American Bitcoin Booked $126M in Q3 Without Buying a Single Sat of Spot BTC

$126 million in net income. $221 million in adjusted EBITDA. And not one satoshi bought on the spot market.

American Bitcoin's November 4 release put three numbers in the same paragraph that don't normally share space. The company grew its Bitcoin reserve last quarter entirely through its own rigs. No ATM share sales. No treasury buys. Just machines, power contracts, and hash rate.

That's a harder trick than it sounds in 2026.

Most public miners are doing the opposite right now. They're selling coins to fund AI data center buildouts because compute contracts pay in dollars and dollars pay the electricity bill. American Bitcoin went the other direction and stacked.

The Quarter, Line By Line

Here's what the release actually says. Roughly $126 million in net income for the quarter. About $221 million in adjusted EBITDA. And roughly 5% growth in satoshis per share.

That last one is the number you should care about, and almost nobody quotes it in the headline.

Bitcoin per share isn't a GAAP concept. It's a ratio, and you can compute it yourself in three lines. Here's the relevant code.

def sats_per_share(btc_held, diluted_shares): return (btc_held * 100_000_000) / diluted_shares

Run that against the last four quarters of any public miner. If the line slopes up, shareholders are getting more Bitcoin per unit of ownership. If it's flat while the total treasury grows, the company is running on a treadmill. It's buying coins with freshly printed shares and calling the result growth.

American Bitcoin says it didn't touch its at-the-market program during the period. No issuance. So the 5% gain in sats per share came from production rather than a capital markets maneuver. That's the clean version of the story, and management deserves the credit for it.

But one quarter of discipline isn't a policy. Ask again in four quarters.

Worth noting too that skipping spot buys also means skipping the premium. Companies that raise equity at 1.8x net asset value and immediately buy coins are printing sats per share out of thin air. It's an accretive trade right up until the premium collapses and the ATM window shuts. Sitting out that game for a quarter is either restraint or a missed opportunity, depending on where you think the stock trades relative to its reserve.

Where the Other $95 Million Lives

Now the part the press release buries.

Net income of $126 million against $221 million of adjusted EBITDA leaves roughly $95 million of the operating picture sitting outside the GAAP line. That gap isn't automatically a red flag. It's just the place where definitions live.

Some of it's ordinary. Depreciation on an ASIC fleet is real money and it's non-cash, so adding it back is defensible. Interest on equipment financing is also real, and also added back. But "adjusted" is a word that carries a lot of freight in mining. Every operator defines it differently. Two companies can post the same EBITDA figure and mean entirely different things, which makes cross-company comparisons nearly useless unless you align the definitions first.

Read the reconciliation table. Then read it again. If the adjustments are depreciation and interest, fine. If they include impairment of digital assets or mark-to-market on derivatives, you're looking at a number that tracks the Bitcoin price rather than the business.

And that's the trap with any treasury company. A single strong quarter is a snapshot of a price. It isn't a run rate.

Management also points to producing Bitcoin at a structural discount to market. Plausible, when power contracts are cheap and the fleet is current. But difficulty climbs. Old machines drop out of profitability without warning, and seasonal electricity costs hit every summer. The discount isn't structural. It's contractual, and contracts expire.

So ask the uncomfortable question. What's the all-in cost per coin? And does that figure include site overhead, staff, and the depreciation they just added back to EBITDA? Most miners quote cash cost and quietly skip the rest. Does the 5% hold if difficulty jumps 8% in a single quarter? Nobody in that release answers it.

The Bet Against the AI Pivot

Here's the strategic picture, and it's the most interesting thing in the document.

Plenty of miners spent 2025 and 2026 converting racks to GPUs and signing dollar-denominated compute deals. Those contracts are boring. They're also predictable, which is exactly what a public equity investor wants. The trade-off is that you stop being a Bitcoin company and become a landlord with a power bill.

American Bitcoin went the other way. It's accumulating coins and refusing to sell them to fund anything. That's a directional bet on BTC price with an operating business bolted on top.

Which is the better business? Depends entirely on what you think Bitcoin does over the next 24 months. If you're bullish, the accumulator wins because every mined coin is upside. If you're neutral, the AI landlord wins because the cash flow doesn't care about the halving.

Either way, understand what you own. Public shareholders hold a claim on a company, not a claim on specific coins in a reserve. The company can pledge those coins. It can sell them. It can issue shares against them tomorrow morning. Sats per share is your only real defense against that, and it only works if you track it every single quarter.

What to Check Next Quarter

Six numbers. Mined BTC. Treasury change. Cost per coin. Network difficulty. Diluted share count. Cash on hand.

If mined BTC grows and sats per share grows with it, the model works. If the treasury climbs while sats per share stalls, you're watching dilution dressed up as accumulation. That's the whole test. You don't need an indexer for it. A spreadsheet and two earnings releases will do.

The $126 million is real. The $221 million is a definition. Know which one you're pricing.