycle for miner fleets.

3. AI/HPC revenue inflection point: CORZ targets full delivery of its CoreWeave project with 590 MW capacity in early 2027. WULF’s Lake Mariner site expansion is still ongoing. The market will closely watch whether contracted revenue can convert into actual billing and whether margins can reach the target of above 85%.

4. Leverage differentiation creates M&A catalysts: Miners with healthy balance sheets and good liquidity (such as HIVE and CLSK) are expected to become acquirers. However, even CLSK has taken on a substantial amount of convertible debt ($1.15 billion, zero coupon) to fund its transition to AI infrastructure.

5. Changes in regional distribution and regulatory environment: The U.S. continues to expand its market share. Paraguay and Ethiopia are becoming emerging mining hubs. Regulatory enforcement actions in China’s Xinjiang may prompt hash rate to shift overseas. Texas’ SB 6 bill (signed in June 2025) imposes new requirements on connecting large mining and data center power loads to ERCOT, including mandatory remote power shutoff capability.

6. Industry integration: We expect more M&A activity in 2026. The efficiency gap between leading fleets (around 15 W/T) and lagging fleets (above 25 W/T) is already large enough that directly acquiring efficient capacity could be cheaper than upgrading older operating facilities.

Appendix: Methodology

Denominator: The number of BTC produced from self-mining during the quarter.

Allocation ratio: Self-mining revenue / total revenue. This ratio is applied to allocate selling, general, and administrative expenses (SG&A), depreciation and amortization (D&A), share-based compensation (SBC), interest and taxes.

All-in cost per BTC = electricity cost (net of curtailment compensation) + SG&A (excluding SBC) + D&A + net interest + income tax + SBC — — Where applicable, all items are allocated in proportion to mining revenue.

Cash cost per BTC = cost of revenue (excluding D&A) + SG&A (excluding SBC) + net interest + income tax — — All items are allocated proportionally.

Electricity cost: Net of curtailment / demand response credit limits. This cost analysis excludes asset impairments, fair value remeasurement, and non-operating items (e.g., BTC remeasurement gains/losses, changes in derivative fair value, debt conversion inducement fees).

Numerical units: Unless otherwise noted, all values are in thousands of dollars. Financial data not denominated in USD has been converted using the average quarterly exchange rate.