At 5:00 p.m. Eastern tonight, MARA Holdings will walk investors through its second-quarter results, and a familiar ritual will follow: headlines will declare the quarter a triumph or a disaster, the stock will gap in one direction, and most holders will be none the wiser about what actually happened inside the business. Miner earnings are the most misread documents in crypto — part industrial filing, part leveraged Bitcoin bet, part AI land-grab pitch — and this week delivers a full season of them. This guide, the nineteenth in our literacy series and a companion to the treasury-company earnings guide, gives you the five numbers that matter, the two accounting traps that generate fake headlines, and a worked example you can apply to tonight’s print.
Why miner earnings are structurally weird
A Bitcoin miner is three businesses stapled together. It is an industrial operator that converts electricity into hashes, where the metrics are cost per coin and uptime. It is a leveraged Bitcoin fund, because most miners now hold coins they mine (and mark them to market), so the BTC price on the last day of the quarter can swing reported earnings more than anything management did. And increasingly it is a data-center developer, renting power and buildings to AI tenants — the pivot that produced deals like Hut 8’s $9.8 billion, multi-year AI lease announced in July, which moved the whole sector, per 24/7 Wall St. Each business deserves its own P&L. The quarterly report smashes them into one number — which is why the number alone tells you almost nothing.
The five numbers that matter
| # | Number | Where it hides | What it tells you |
|---|---|---|---|
| 1 | BTC mined vs. network difficulty | Production update / MD&A | Operational execution, normalized for luck and the difficulty tax |
| 2 | Cost to mine one BTC (energy cost, then all-in) | MD&A, sometimes only derivable | The margin of survival if price falls; compare to spot, not hope |
| 3 | BTC held — and BTC sold | Balance sheet + cash-flow statement | Whether the miner is a believer, a forced seller, or quietly funding capex with coins |
| 4 | HPC/AI revenue — contracted vs. delivered | Segment note / press-release bullets | Whether the pivot is revenue or renderings; count only signed, dated dollars |
| 5 | Power position (MW contracted, cost per kWh) | Operations update | The real moat — in 2026, energy contracts are worth more than mining rigs |
Number 1 needs the denominator. Mining is a zero-sum tournament: if the network’s difficulty rises 10% and a miner’s production falls 8%, that miner gained relative share. This quarter’s reports arrive after a difficulty tape we have covered closely — including July’s unusually small 0.74% cut — so a flat coin count can be a good result. Number 3 is where narratives die: a miner that preaches conviction while its cash-flow statement shows every mined coin sold is telling you its real cost structure in the only language that cannot spin. And Number 4 has a hard rule: a lease is revenue when a tenant with a name signs for dollars with a date. “Exploring HPC opportunities across our portfolio” is not a segment; it is a slide.
The two accounting traps
Trap one: fair-value whiplash. Since the FASB rule change, miners mark their Bitcoin holdings to market through earnings every quarter. A miner holding 50,000 BTC swings its paper profit by $50 million for every $1,000 of quarter-end price movement — dwarfing operations. The June 30, 2026 quarter-end printed near $63–64K against a March 31 print near $80K, so every HODLing miner will report enormous fair-value losses this season regardless of how well it mined — the same mechanics that produced the loss headlines across the sector in May, per Blockhead’s review of the March quarter, when CleanSpark’s fiscal Q2 showed $136.4 million of revenue against a $378.3 million net loss dominated by marks, per StockTitan. Strip the mark, rebuild the operating result, then decide. Trap two: the calendar. Not all “Q2s” are the same quarter — CleanSpark’s fiscal “Q2 2026” ended March 31, while MARA’s calendar Q2 ended June 30. Comparing them as if they covered the same three months is a category error that even paid research commits.
The silent killer: share count
There is a sixth number, and it belongs to every miner conversation even though it appears in none of the headlines: diluted shares outstanding, this quarter versus four quarters ago. Mining is brutally capital-intensive — rigs depreciate in three to four years, difficulty compounds relentlessly, and AI build-outs are measured in billions — and since most miners cannot fund that from operating cash flow, they fund it from you. At-the-market equity programs let a miner sell new shares into every rally, which is why a company can grow its hashrate 40%, grow its Bitcoin stack 20%, and still leave a long-term holder worse off per share than a year earlier. The test is always per-share, never headline: hashrate per share, BTC held per share, and — for the AI hopefuls — contracted megawatts per share. A miner that grows those three while the share count crawls is compounding for its owners; one that grows them only alongside a ballooning denominator is a treadmill that converts shareholder dilution into industrial statistics. When tonight’s call turns to expansion plans, translate every announced exahash and megawatt into “funded by what?” — retained earnings, debt, coin sales, or new paper. The four answers carry four very different meanings for the person actually holding the stock, and management will volunteer none of them unprompted.
Worked example: tonight’s MARA print
Apply the framework to what Wall Street expects at 5:00 p.m. ET: $209.4 million in revenue and $0.17 in EPS, with analysts citing weaker Bitcoin assumptions and “mixed execution” on the HPC/AI transition, per Yahoo Finance’s preview. Before the call, write down four questions. One: BTC mined against the quarter’s difficulty path — and against the 1,527-coin consensus the sector prints as a benchmark for peers, per the Riot preview’s comparable math. Two: coins held versus sold — MARA historically HODLs; any change in that policy is the real headline, whatever the EPS line says (and after Strategy’s own sale this week, treasury policy is the sector’s live question). Three: AI dollars with dates — contracted megawatts, named counterparties, revenue recognized this quarter, not “pipeline.” Four: the fair-value mark — compute earnings ex-mark before reacting. If the stock gaps on a number the mark explains, the gap is noise; if it gaps on cost-per-coin or a HODL-policy change, it is signal.
Five rules for reading any miner quarter
Rule 1: Normalize by difficulty before judging production. Coin counts without the network denominator are advertising. Rule 2: Rebuild earnings without the fair-value mark. The mark is the Bitcoin price wearing a costume; you already know the Bitcoin price. Rule 3: Read the cash-flow statement before the press release. Coins sold, capex paid, and shares issued are the three numbers management least wants to discuss. Rule 4: Count only signed AI dollars. The pivot is real — Hut 8’s lease proves it — but reality arrives contract by contract, and the sector currently trades renderings at contract multiples. Rule 5: Treat delays as data. Earnings calls that move without explanation, like Riot’s this week, are usually logistics — but “usually” is a probability, not a promise, and the gap between the two is where risk lives.
FAQ
Why do all Bitcoin miners report huge losses in the same quarter?
Fair-value accounting marks their BTC holdings to the quarter-end price through earnings. When Bitcoin ends a quarter well below where it started, every holding miner reports large paper losses simultaneously — regardless of operational performance. Strip the mark to see the business.
What is a good cost to mine one Bitcoin in 2026?
Direct energy cost per coin varies widely with power contracts; the honest comparison is each miner’s all-in cost against spot. A miner whose all-in cost sits well below $64K spot has margin of survival; one above it is consuming its balance sheet each quarter.
Is the AI/HPC pivot real revenue?
For some miners, yes — signed multi-year leases like Hut 8’s $9.8B deal are contracted dollars. The rule: count named tenants, dated contracts, and revenue recognized this quarter. Treat everything else as pipeline marketing.
What should I watch in MARA’s report tonight?
Wall Street expects $209.4M revenue and $0.17 EPS. The four tells: BTC mined vs. difficulty, any change in HODL policy, AI revenue with dates, and the size of the fair-value mark inside the headline number.
Why did Riot delay its earnings call?
The company rescheduled its August 5 call and promised a new date in a subsequent release, without stating a reason. Delays are usually scheduling logistics — our Q3 marker tracks whether a new date lands within the week, which historically separates routine from trouble.
Investment disclaimer: This article is for informational and educational purposes only and does not constitute investment, financial, legal, or tax advice. Bitcoin and cryptocurrencies are volatile assets; you can lose some or all of your capital. Always do your own research and consult a licensed financial advisor before making investment decisions.