Bit Digital reported cloud services revenue of $23.806m for the quarter to 30 June 2026 and its earnings release called that a 42 per cent sequential increase driven by “new contracts entering service”. The release does not mention that $12.3m of the figure is a termination fee from a customer that left. Strip it out and cloud revenue was $11.506m against $16.767m in the prior quarter: down a third, not up two fifths.
The fee came from what the filings call the Initial Customer. Its agreement was unwound, $2.2m was written off as bad debt, and a fixed termination fee of $12.3m was recognised as revenue, the 10-Q noting the quarter “was substantially offset by $12.3 million of termination fee revenue recorded”. After the quarter ended the fee was renegotiated upward to $15.7m, still unpaid when the filing went in. So 51.7 per cent of the cloud segment and 38.3 per cent of total revenue of $32.113m is money from a departing customer. Three customer orders were discontinued in the half, costing $5.1m between them.
Note what is not being said. There is no going-concern language, no material weakness, and the auditor's opinion on the 2025 accounts and internal controls is unqualified. And the subsidiary's September quarter, filed this morning, puts cloud revenue at $18.106m with no termination fee in it — a genuine recovery on the $11.506m. The company is not falling over. It is describing itself generously.
Mining is finished here
Mining was 7.4 per cent of the quarter's revenue and is winding down. The fleet shows why: 21,354 machines with 2.8 EH/s of nameplate capacity, of which roughly 0.4 EH/s was active — about 14 per cent utilisation. On cash it still works: mining cost of revenue excluding depreciation was $1.761m against 32.3 bitcoin received, so $54,520 a coin against $73,406 realised. The cash cost itself is down 39 per cent on a year earlier as expensive hosts were exited, which is a genuine improvement.
Fully loaded it does not. The company allocates neither depreciation nor overheads by segment, so any such figure is an estimate: weighting both by mining's share of revenue gives $129,167 a coin, and $231,972 if the fleet is still depreciating. Either way it is far above $73,406, and the gap is stranded fixed cost.
Three convertible issues totalling $690m of face value, and not one capped call bought against any of them.
The treasury, the leverage and the vote
The headline “approximately 164,310.5 ETH” is a composite of three balance-sheet lines on three accounting bases. Only 75,757.5 ETH is carried as a fair-value digital asset, at $118.899m against a cost basis of $267.526m — 55.6 per cent underwater. A further $105.616m sits as a collateral receivable, where the lender may “sell, pledge, rehypothecate, assign, invest, use, commingle or otherwise dispose of” the tokens; 49,000 LsETH is pledged against a $50m loan recallable on seven days' notice.
Debt carrying value reached $440.055m in June against $110.291m of convertible notes and no term debt at all six months earlier, and $690m of convertible face is outstanding once August's issue is counted. We searched all three issues for capped calls and found none; instead there is a $120m zero-strike call option at the subsidiary, which hands settlement timing to the counterparty and is not a dilution hedge. One $20m facility carries an effective interest rate of 50.6 per cent. Dilution depends on the basis: cover-page shares up 12.20 per cent year on year, period-end 24.91, weighted-average basic 69.18. And one million preference shares, carried at $9.05m, hold fifty votes each.
What is signed is substantial: $1.008bn of remaining performance obligations, 92.5 per cent of it colocation, anchored by a ten-year agreement worth roughly $865m. But 51.5 per cent falls after 2030, cloud obligations are $75.146m and run dry after 2028, and the $160m Paris contract has only a binding term sheet for its funding, with “no definitive agreements”. The “$200m of annualised” cloud revenue has no filed counterpart.
| Total revenue, quarter | $32.113m, up 25.2% year on year |
| Cloud revenue, and the fee inside it | $23.806m, of which $12.3m termination fee |
| Net loss attributable | $(107.212)m, $(0.31) a share |
| Adjusted EBITDA | $(63.564)m, token marks not added back |
| Active vs nameplate hash rate | 0.4 of 2.8 EH/s |
| Cost to mine, cash and loaded | $54,520; $129,167–$231,972 a coin |
| ETH fair value against cost | $118.899m against $267.526m |
| Collateral receivable | $105.616m, rehypothecation permitted |
| Convertible face, capped calls | $690m; none |
| Performance obligations | $1.008bn; 51.5% after 2030 |
All figures from Bit Digital's Form 10-Q for the quarter ended 30 June 2026, its 2025 Form 10-K, the furnished earnings release of 13 August 2026, and WhiteFiber's Form 10-Qs of 12 August and 10 November, retrieved from SEC EDGAR on 9 and 10 November 2026. Dollars throughout. Bit Digital itself had filed no 10-Q for the September quarter as at today, so its own figures here are four months old; the September-quarter figures are its subsidiary's, filed on the morning of 10 November 2026. Fully-loaded cost to mine is our allocation, the upper figure assuming three-year straight-line depreciation on the fleet's gross cost. Contested: the release reports gross profit of $18.6m where the segment note reports $18.294m and the income statement carries no gross profit line; the $46.0m LsETH impairment appears in the income statement's three-month column while the note reports nil for the quarter and $46.0m for the half, a contradiction internal to the 10-Q; and 31 December 2025 working capital is $500.8m in the 10-K and $85.1m in the 10-Q, the difference being the current classification of $415.7m of digital assets. Not a recommendation and not a forecast.