Canaan's $32 Million Quarter, an $11 Million Guide, and a Gigawatt Bet the Tape Doesn't Believe

The Bitcoin rig business is collapsing faster than management can pivot — so the story now rests entirely on power resources, a consumer heater, and management buying its own stock with its Bitcoin.

CAN · Earnings Call · 2026-09-08

A mining-machine business that is melting

Canaan's second-quarter print was ugly at the headline and uglier at the guide. Total revenue of roughly $32 million came in “below our previous guidance range of USD 35 million to USD 45 million” — Nangeng Zhang, Chairman and CEO · 2026-09-08, with only about $14 million of that from product sales. The company sold 2.5 exahash per second into a falling market, and mining revenue ($18 million, 243 Bitcoin) is now more than half the business. The cleanest explanation of the quarter is management's own: elevated industry inventory meeting a price war, with competitors discounting to accelerate cash collection. Then came the real tell — third-quarter revenue guidance of $11-15 million. A guide that implies roughly a 60% sequential decline is a statement about the state of the rig market, not about execution. Notably, the A16 series, which dominated the narrative just two quarters ago as the flagship ramp, has quietly fallen out of the top of this company's keyword stack — the company's own language has moved on.

The pivot deserves the attention, and the skepticism

With the rig business in free fall, the transformation becomes the whole equity story. Project ABC, the U.S. mining venture with WindHQ, is now the company's number-one keyword — installed hash rate reached 4.85 exahash per second at the end of July, up 10% from March, and Canaan pulled “$5.2 million in cash from Project ABC, including sales collections and cash distributions” — James Cheng, CFO · 2026-09-08 during Q2 (cumulative $8.4 million). Management is emphatic that it is optimizing for cash generation, not book value: it took roughly $4 million of equity-investment losses precisely because it retired old machines rather than protect their carrying value. The bigger swing is power resources. Canaan is trying to turn years of U.S. site-scouting into long-term, low-cost power, and it continues to wave the gigawatt scale flag. This is not new — the promise dates to at least February, when NG said “we are quite confident in the gigawatt-level power opportunities, which is based on our results from work at this stage” — Nangeng Zhang, Chairman and Chief Executive Officer · 2026-02-10 — and the opacity is deliberate and recurring. In May he explained “we sit down with our compliance advisers and agree that it's better to announce details after some important commercial and legal documents are formally signed” — Nangeng Zhang, Chairman and CEO · 2026-05-19. What is genuinely new this quarter is the framing that Canaan may participate in the AI/HPC buildout at all; the fresh keyword GPU cloud rental surfaced as analysts pushed on structure. NG's answer was refreshingly honest: “I think it's still too early to decide whether we will focus on co-location or GPU cloud rental.” — Nangeng Zhang, Chairman and CEO · 2026-09-08 The company is openly saying it will not build end-to-end.

first of all, we should try to secure gigawatt level power and with this on hand and then we can start to talk about the next step.
James Cheng, CFO · 2026-09-08

Capital allocation: Bitcoin in, shares out

The quarter's most interesting move was financial, not operational. Liquidity actually improved — “At the end of the second quarter, we had $66 million in cash, up about $23 million from $43 million” — James Cheng, CFO · 2026-09-08 — and in late August Canaan sold all of its Ethereum plus 54 Bitcoin to raise roughly $13.9 million, then recycled it into buybacks. Year to date it has repurchased 16.4 million ADSs for $7.4 million. The keywords use of capital and stock repurchase are doing here what they have done for several quarters, but the conviction behind them matters given the share price and the extended Nasdaq minimum-bid grace period (deadline pushed to January 11, 2027). Management is explicitly choosing digital-asset monetization over passive holding — a genuine change in treasury posture. The quieter pivot is the consumer product line. Avalon Home generated only about $1 million in Q2 revenue, but NG leaned into the seasonality lesson — “After the heating season ended, sales of our home products came down...” — Nangeng Zhang, Chairman and CEO · 2026-09-08 it is not very easy to sell heaters in the summer — and is prepping mass production for the Christmas shopping season. Note that Avalon Home series shows up this quarter as a decliner in the company's momentum history, a reminder that enthusiasm has cooled even as the narrative persists.

The tape is voting the other way

Here is the contrast that makes this dossier interesting: Canaan is leaning hard into themes that the broader market's own tape has recently turned against. Across Know Trend's market-theme price movers, megawatts of power is a 360-day advancer but a 90- and 30-day decliner, meaning the move is stale. AI data centers likewise ranks as a long-window winner yet shows a broadly negative 30-day breadth — the AI-datacenter trade is rolling over just as Canaan tells investors it wants in. Crypto-adjacent themes are no help either: price of Bitcoin, cost to mine, and self-mining hash rate all sit in the 90-day decliner bucket. So the company is betting a $325 million market cap on a power/data-center future while its core demand signal is deteriorating and the market's enthusiasm for that same future is fading. That is not a reason to dismiss it — step by step is a fair description of how management says it will proceed — but it is the reason the risk sits where it does. Revenue is collapsing, the pivot is unmonetized, and the only concrete support for the equity right now is a buyback funded by selling crypto and a power pipeline management still won't describe in detail. Worth watching closely; not yet worth believing.