Public Bitcoin (BTC) mining companies sold more than 25,000 BTC in the first quarter of 2026, liquidating treasury reserves to repay debt and fund a generational pivot toward AI data center infrastructure.

The sell-off has sparked fierce debate over whether corporate treasuries are abandoning Bitcoin or simply managing balance sheets under post-halving stress. The answer depends on which type of company is doing the selling.

The exodus also raises questions about Bitcoin's long-term network security as hashrate declines and transaction fees fail to replace shrinking block subsidies.

TL;DR

• Major Bitcoin miners including MARA, Riot, Bitdeer, and Bitfarms sold billions of dollars in BTC during Q1 2026, with some reducing holdings to zero
• Post-halving mining costs now exceed Bitcoin's market price, pushing miners toward AI data center contracts worth over $70 billion collectively
• Strategy (formerly MicroStrategy) accounts for roughly 98% of all recent corporate Bitcoin buying, exposing a dangerous concentration risk in the treasury trend

Who Holds Bitcoin on Corporate Balance Sheets

The universe of public companies holding Bitcoin splits into two distinct categories.

The first consists of miners who produce BTC as part of their core operations. The second includes non-mining companies that adopted Bitcoin as a treasury reserve asset.

Among miners, MARA Holdings held the largest position heading into 2026 with roughly 53,800 BTC.

Riot Platforms maintained approximately 19,200 BTC. CleanSpark carried around 13,900 BTC, while Hut 8 held close to 10,100 BTC.

Bitdeer Technologies and Bitfarms held smaller treasuries but made the most dramatic moves. Core Scientific also carried a sizable reserve of roughly 9,600 BTC at peak levels. IREN, formerly known as Iris Energy, held a modest amount relative to its power capacity.

On the non-mining side, Strategy dominates the landscape with 766,970 BTC. Twenty One Capital, the Tether and SoftBank-backed entity that began trading on the NYSE in Dec. 2025, holds 43,514 BTC. Metaplanet, a Tokyo-listed company, reached 40,177 BTC after adding 5,075 BTC in Q1 2026. Coinbase holds roughly 15,389 BTC, Tesla retains 11,509 BTC with no recent activity, Block carries 8,780 BTC, and GameStop holds 4,710 BTC.

The distinction matters because miners and non-mining treasury holders face completely different economic pressures. Miners must cover electricity costs, equipment depreciation, and debt service from operations that have turned unprofitable. Non-mining holders can afford to sit on their reserves indefinitely.

Also Read: Solo Miner Nets $210K Bitcoin Reward On Tiny Hashrate, Against 28,000-to-1 Odds

The Great Miner Liquidation of Q1 2026

MARA executed the largest single corporate Bitcoin sale in history between Mar. 4 and Mar. 25.

The company offloaded 15,133 BTC for approximately $1.1 billion at an average price around $72,700 per coin.

That represented 28% of MARA's total holdings. The treasury dropped from a peak of roughly 53,800 BTC to around 38,700 BTC.

CEO Fred Thiel described the sale as a capital allocation decision designed to strengthen the balance sheet.

Proceeds went toward repurchasing $1 billion in face value of zero-coupon convertible notes at a 9% discount to par. The move cut MARA's total convertible debt from $3.3 billion to $2.3 billion and captured $88.1 million in economic gain.

MARA simultaneously cut 15% of its workforce. Its stock rose 10% on the announcement as investors rewarded the deleveraging.

Riot sold 3,778 BTC for $289.5 million in Q1 2026 at an average of $76,626 per coin. That was 2.5 times what it mined during the quarter. Holdings fell to 15,680 BTC, down 18% year over year.

Riot had already sold 1,818 BTC in Dec. 2025 and another 1,080 BTC in Jan. 2026 to fund a land acquisition in Rockdale, Texas.

Bitdeer went furthest. The company reduced its Bitcoin reserves to zero by Feb. 20, liquidating approximately 2,000 BTC accumulated over prior years.

Bitdeer now sells all production immediately, treating Bitcoin as inventory rather than a treasury asset.

Bitfarms CEO Ben Gagnon made the starkest public declaration on the company's Q4 2025 earnings call. He said the company was no longer a Bitcoin company and planned to hold no Bitcoin in time. Bitfarms plans to rebrand as Keel Infrastructure under the ticker KEEL, re-domicile from Canada to the United States, and pursue a 2.2-gigawatt AI data center pipeline.

Core Scientific sold roughly 1,900 BTC for $175 million in Jan. 2026. Holdings dropped from a peak of 9,618 BTC to approximately 630 BTC, with plans to sell the remainder this year.

Also Read: Bitcoin Rallies Past $69K On Thin Volume, Glassnode Reports

Why Miners Can No Longer Afford to Hold

The Apr. 2024 Bitcoin halving slashed block rewards from 6.25 to 3.125 BTC. That cut miner revenue per block in half overnight. But operating costs did not fall.

CoinShares estimated in its Q1 2026 mining report that the weighted average cash cost to produce one Bitcoin among publicly listed miners reached roughly $80,000 in Q4 2025.

That figure rose to an estimated $80,000 to $88,000 range in Q1 2026.

Bitcoin traded between $60,000 and $69,000 during most of this period. The gap means most miners lose $11,000 to $19,000 on every coin they produce.

Network hashrate fell from a peak of approximately 1,160 exahashes per second in Oct. 2025 to around 920 EH/s. Three consecutive negative difficulty adjustments followed. Hash price collapsed to roughly $28 to $30 per petahash per second per day in early Mar. 2026, a five-year low.

The key cost dynamics pushing miners underwater include:

• Electricity prices averaging $0.045 to $0.065 per kilowatt-hour across major North American facilities, up 15% to 20% year over year
• Equipment depreciation accelerating as next-generation ASIC machines from Bitmain and MicroBT render older rigs obsolete faster
• Rising interest costs on convertible debt issued during the 2024 and 2025 expansion cycle
• The FASB fair value accounting standard forcing quarterly mark-to-market reporting on crypto holdings through income statements

That last point created headline-grabbing losses across the sector. MARA posted a $1.5 billion negative fair value change on digital assets in Q4 2025.

Hut 8 recorded a $401.9 million mark-to-market loss. Strategy reported a $17.4 billion operating loss in Q4 2025. While these figures are non-cash, they spook investors and amplify pressure to deleverage.

CleanSpark stands as one of the more efficient operators with a direct mining cost of roughly $36,100 per BTC. Even so, its all-in cost ratio exceeds 100% of mining revenue when depreciation is included. IREN, at approximately $41,000 per BTC all-in, is one of the rare miners still generating positive margins.

Also Read: Strategy Resumes Bitcoin Buying With 4,871 BTC For $330M

The $70 Billion AI Pivot

The most consequential shift in 2026 is not the Bitcoin selling itself but what it funds. Mining companies possess exactly what hyperscalers need: access to cheap power at scale, land with permits already in hand, and cooling infrastructure.

The economics explain the pivot clearly.

AI hosting generates $1 million to $4 million per megawatt annually with operating margins above 85% and multi-year revenue visibility. Bitcoin mining now generates negative returns for most operators.

The largest contracts signed by former miners include:

• IREN secured a $9.7 billion five-year contract with Microsoft for GPU cloud infrastructure, with a $1.9 billion prepayment
• Hut 8 signed a $7 billion 15-year lease backed by Google for 245 megawatts at its River Bend campus in Louisiana
• Core Scientific contracted roughly $10 billion over 12 years with CoreWeave across 590 megawatts
• Cipher Digital signed deals worth approximately $9.3 billion total, including a 15-year 300-megawatt lease with AWS
• TeraWulf secured $12.8 billion in long-term customer contracts, with Google holding a 14% equity stake
• Applied Digital leased 600 megawatts representing approximately $16 billion in contracted revenue over 15 years

Even companies that initially resisted the pivot have joined the migration. Riot signed a 10-year agreement with AMD for 25 megawatts at its Rockdale facility. MARA partnered with Starwood Capital for a data center joint venture targeting one gigawatt near-term.

CoinShares projects that publicly listed miners could derive up to 70% of total revenue from AI by end of 2026.

Miners with secured HPC contracts now trade at 12.3 times next-twelve-months sales versus 5.9 times for pure-play miners.

CleanSpark sold 97% of its February production to fund AI expansion. CEO Matt Schultz framed the company's model as one where Bitcoin mining funds the platform, AI monetizes it, and digital asset management optimizes it across all cycles.

Also Read: Crypto Developer Numbers Crash To 2017 Levels But That May Not Be Bearish

Strategy Stands Alone as the Dominant Buyer

The contrast between the mining sector's retreat and Strategy's continued accumulation could not be starker. Michael Saylor's company added approximately 94,473 BTC worth $7.65 billion in Q1 2026 alone. That was roughly 40% of what it purchased in all of 2025.

Total holdings now sit at 766,970 BTC at an aggregate cost of approximately $58 billion.

The average purchase price across the entire position is $75,644. That represents 3.8% of all Bitcoin in circulation.

Saylor told CNBC in Feb. 2026 that the company's leverage ratio remained half that of a typical investment-grade company. He said Strategy held two and a half years of dividend payments in cash.

But Strategy's dominance reveals a fragility in the broader corporate Bitcoin thesis. CryptoQuant research found that non-Strategy treasury companies purchased a combined 1,000 BTC in the 30 days ending late Mar. 2026. That represents a 99% decline from the Aug. 2025 peak of 69,000 BTC.

The number of active corporate buyers dropped 76%, from 54 separate purchases in a single 30-day window to just 13. Strategy alone accounts for roughly 98% of recent corporate Bitcoin purchases and holds approximately 65% of all publicly held corporate BTC.

Other companies still accumulating:

• Metaplanet added 5,075 BTC in Q1 2026 at an average of $78,000, reaching 40,177 BTC total
• Twenty One Capital holds 43,514 BTC after going public in Dec. 2025
• Coinbase conducts weekly purchases and holds roughly 15,389 BTC
• Block adds gradually, now carrying 8,780 BTC

Tesla continues to hold 11,509 BTC with no activity since mid-2022. GameStop holds 4,710 BTC but deployed most of its position into a covered-call options strategy. CEO Ryan Cohen teased what he called a transformational acquisition he found more compelling than Bitcoin.

Also Read: Bitmine Acquires 71K ETH In One Week Before NYSE Listing, Nears 4% Of Supply

The Bull-Bear Debate Over Corporate Bitcoin Faith

Geoff Kendrick, Standard Chartered's global head of digital assets research, has emerged as the most prominent institutional voice on the bearish side. He argued in a Dec. 2025 note that buying by treasury companies was likely over because valuations no longer supported further expansion.

He cut his 2026 year-end BTC price target from $300,000 to $150,000 in Dec. 2025, then further to $100,000 in Feb. 2026. He warned of a possible dip to $50,000 first.

The bear case rests on several structural concerns:

• Strategy's market-cap-to-net-asset-value ratio dropped below 1.0 times for the first time since 2023, approaching what analysts call a danger zone
• The company faces $750 million to $800 million in annual preferred dividend obligations across two classes of preferred stock
• CEO Phong Le acknowledged in a risk disclosure that Strategy could sell BTC under specific crisis conditions
• Exchange whale ratios surged from 0.34 in Jan. to 0.79 by late Mar. 2026
• Apparent demand remained negative by roughly 63,000 BTC, meaning new buying fails to keep pace with supply and selling

The bull case rests on different structural arguments. Matt Hougan, Bitwise's chief investment officer, wrote in a client note that concerns about forced selling were wrong. Public companies still collectively hold approximately 1.16 million BTC, representing 5.4% of total supply.

Aggregate Q1 2026 buying between 62,000 and 68,000 BTC actually exceeded Q4 2025 levels.

Several analysts argue the AI pivot could ultimately reduce sell pressure. If miners generate sufficient cash flow from AI hosting to cover operating expenses, they would no longer need to sell mined BTC.

Kendrick himself noted that a key milestone for 2026 would arrive when Bitcoin-backed lending exceeds $100 billion. That would create a cycle of less selling pressure, more utility, and higher prices.

Also Read: Polymarket Launches Own Stablecoin In Major Platform Overhaul

What the Earnings Numbers Show

The Q4 2025 and Q1 2026 financial results paint a picture of an industry in painful transition.

MARA reported full-year 2025 revenue of $907 million, up 38% year over year, but a net loss of $1.31 billion driven by digital asset fair-value changes.

Riot posted record full-year 2025 revenue of $647 million, up 72%, but a net loss of $663 million. Hut 8's full-year net loss hit $248 million. CleanSpark's fiscal Q1 2026 showed a net loss of $378.7 million despite being one of the most efficient operators.

Strategy reported the most dramatic figures: a $17.4 billion Q4 2025 operating loss and $12.6 billion net loss, almost entirely from non-cash mark-to-market effects.

Its stock declined roughly 60% over 12 months. Institutions reduced MSTR holdings by $5.4 billion in a single quarter.

On the revenue side, the AI pivot is beginning to show results. IREN's Q1 fiscal 2026 revenue surged 355% year over year to $240.3 million with a record net income of $384.6 million.

Core Scientific's AI colocation reached 39% of total revenue in Q4 2025. TeraWulf's HPC revenue accounted for 27% of its total.

These early numbers validate the thesis that mining infrastructure can be repurposed for higher returns. But the transition period is punishing.

Also Read: Franklin Templeton Launches Crypto Unit Targeting Pension Funds

Can Bitcoin's Network Absorb the Miner Exodus

The question hanging over the entire sell-off is whether the departure of major public miners poses a real threat to Bitcoin's network. The short answer is complicated. The protocol was designed to survive exactly this kind of shakeout, but the scale and speed of the current transition are testing assumptions that have never been stressed at this level.

Bitcoin's total hashrate fell roughly 4% in Q1 2026, marking the first first-quarter decline in six years.

From a peak of approximately 1.16 zettahashes per second in Oct. 2025, computing power dropped to around 920 to 1,000 exahashes per second depending on the measurement window.

That decline followed five consecutive years of double-digit growth.

Mining difficulty dropped 7.76% in its latest biweekly adjustment on Mar. 21, the second-largest negative adjustment of 2026. An 11.16% plunge in Feb. was even steeper. Difficulty now sits nearly 10% below where it started the year.

The difficulty adjustment mechanism is Bitcoin's built-in shock absorber. Every 2,016 blocks, roughly every two weeks, the protocol recalibrates to keep block production on a 10-minute schedule.

When miners leave, difficulty drops. Remaining miners find blocks more easily. The system self-corrects.

But three risks emerge when the correction happens at this pace:

• A sustained hashrate decline reduces the theoretical cost of a 51% attack, even if such an attack remains economically impractical at current scale
• Slower block production between adjustment periods can temporarily increase confirmation times and fee volatility for users
• Geographic concentration intensifies, since publicly listed U.S. miners have accounted for over 40% of global hashrate, and their retreat could leave the network more dependent on private operators in Russia, China, and emerging markets

The geographic shift is already underway. The Hashrate Index's Q2 2026 heatmap shows global hashrate declining further to 1,004 exahashes per second.

The top three countries, the United States, China, and Russia, still control roughly 68% of total hashpower. But emerging markets including Paraguay, Ethiopia, Oman, and Kyrgyzstan have entered the global top 10 for the first time.

Private and state-backed miners are filling part of the gap. CoinShares noted that network hashrate has proven remarkably resilient despite the margin squeeze, likely supported by state-backed operations with strategic rather than profit-driven mandates, operators with access to exceptionally cheap power, and ASIC manufacturers plugging unsold inventory into their own facilities.

These participants do not need to generate quarterly profits for shareholders. They mine for different reasons.

Smaller private operators face a brutal calculus.

The average all-in cost to produce one Bitcoin reached roughly $88,000 per the Checkonchain difficulty regression model. Bitcoin trades near $67,000 to $69,000.

The math works only for operators with sub-$0.045 per kilowatt-hour electricity contracts and latest-generation hardware running below 15 joules per terahash. Everyone else mines at a loss.

The efficiency divide creates a Darwinian dynamic:

• Industrial miners in Texas, Paraguay, Iceland, and Oman with stranded energy access remain cash-positive even below $70,000
• Mid-generation hardware operators running machines at 25 to 30 joules per terahash are well below breakeven and shutting down
• Retail miners paying commercial electricity rates above $0.08 per kilowatt-hour have been effectively priced out of the network
• An estimated 252 exahashes of marginal capacity already sits offline, mostly legacy hardware that will never return

CoinShares still forecasts hashrate growth to around 1.8 zettahashes per second by end of 2026, but only if Bitcoin recovers toward $100,000.

Without that price recovery, hashrate growth flatlines or contracts further. Next-generation hardware from Bitmain and SEALMINER running below 10 joules per terahash is expected at scale through the first half of 2026, which could widen the efficiency gap and accelerate fleet turnover among survivors.

The transaction fee picture adds another layer of concern. Fees contributed roughly $300,000 per day to miner revenue in late 2025, comprising less than 1% of total miner income. That share collapsed from approximately 7% in 2024, when Ordinals, BRC-20 tokens, and Runes created intense demand for block space. With that on-chain activity boom now faded, miners depend almost entirely on the 3.125 BTC block subsidy plus Bitcoin's market price.

This creates a structural vulnerability. The block subsidy will halve again in 2028 to 1.5625 BTC. If transaction fee revenue does not grow substantially before then, the economic incentive to secure the network weakens further. The Bitcoin community will need to build applications and use cases that generate sustained demand for on-chain block space.

For everyday Bitcoin users, the near-term effects remain modest. Block times have stayed close to the 10-minute target thanks to difficulty adjustments. Transaction fees remain low during this period of reduced on-chain activity. Confirmation speed has not meaningfully deteriorated.

The real risk is not immediate. It is cumulative.

VanEck's head of digital assets research Matthew Sigel noted that miners are sitting on a gold mine in terms of their power capacity's value for AI applications. Every megawatt that moves from Bitcoin hashing to AI hosting reduces the network's long-term security budget. If AI generates more stable and higher-margin revenue than mining, rational operators will keep reallocating capital.

The aggregate result is a lower hashrate and a reduced security budget, even as the protocol itself remains technically unchanged.

History offers some comfort. VanEck's research found that Bitcoin has posted positive 90-day forward returns 65% of the time during periods of shrinking hashrate. The 2021 China mining ban wiped out roughly 50% of global hashpower overnight, and the network recovered within months. The current decline, while structural rather than regulatory, is far smaller in magnitude.

The protocol will survive the miner exodus. That much is clear from its design. The more pressing question is whether Bitcoin's security model can remain robust over the next decade as block subsidies approach zero and fee revenue fails to compensate. The AI pivot has turned that long-term theoretical concern into a near-term operational reality.

Also Read: Bybit Rolls Out Crypto-Powered Fiat Transfers

Conclusion

The corporate Bitcoin selling wave of 2026 does not represent a wholesale loss of conviction in BTC as a treasury asset. The mining industry's economic model broke after the Apr. 2024 halving, and companies with access to cheap power discovered they can generate far more value hosting AI workloads.

The selling is purposeful: retiring debt, funding construction, and financing a generational infrastructure pivot.

The deeper risk lies in concentration. Strategy's dominance as essentially the sole large-scale corporate buyer creates fragility the market has not fully priced. If its net asset value ratio continues deteriorating or preferred dividend obligations become strained, the never-sell thesis faces its first genuine test.

Meanwhile, the emergence of specialized treasury vehicles like Twenty One Capital and Metaplanet suggests demand for corporate Bitcoin exposure is migrating from miners to purpose-built companies. The most important number in this landscape may not be any company's BTC balance but the $70 billion in AI contracts that former miners have signed with counterparties like Microsoft, Google, and Amazon.

Read Next: Is The Worst Over For Stocks? Tom Lee Says 95% Of War Sell-Off Is Done