MARA vs Bitcoin and Why the Miner Is Not Leveraged Bitcoin

From 9 November 2021 to the Friday 28 August 2026 close, the bitcoin network's hashrate expanded from roughly 162 EH/s to 914 EH/s, a 5.6x build-out in four and a half years. Bitcoin gained 16.2% over that same stretch. Hashprice, which is the revenue a miner earns per unit of computing power it runs, fell 88.6%. MARA Holdings fell 86.0%.

Three of those four numbers move together, and the coin is not one of them. The miner tracked the hashprice, and that single alignment answers the question retail keeps asking about owning a bitcoin miner as a geared way to own bitcoin. MARA is a different instrument from the asset it mines, and the gap has widened far enough that through 2026 the two have been moving in opposite directions.

Why the Leverage Thesis Worked in 2023 and Broke Afterward

Give the thesis its due, because for one full year it was correct and it was correct by a lot. Through calendar 2023 MARA returned 591% against bitcoin's 156%, roughly 3.8x the coin, and anyone who bought the miner instead of the asset was rewarded for it. That year is where the "miners are leveraged bitcoin" idea was born, and the pages still arguing it in search results were written in 2023 and early 2025, before the rest of the data existed.

The rest of the data is re-derived below from dated session closes, because every third-party year-to-date figure checked against those closes came back wrong, in both directions.

A leveraged instrument loses when the underlying loses and wins when it wins. MARA lost 28.61% in the single best bitcoin year of the cycle, then took seven times bitcoin's loss on a coin that finished 2025 nearly flat. And in 2026 the relationship inverted the other way, with the miner up while the coin is down. That is not leverage. That is a different business whose price happens to rhyme with bitcoin some of the time.

What a Miner's Stock Actually Tracks

Think of hashprice as the wholesale rate a miner is paid for the only product it makes. Bitcoin's price is one input into that rate. Network difficulty is the other, and difficulty moves against the miner every time a competitor plugs in a machine. Installed capacity reached 914 EH/s on Blockchain.com's network hashrate chart, and every one of those exahashes is competing for the same block reward. The four-year halving cuts the block subsidy on a schedule that has nothing to do with price, and the rest of the compression comes from everybody else's capital expenditure.

MARA's own quarter shows the treadmill better than any chart. In Q2 2026, ended 30 June, energised hashrate rose 22% to 70.3 EH/s across a fleet of roughly 440,000 rigs, and revenue fell 27% year over year to $174.9 million. Energy as a share of revenue from owned mining went from 34.1% to 54.9% in twelve months.

Cost per petahash per day came in at $27.7 against a Q2 hashprice near $35.66. The most recent published reading on Luxor's hashprice index is $38.29, dated 22 August 2026, which leaves a margin thin enough that a single difficulty adjustment can eat most of it.

Production tells the same story from the other side, and the two bases get confused constantly. MARA produced 2,422 BTC company-wide in Q2 2026, up 3% year over year, but only 1,260 of those came from facilities it owns. The rest arrived through joint ventures and hosting arrangements where MARA does not keep the full economics. More machines, more hashrate, more coins, less money.

MARA Has Been a Net Seller of Bitcoin

This is the part that breaks the "buy the miner to get more bitcoin" argument outright. Across the first half of 2026, according to the 10-Q filings on the SEC's EDGAR system, MARA sold 23,093 BTC for $1.6 billion at an average of $70,631, including 15,133 BTC for roughly $1.1 billion in March alone to fund a convertible note repurchase. The HODL policy adopted on 25 July 2024 was relaxed during 2025 and abandoned in 2026, in the filing's own language.

Holdings finished the June quarter at 35,577 BTC, down 34% in six months from a 31 December 2025 peak of 53,822, with 9,270 coins loaned out or pledged as collateral. Bitcoin per share fell from 0.0001418 to 0.0000921 on the 28 August share count. A shareholder who sat still through that period ended up with a smaller claim on a smaller pile.

Share count is the older half of that problem and it deserves an accurate telling rather than a frightening one. MARA went from 174,266,313 shares on 4 August 2023 to 386,299,297 on 30 July 2026, a 121.7% increase in three years that damaged the per-share case regardless of what bitcoin did. But issuance has close to stopped, at 2.1% since October 2025 and zero at-the-market shares sold in the first half of 2026. The dilution is historical and the tap has been shut for ten months, which is the same tension that runs through how mNAV works for treasury stocks and through every equity that markets itself as a claim on a coin, including the Phemex profile of Strategy's 2026 stock.

One more number needs its context or it misleads. MARA reported a Q2 2026 net loss of $$611.3 million and EPS of$$(1.60), and most of that is mark-to-market rather than operations. The filing states that the fair value of its bitcoin fell $343.0 million in the quarter and $1.4 billion across the half. That is an accounting consequence of carrying bitcoinon the balance sheet through a drawdown, not evidence that the mining business collapsed.

MARA Is Not a Pure Bitcoin Miner Anymore

Anyone valuing MARA as a mining pure-play in 2026 is valuing a company that stopped existing. Three dated transactions, all sitting in MARA's own press release archive, moved it into AI and high-performance computing. It bought a controlling stake in Exaion for $174.5 million cash on 20 February 2026, formed a venture with Starwood Digital Ventures on 26 February targeting more than a gigawatt near term, and agreed to acquire Long Ridge Energy for roughly $1.5 billion on 29 April, a 485 MW gas plant in Ohio that had not closed as of 30 June 2026 and is awaiting FERC approval. A 2 GW site in Matagorda County, Texas followed after quarter end, against a stated target of about 4.8 GW.

The 10-Q books a 2026 Restructuring Plan costing $47.6 million in the first half, explicitly to "reallocate resources toward AI initiatives." CEO Fred Thiel put the logic plainly in the 6 August 2026 results, saying the company does not see bitcoin mining and AI infrastructure as competing businesses but as "different applications of the same underlying asset", and naming that asset as power. As a description of the business that is accurate, and it is also an admission that the stock is partly a power and datacenter equity riding the same demand that repriced the AI chip names in the Phemex comparison of Marvell and Broadcom.

The problem is not the pivot itself. Almost every large listed miner has made some version of it, and several announced theirs earlier. The problem is that MARA pivoted late and without an anchor tenant, while peers signed named, revenue-producing leases first, which is the model behind the Nasdaq-100 entry covered in the Phemex writeup on CoreWeave. MARA's AI capacity does not deliver until mid-2028. A buyer in 2026 is paying for gigawatts that produce nothing for roughly two years and have no signed customer attached to them.

The Volatility Numbers and What That Beta Actually Measures

Our own calculation across 667 paired daily sessions, computed from raw closes rather than lifted from any data provider, puts the relationship in numbers.

Read the first three rows together and the trade looks bad on its own terms. You are taking almost double the volatility for barely more directional exposure to bitcoin, and roughly 40% of MARA's movement is being driven by something that is not bitcoin at all. An independent study published on 29 August 2026, using rolling 90-day betas across seven listed miners, reached a compatible conclusion and found something more damaging for the sector than for MARA alone. Correlation to the Nasdaq-100 exceeded correlation to bitcoin for every one of the seven.

One trap is worth naming, because secondary write-ups keep falling into it. The "Beta 5.36" figure that circulates on MARA quote pages is a five-year monthly beta against the S&P 500. It is not a bitcoin beta, it never was, and reading it as one overstates the coin relationship by more than four times.

A second trap ran on the Friday itself. MARA closed at $10.67 against a Thursday 27 August close of $11.87 on 41.8 million shares, inside a 52-week range of $6.66 to $23.45, per stockanalysis.com's dated session history for MARA. Two widely syndicated recaps that day put the move at 6% and the price at $11.12. Both were intraday quotes captured before the bell, and the closing figures in the table above are the ones that survive.

Frequently Asked Questions

Is buying a bitcoin miner the same as buying leveraged bitcoin?

No, and the practical difference is that miner leverage is undefined and resets against you. A perpetual position gives you a leverage ratio you chose and a liquidation price you can see, while a miner's effective gearing changes every time network difficulty rises, a competitor energises a fleet, or management sells treasury coins to service debt.

Which is riskier to trade, MARA or bitcoin?

MARA, on almost every measure a trader cares about, because it carries roughly double bitcoin's volatility plus single-issuer risk that the coin does not have. It also stops trading at the closing bell and gaps over weekends and holidays, so a bitcoin move on a Saturday reaches the stock only as a Monday gap you had no way to manage.

Where can you check hashprice yourself?

Luxor's Hashrate Index publishes it daily in dollars per petahash per day, and it is the single most useful number for anyone holding a miner. Compare it against the company's disclosed cost per petahash per day from the latest quarterly filing, because the spread between those two figures tells you more about the stock than bitcoin's chart does.

Do all bitcoin miners trade like MARA?

Directionally yes, though MARA sits at the extreme end. It was the highest-beta of the seven miners in the 29 August study at 1.10 with a 0.48 bitcoin correlation, and what separates the group in 2026 is power portfolio and signed tenant contracts rather than hashrate.

Bottom Line

MARA is a power and datacenter business with a bitcoin mining division attached, and pricing it as a bitcoin proxy has cost people money in three of the last four calendar years. Three things decide the stock from here and none of them is bitcoin's price. Hashprice against that $27.7 per petahash per day production cost, the FERC decision that closes Long Ridge, and the first named tenant for AI capacity that produces nothing until mid-2028. If bitcoin exposure is what you actually want, bitcoin closed Sunday 30 August at $77,697 and you can hold it directly at a leverage ratio you set. Buying the miner instead gets you a leverage ratio nobody publishes, recalculated against you every time somebody else plugs in a machine.

This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency trading involves substantial risk. Always conduct your own research before making trading decisions.