Skip to content
Data provided by CoinGecko
Crypto News

MARA Q2: more hashrate, smaller Bitcoin stack

Mira Chen Mira Chen 6 min read

Comic illustration of a mining hall converting toward glowing compute racks with blank coin stacks and a blank ledger

MARA Holdings just printed a quarter that grows the mining machine and shrinks the Bitcoin treasury at the same time. In its Q2 2026 shareholder letter filed with the SEC on August 6, 2026, the company said energized hashrate rose 22% year over year to 70.3 EH/s, while Bitcoin holdings fell 29% to 35,577 BTC (about $2.1 billion at the letter’s June 30 spot of $58,524). It mined 2,422 BTC and sold 2,213 BTC. That is not a “stack forever” print.

Bitcoin itself is still stuck in a quiet band. Our CoinGecko market snapshot fetched at 2026-08-07T00:00:43+00:00 showed Bitcoin near $64,269, down about 0.5% over 24 hours. Coinbase BTC-USD top-of-book around 2026-08-07T00:09:23Z sat near $64,339 with a thin best bid of about 0.036 BTC. The tape is calm. The miner balance sheets are not.

CleanSpark’s same-day fiscal Q3 2026 earnings release rhymes: mining revenue soft, power and AI infrastructure loud. The Block’s wrap flagged both names for double-digit revenue drops beside the AI pivot talk. The primary documents are what matter.

What happened

MARA’s letter is a classic Item 2.02 earnings package: operations still large, economics harder, treasury smaller.

  • Revenue fell 27% to $174.9 million in Q2 2026 from $238.5 million in Q2 2025.
  • Net income swung to a $611.3 million loss from $808.2 million of net income a year earlier.
  • Adjusted EBITDA fell to a $360.9 million loss from $1.2 billion.
  • Energized hashrate rose to 70.3 EH/s from 57.4 EH/s year over year (down 3% from 72.2 EH/s in Q1 2026).
  • Blocks won rose 1% to 700; BTC produced rose 8% quarter over quarter to 2,422.
  • Purchased energy cost per BTC at owned sites rose to $38,690 from $33,735 in Q2 2025, with owned-site power at $0.04/kWh.
  • Holdings: 35,577 BTC, including 9,270 BTC loaned or pledged as collateral. No BTC purchased in the quarter. Cash and cash equivalents: $421.3 million. Combined unrestricted cash and BTC (including loaned/pledged): about $2.5 billion.
  • Digital-asset management: 4,742 BTC loaned, generating about $4.3 million of interest income in the quarter.

CleanSpark’s release lands on the same day with a different fiscal calendar (quarter ended June 30, 2026):

  • Revenue $138.0 million, down 30.5% year over year from $198.6 million.
  • Net loss $239.8 million, or $0.89 per basic share.
  • Adjusted EBITDA loss $113.0 million.
  • Bitcoin HODL value $814.9 million (current, non-current, and counterparties for collateral).
  • Headline infrastructure print: a signed 20-year, $6.6 billion triple-net lease at Sandersville with a high investment-grade tenant; equity portion described as fully funded and long-lead items prepaid.

Two miners. Same message in different accents: the ASIC floor still runs, the income statement hurts, and management wants the market to price power and compute optionality, not another pure HODL victory lap.

Context

This fits the 2026 miner pattern we have been watching across the desk. Spot Bitcoin is not melting up. Our snapshot still shows a mid-$64k handle, far from the kind of price that papered over thin margins in earlier cycles. When energy cost per coin rises and network difficulty outruns hashrate growth, as MARA notes, “grow EH/s” stops automatically growing the stack.

MARA sold almost as many coins as it produced. That is the line traders should not skim. A miner can raise hashrate, win slightly more blocks, and still choose liquidity over accumulation. Part of the remaining stack is already working as loaned or pledged collateral. That is closer to a balance-sheet tool than to a religious cold-storage vault. For readers who want the base asset timeline, our Bitcoin history page is the hub; for how thin books behave when headlines hit, see market depth and counterparty risk.

CleanSpark makes the pivot less abstract by attaching a long-dated lease number. A 20-year triple-net lease is not the same thing as shipping AI revenue today, but it is a sharper primary signal than slide-deck megawatts. MARA’s letter still talks in campus-conversion language: mining sites becoming AI/HPC campuses, every megawatt pushed to its highest-value use. Fair. Until more named contracts and dollars show up beside the EH/s table, the equity story and the Bitcoin treasury story are diverging.

We have already watched corporate Bitcoin sellers this month when Strategy funded dividends and preferreds with BTC (see our earlier Strategy sale reaction). Miners selling production is a different mechanism, but it rhymes: Bitcoin on the corporate balance sheet is being treated as working capital again, not only as a monument.

Sector tape context stays soft. Coinbase BTC-USD 24h range on the venue stats print we pulled sat roughly $64,087–$64,944 with last near $64,344. That is a narrow day. Miner filings are doing the loud work.

Our read

My stance: MARA’s Q2 confirms the Bitcoin-treasury miner is giving way to the power-and-compute miner. Hashrate up, holdings down, nearly one-for-one sell-through of coins mined, and a letter that spends its emotional energy on AI infrastructure scarcity. That is a capital-allocation tell, not a vibes tell.

I am not saying Bitcoin mining is over for MARA. Blocks won and BTC produced still rose on a quarterly basis. Cost per petahash per day even fell 4% year over year. The operating machine works. What changed is the preference revealed by the treasury line: liquidity and collateral utility beat another 2,000+ BTC added to the pile.

CleanSpark’s Sandersville lease is the corroborating sector signal. When two large public miners print soft mining revenue on the same day and both lean into long-duration power monetization, the market should stop treating “more EH/s” as a synonym for “more BTC forever.”

Falsifiable claim: By 2026-11-15 UTC, in MARA’s next quarterly shareholder letter (or accompanying 8-K exhibits), either (1) disclosed Bitcoin holdings are at or above 35,577 BTC after a quarter that includes open-market BTC purchases, or (2) MARA discloses a signed multi-year AI/HPC customer contract covering at least 100 MW with economics or a named counterparty in the primary filing. If holdings keep falling without purchases and no ≥100 MW contracted AI/HPC print appears, this Q2 read stands: treasury accumulation is no longer the lead product.

What to watch next

  1. Next MARA holdings line. Did they sell through production again, or did buys return? Watch loaned/pledged BTC as a share of the stack.
  2. Sandersville follow-through. CleanSpark already claims equity funded and long-lead gear prepaid. Look for ready-for-service timing and whether the unnamed “high investment-grade tenant” ever becomes nameable in a filing.
  3. Energy cost per BTC vs spot. MARA’s owned-site energy cost of $38,690 only works if Bitcoin holds a healthy premium to that all-in power bill. Our snapshot mid-$64k handle still clears it; a deeper range break would not.
  4. Equity narrative vs BTC flow. If miner stocks trade the AI lease headlines while spot Bitcoin stays range-bound, treat that as a sector multiple story, not a Bitcoin breakout story. For reading those cross-currents cleanly, keep our market snapshot guide and the Bitcoin news category nearby.

Bottom line for the desk: the August 6 miner filings are not a bullish Bitcoin treasury surprise. They are a balance-sheet confession that power is the scarce asset managers want priced, and Bitcoin is again a coin they are willing to sell when the quarter demands cash.