Top Bitcoin miner MARA open to selling entire $3.8 billion BTC stash creating a new liquidity test

MARA Holdings just rewrote the playbook that has defined Bitcoin mining over the past four years, and the potential outcomes matter for the entire crypto industry.

The company’s March 2 filing authorizes balance-sheet sales of its entire 53,822 BTC treasury, representing a complete reversal of its 2024 “retain all mined and purchased Bitcoin for the foreseeable future” policy.

Bitcoin trades around $68,000, down nearly 46% from late-2025 highs, while market depth has thinned to levels where modest selling creates an outsized impact.

The timing raises a question: what happens when one of the industry’s largest holders treats Bitcoin as working capital rather than as a matter of conviction?

The policy that wasn’t supposed to change

MARA’s 2024 10-K positioned it alongside Strategy as a Bitcoin maximalist.

The pivot began in late 2025, when MARA sold roughly 4,076 BTC for $413.1 million, at an implied average of $101,000 per BTC. The 2026 filing permits balance sheet sales, making Bitcoin “a readily convertible source of liquidity.”

Three factors sharpen the stakes.

First, 15,315 BTC are loaned or pledged as collateral, representing 28% of holdings. That leaves 38,507 BTC unrestricted: $2.6 billion or 60 days of post-halving issuance.

Second, MARA recorded a $422.2 million fair-value decline in 2025 and a $69.1 million trading loss.

Third, MARA partnered with Starwood Capital to develop AI data centers targeting 1 GW, with a path beyond 2.5 GW, a capital-intensive infrastructure that pulls liquidity needs forward.

The logic: fund operations and AI by selling BTC instead of diluting shareholders. The trade-off transforms MARA from a Bitcoin ETF into a capital allocator holding volatile assets.

The timing isn’t random

Regarding “why now?”, three drivers converge.

First, balance sheet pressure. Post-halving, rewards were cut to 3.125 BTC, while difficulty and energy costs squeezed margins.

Output fell 7% to 8,799 BTC despite growing hashrate to 66.4 EH/s. When Bitcoin drops from the $76,000 to $126,000 range to $60,000, liquidity becomes urgent.

The company faces $350 million in convertible notes maturing in 2027.

Second, AI capex. MARA’s Starwood partnership targets sites toggling between Bitcoin mining and AI compute. Starwood leads design and construction; MARA contributes sites and retains up to 50% ownership.

This bets power-to-compute monetization beats post-halving mining returns.

Third, market microstructure. Liquidity has deteriorated since late 2025, with spot volumes running 25% to 30% below year-ago levels. MARA, as a discretionary seller, doesn’t need to crash markets. Instead, it creates an overhang narrative when sentiment is fragile.

MARA formalized this not despite weak conditions, but because weak conditions make BTC sales credible versus costlier funding.

The overhang isn’t just MARA

Public miners collectively hold 116,697 BTC, down 4.42% month over month.

MARA’s 53,822 BTC represents nearly half of the total. The broader pool includes Riot Platforms (18,005 BTC), CleanSpark (13,513 BTC), Hut 8 (10,278 BTC), and Core Scientific (2,537 BTC).

Core Scientific expects to monetize “substantially all” holdings in 2026. In January, it sold 1,900 BTC for $175 million at $92,000 per coin. Bitdeer liquidated its entire treasury in late February.

Miners now treat Bitcoin as inventory to monetize when AI infrastructure economics beat hash-rate expansion.

The question is how quickly and at what scale others will follow, and three scenarios frame the range.

In the conservative scenario, miners sell production, but keep treasuries intact. A 10% non-MARA drawdown equals 6,287 BTC or 14 days of issuance.

In a moderate case, miners fund AI capex by selling 5% to 10% of their holdings. For MARA, that’s 2,700 BTC to 5,400 BTC, or 6 to 12 days of issuance. This is equivalent to $180 million to $361 million.

A 25% collective drawdown releases 29,174 BTC, or 65 days of issuance.

In the aggressive scenario, a 50% drawdown would put 58,349 BTC into markets, equivalent to 130 days of new supply. The risk is narrative, not volume.

Bitcoin’s 24-hour volume exceeds $50 billion, but when multiple miners become known sellers during macro stress, impact runs through sentiment and derivatives positioning rather than spot prints.

MARA’s filing permits others to follow without appearing distressed.

What the shift reveals

On top of the three scenarios, three competing narratives emerge.

The first is the AI pivot: miners repurpose power infrastructure into data centers, using Bitcoin as fuel to fund them.

MARA’s Starwood partnership targets AI-capable infrastructure with toggle economics. This is a strategic reallocation, consisting of power certainty to capacity certainty.

The second narrative is the tactical risk management: after $422.2 million in fair-value declines and $69.1 million in trading losses, MARA treats Bitcoin as a managed position.

Thin depth and macro sensitivity increase the value of discretionary liquidity tools.

The last narrative is a structural regime shift: the end of miner HODL. The contrast between 2024’s “retain all BTC” and 2026’s “may buy or sell from time to time” signals that miners behave like capital allocators, optimizing returns across mining, grid services, and AI leases.

Each narrative carries different supply implications.

If an AI pivot happens, the BTC sales fund transitions. In this case, supply pressure is front-loaded but finite.

In case the risk management narrative is the one moving forward, sales track volatility, making miners countercyclical sellers.

Lastly, a regime shift would mean that the approximately 117,000 BTC miner treasury becomes subject to active management, changing baseline assumptions about supply absorption.

The clock that matters

The next clarity window is MARA’s 10-Q form for the first quarter, projected mid-May.

Investors will scrutinize how much BTC was monetized post-policy change, whether AI milestones tie to treasury drawdowns, and what guidance on minimum reserves or sell cadence is provided.

The gap until May creates a narrative vacuum that macro conditions will fill.

Bitcoin trades in risk-off shape, driven by energy shocks and inflation fears, exactly when “who might be forced to sell” dominates.

MARA’s filing doesn’t say it will sell a majority. Still, authorization alone creates price-sensitive reference when liquidity is thin enough that the execution method determines whether a $1 billion sale is absorbed quietly or amplifies downside.

Starwood’s timeline adds urgency. The partnership targets 1 GW near-term, with a path to 2.5 GW, but “near-term” is undefined.

If MARA accelerates construction to capture AI demand, funding needs compress. If slower buildouts, BTC sales may stretch over years. That determines whether MARA’s treasury becomes a multi-year drag or a one-time recapitalization.

If the first-quarter earnings reveal multiple miners expanding sale authorizations or linking BTC monetization to AI capex, markets will reprice the entire miner treasury base as supply overhang rather than strategic reserve.

That repricing doesn’t require actual selling, it just means investors stop treating miner holdings as locked supply.

What’s actually at stake

MARA’s shift matters less for what it permits than what it signals.

For four years, miners positioned treasuries as differentiators by aligning equity performance with BTC appreciation. That worked when Bitcoin rallied, capital was cheap, and post-halving economics were theoretical.

Now Bitcoin trades nearly 50% off highs, capital markets favor AI over crypto, and post-halving margins are tighter than modeled.

If MARA executes AI pivots successfully and uses BTC sales as one-time funding, the treasury drawdown story ends cleanly. If AI projects drag on or Bitcoin recovers faster than expected, miners may have sold reserves at cyclical lows to fund underperforming projects.

For crypto markets, stakes are clear.

Miner treasuries were among the last bastions of non-speculative Bitcoin demand, representing entities that accumulated Bitcoin for operational purposes.

If that cohort shifts to active management, Bitcoin loses a structural bid and gains a structural seller. When the world’s largest Bitcoin miner by holdings formalizes its ability to sell its entire stack, it’s a signal that even believers are hedging.