Canaan Pledged 63% of Its Bitcoin to Borrow. The Ledger Shows What the 'Treasury' Story Hides

On Sept. 8, Canaan filed a quarterly disclosure that contained one number worth more than the rest of the document combined. Of the 1,915.5 BTC the company claimed at June 30, 1,117 BTC sat pledged against secured term loans, and another 100 BTC had been parked in a fixed-term product. That is 63.5% of the headline balance. Not held. Not free. Encumbered. The same filing reported second-quarter revenue of $31.9 million, against guidance of $35 million to $45 million issued in May. Product revenue — the machines that are supposed to be the business — collapsed to $13.6 million from $42.9 million in the first quarter. Third-quarter guidance landed at $11 million to $15 million. That range is not a correction. It is a different company. And the disclosure tells you, quietly, what that different company actually owns. We do not build in the dark; we audit the light. So let's audit. Canaan is a Bitcoin mining-equipment manufacturer. That sentence used to describe a business with gross margins, inventory cycles, and a customer base of miners who bought ASICs on the assumption that network hashprice would cover the amortization schedule. That assumption has been losing ground for two years, and the equipment business has been repricing around it in real time. When machine demand weakens, a hardware vendor has three levers: cut prices, sell more compute to itself, or lean on the balance sheet. Canaan has pulled all three in sequence. Through 2024 and 2025 it built a self-mining operation. By mid-2026 it was producing coins rather than only selling the machines that produce them. That pivot converts an equipment maker into a leveraged commodity producer. The asset base gets re-denominated. Inventory becomes rigs deployed in Ethiopia and elsewhere. Accounts receivable become mining revenue. And cash — the thing that pays salaries, suppliers, and creditors — gets progressively replaced by an asset whose price is set in a 24-hour market over which management has no influence whatsoever. Treasury-company framing came next, and in a bull market that framing is rewarded. Equity holders in firms holding Bitcoin have historically been valued on a per-coin basis, sometimes at a premium to net asset value, on the theory that the corporate wrapper supplies leverage, tax efficiency, and access. The wrapper is real. The leverage is real. What tends to go missing from the marketing is the encumbrance schedule. The June 30 balance sheet separates two entries that deserve to be read as unrelated assets. The first is cryptocurrency receivables: $70.9 million. That figure comprises the 1,117 BTC pledged as collateral for secured term loans plus the 100 BTC in the fixed-term product. Both are contractual claims with conditions attached, not discretionary holdings. The pledged coins cannot be sold to fund operations without triggering repayment obligations or margin consequences. The fixed-term product carries a maturity date. Neither behaves like a reserve. The second is cryptocurrency assets: $47 million, corresponding to 698.5 BTC. The distinction matters because the "1,915.5 BTC treasury" headline sums two things a CFO would never treat as one thing. Collateral is a liability-management instrument. A treasury is a reserve. Adding them produces a number that is arithmetically correct and analytically useless. Cash told a cleaner story. Canaan reported $66 million on hand at June 30, up from $43.5 million at March 31 — a sequential improvement of $22.5 million — yet still below the $80.8 million held at the end of 2025. Against a $97.6 million quarterly net loss, that cash build looks anomalous until you decompose the loss. The $97.6 million included $25.3 million in inventory and prepayment write-downs and purchase-commitment provisions, plus $9.2 million in property and equipment impairment. Those are noncash charges. They describe machines and prepaid silicon now worth less than the price paid, not dollars that left the building. Based on my audit experience with hardware vendors through the 2018 and 2022 drawdowns, this is the precise accounting signature of a manufacturer caught with the wrong inventory at the wrong point in the cycle: the write-down is the moment the market's new price is admitted into the ledger. Mining softened the damage. Canaan produced 243 BTC and booked $17.7 million of mining revenue in the quarter. Management stated the segment contributed positively before depreciation. Read that as the narrow gauge it is. "Positive before depreciation" excludes the cost of the rigs themselves, which is the dominant capital cost in any mining operation. It answers whether electricity and hosting are covered. It does not answer whether the deployed capital earns its cost of capital. There is also a physical constraint. Canaan has disclosed that paused Ethiopia capacity represented nearly 35% of its July operating hashrate. Hashrate that is not running produces no coins and burns the depreciation clock regardless. A 35% idle share is not a rounding error; it is the difference between self-mining as a hedge and self-mining as a stranded asset. Then came the post-quarter transactions. In late August, Canaan sold 3,952 ETH and 54 BTC for approximately $13.9 million and used part of the proceeds for share repurchases. By Sept. 8, buybacks under the existing program totaled roughly 16.4 million American depositary shares for $7.4 million, including $5.4 million spent in late August alone. Read that sequence without the press-release adjectives. The company sold crypto assets — the liquid portion, not the pledged collateral — to buy back its own equity. In a bull market that is defensible: management is signaling that the shares trade below intrinsic value. But the funding source is not operating cash flow. It is the liquidation of a reserve that had been accumulating for eighteen months, and the June treasury snapshot predates the sales entirely. Codifying the intangible — how hash becomes a balance-sheet line — is the whole game here. Machines are inventory. Deployed machines are property and equipment. Coins produced are either revenue or assets. Coins pledged are receivables. Each reclassification changes how identical underlying economics appear to a screen, and each one is a choice. The consensus reading of this quarter is that Canaan is a struggling hardware company sitting on a Bitcoin cushion. I would invert it. Canaan is a Bitcoin-adjacent financing operation that happens to manufacture rigs, and the rig business is now the division that consumes capital rather than generating it. The bear case everyone recites is "hardware demand is cyclical." True, and uninteresting. The case worth pricing is structural. The pledged BTC is not a cushion, because it is already promised to lenders. The buyback is not a signal of undervaluation, because it is funded by selling the reserve that would otherwise absorb a bad third quarter. And guidance of $11 million to $15 million implies a product revenue line that leaves mining carrying the load on a base with more than a third of capacity paused. Here is the sharper point. In bull markets, crypto-equity investors price treasuries at the top of the stack and liabilities at the bottom. The ledger remembers what the narrative forgets: 1,217 of 1,915.5 coins were spoken for on June 30. Any thesis built on "BTC per share" that fails to net pledged collateral is not conservative. It is unread. The datapoint that matters next is not the third-quarter revenue print. It is whether the pledged 1,117 BTC gets rolled, reduced, or called when the secured term loans mature — and whether Canaan funds operations or further buybacks from the remaining 698.5 coins. Watch the receivables line, not the treasury headline. When a hardware company starts calling collateral a treasury, the balance sheet is the only honest quarter of the story.