Canaan

Phase A — Understand the business

Lens 1 · Company Overview

Canaan is a fabless ASIC designer whose entire product line is Bitcoin mining machines sold under the AvalonMiner brand, plus a growing self-mining ("mining revenue") business it bolted on as a "second engine." It is a Cayman holding company, HQ in Singapore (28 Ayer Rajah Crescent), with R&D/operations rooted in China and assembly across the PRC, Southeast Asia, and (since Q1 2025) the United States. ADSs trade on Nasdaq under CAN, each ADS = 15 Class A shares, listed since Nov 2019.

How it makes money, FY2025:

• Products revenue (mining machines + parts): $413.8M (78.1% of revenue)
• Mining revenue (self-mining BTC): $113.2M (21.4%)
• Other: $2.7M (0.5%)
• Total revenue: $529.7M, up 96.7% YoY from $269.3M.

Contract structure is the opposite of recurring: machines are sold "first-pay-first-serve," one-off contracts / POs, no long-term agreements, no returns except for major defects, 360-day warranty. Demand is a direct derivative of expected mining economics (BTC price × hashprice). Customers are individual and corporate miners across North America, West Asia, SE Asia, Europe, Africa; 58.9% of FY2025 machine revenue was outside the PRC (up from 34.3% in 2023 as the business fled China's 2021 mining ban). Suppliers: a fabless model on two third-party foundry partners (unnamed; almost certainly TSMC-class) under 3-year framework agreements with ≥50% prepayment and 4–6 month wafer lead times. Competitors: Bitmain and MicroBT (see Lens 3).

The company is tiny as an org: 399 total employees, 189 (47.4%) in R&D. This is a chip-design shop with a global sales footprint, not an industrial behemoth.

Lens 2 · Supply Chain

Named map, input → output:

Upstream (inputs):

• Wafer foundries — 2 partners (unnamed in the 20-F; the leading-edge nodes implied — A16 at "12.8 J/TH" needs ~5nm-class — point to TSMC and/or Samsung; not disclosed, so ). 3-year frameworks, ≥50% prepay, 4–6mo lead. This is the single largest chokepoint — "in 2025, we mainly relied on two third-party foundry partners".
• Packaging & testing (OSAT) — separate "leading" partners, net-30 settlement.
• Components — circuit boards, PMU boards, cooling fans, heat sensors, aluminum casing, PSUs (sourced/procured, integrated in-house).

Midstream (Canaan): Front-end + back-end IC design in-house → wafer out to foundry → OSAT → in-house assembly at a 7,378 m² PRC plant + assembly partners in SE Asia and the US. As of 2025-12-31, mining equipment net carrying value was $26.7M in North America and $4.5M in Ethiopia — the self-mining fleet's geography.

Downstream (customers / end market):

• Distributors (significant in SE Asia) + direct sales + an online store reaching ~100 countries.
• Corporate miners. The one named, material counterparty: Cipher Mining — both a ~7% shareholder (Lens 7) and a customer/JV partner. Cipher bought 6,840 Avalon A15Pro rigs in mid-2025, deployed at its Black Pearl site. WindHQ LLC is the 51% operating partner in the West Texas JVs Canaan just consolidated.
• Tether — co-developing modular mining systems with Canaan + ACME Swisstech (compute/power/cooling separated).

Single-source dependency verdict: foundry concentration (2 partners) is the structural chokepoint; tariff/sanction risk on China-origin hardware is the geopolitical chokepoint driving the US-assembly pivot.

Lens 3 · Competitive Advantages (moats)

The Bitcoin-rig market is a rare-but-shallow oligopoly: Bitmain + Canaan + MicroBT hold >90% of global rig share; the top five control ~68%. So the category has scale/IP barriers — but within it Canaan is the perennial #3, and a distant one.

Real, durable moats:

• IC-design IP + process know-how: 622 patents (185 invention) + 734 trademarks + 127 software copyrights + 82 IC layout rights as of 2025-12-31. Multi-generation silicon data is genuinely hard to replicate — "high barriers to entry" is not pure puffery here.
• Time-to-market cadence: new AvalonMiner series almost annually (A8 2018 → A16 2025).
• Foundry access: secured leading-node capacity is itself a moat for a sub-scale player.

Where the moat is thin:

• No switching costs, no network effects, no brand pricing power. Buyers choose on $/TH and J/TH at the moment of purchase; Canaan itself says ASP is set by "supply and demand … Bitcoin price". Machines are commoditized hash.
• Bargaining power is weak on both sides: customers pay first-come-first-serve with no contracts (Canaan needs them more), and Canaan must prepay ≥50% to foundries (foundries need Canaan less). It is squeezed at both ends.
• Bitmain out-scales it on R&D budget, volume, and balance sheet — the 20-F explicitly concedes competitors "have stronger brand names, greater access to capital … more resources".

Net: a legitimate technology moat (IP/know-how/foundry access) wrapped around a commodity end-product in a price-taking, BTC-levered demand pool. The moat keeps Canaan in the top 3; it does not give it pricing power or margin stability.

Lens 4 · Segments

The research layer's segments.csv is empty, so segmentation is grounded in the 20-F. Canaan reports by revenue line and by computing power, not by formal operating segments.

By revenue line (USD M):

Trend & cause: Products +85% YoY in 2025 on both volume and ASP recovery; mining +157% YoY on more energized hash + higher average BTC price. Mining is the faster-growing leg and is being deliberately scaled as the "second engine" to absorb inventory and smooth the cycle.

By computing power sold (the real unit economics):

This is the single most important table in the filing: 2025 was the first year ASP/TH ($11) exceeded cost/TH ($10) — the A15/A1566 generation (32.0M of 35.0M TH/s sold) is the workhorse that flipped unit economics positive. That is the mechanical engine behind the gross-profit turn in Lens 5.

By geography: the 20-F gives revenue-mix only at the machine level — 58.9% ex-PRC in 2025 (vs 69.7% in 2024, 34.3% in 2023). The dip from 69.7% → 58.9% reflects a large US order cycle and growing self-mining (US/Ethiopia), not a return to China.

---

Phase B — Measure performance

Lens 5 · Earnings Result

FY2025 (the latest 20-F print):

• Revenue $529.7M, +96.7% YoY — a near-doubling.
• Gross profit $41.2M (7.8% margin) — the headline: first gross profit after gross losses of $(84.3)M in 2024 and $(240.8)M in 2023. Driven by ASP recovery + a collapse in inventory write-downs (from $100.6M in 2024 to $18.6M in 2025).
• Loss from operations $(112.2)M (vs $(227.1)M) — opex held roughly flat at $153.4M while revenue doubled → operating leverage is real (R&D fell to 11.9% of sales from 22.8%; G&A to 12.9% from 26.6%).
• Net loss $(210.3)M (vs $(249.8)M). The gap between the $112M operating loss and the $210M net loss is almost entirely non-operating, crypto/financing-driven: $(46.6)M fair-value loss on preferred-share instruments, $(28.2)M "excess of fair value of convertible preferred shares," $(11.4)M crypto mark, $(9.5)M derivative mark. The operating business lost far less than the headline.
• Net loss per ordinary share $(2.99)¢ (≈ $(0.45)/ADS ) vs $(6.13)¢ in 2024 — losses narrowing on a per-share basis despite the share count exploding, because the loss shrank faster than dilution.

Balance-sheet flags:

• Cash $80.8M (down from $96.5M) — thin against a $261.1M operating cash burn. ~3.7 months of runway on cash alone; the business is kept alive by the capital market, not operations.
• Inventories ballooned to $180.8M (from $94.6M) — a near-doubling, the central red flag (Lens 10).
• Accumulated deficit $(660.8)M; total equity $437.4M held up only by $1,177.1M of paid-in capital — i.e. shareholders have funded $1.18B to build $437M of book.
• Debt modest: $52.2M total loans (current + non-current); 900 BTC pledged for secured term loans.

Guidance: 20-F gives no formal numeric FY guidance; management points to "trend information" only. Post-period, Q1 2026 was guided and printed at $62.7M revenue (product $42.9M, mining $19.1M), "in line with guidance" — but down sharply sequentially, confirming extreme order lumpiness.

Unusual vs own history: the gross-profit flip and the collapse in write-downs are the genuine positives; the inventory build into a falling BTC tape is the genuine warning.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts in the research layer (transcripts/ empty), so this is `` + filing tone. Management's narrative arc across 2024→2026 is a clear, traceable pivot:

• 2024–early 2025: survival + "second engine" framing — grow self-mining to absorb inventory risk and "synergize" with machine sales.
• Late 2025 → 2026: an aggressive re-positioning to "upstream power development" + AI/HPC + a crypto treasury. The Cipher deal press release frames a "transition from opportunistic, asset-light mining toward a systematic approach … project pipeline potentially at gigawatt scale". April 2026 updates lead with record crypto treasury and global hashrate, not machine bookings.

Recurring phrases: "democratize access" (Avalon Home), "second engine," "low-cost power," and now "hash-to-heat" and "AI-HPC colocation." Things they've stopped emphasizing: China; pure machine-ASP cyclicality. The tone has shifted from defensive (survive the write-down cycle) to offensive (vertically integrate into power + diversify into treasury/AI) — which is either strategic maturation or narrative diversification away from a hard core business, depending on your priors (Lens 13).

Lens 7 · Comps

Canaan, live: price ≈ $0.34–0.36/ADS (2026-06-13), market cap ≈ $230–255M; 52-week range $0.31–$2.21 (down 84% from the high, near the low). Book equity $437.4M (2025-12-31) → trading at ~0.5–0.6× book. Net cash + crypto treasury ($148M crypto + $80.8M cash as of Q1) is a large fraction of the market cap.

Peer set (BTC-mining-rig makers + the listed miners Canaan increasingly resembles):

The honest read: the only two true rig-maker peers (Bitmain, MicroBT) are private with no disclosed multiples, and I did not pull live MARA/RIOT/CIFR multiples this run — so I will not fabricate them (n/a). What is sourceable and decision-relevant: Canaan trades below book and at a large fraction of net-cash-plus-treasury, which is the market pricing in continued dilution and going-concern risk, not a growth multiple. Analyst targets are wildly dispersed — consensus ≈ $2.95–3.26, low $1.30 (Rosenblatt, 2026-05-20), high $8 (BTIG, 2025-10-14) — i.e. 4–9× upside to consensus from $0.34, the signature of a binary small-cap where the analysts are modeling a BTC-bull recovery the tape is not.

Lens 8 · Stock-Price Catalysts

Pattern over the cycle (mostly ``): CAN is a high-beta BTC proxy with idiosyncratic dilution shocks.

• Up >5% drivers: BTC rallies (machine-demand read-through), new-product launches (A16 announce, Oct 2025), the Cipher/power-integration narrative, treasury-accumulation headlines, AI/HPC optionality.
• Down >5% drivers: BTC drawdowns; the Nasdaq minimum-bid deficiency notice (2026-01-16); serial equity issuance (ATM + registered direct + preferred — each a fresh dilution print); the Q1 2026 revenue "collapse" framing; broad miner de-rating as hashprice fell 66% from the Oct-2025 peak.

What the market actually reacts to: (1) the BTC price, (2) the share count, (3) the listing. Fundamentals (gross-margin turn) move it far less than these three. The 84% drawdown from the 52-wk high tracks the hashprice collapse + dilution, not a deterioration in the product.

---

Phase C — Judge people & books

Lens 9 · Management

• Nangeng Zhang — Founder, Chairman & CEO (age 44). ~16 years in electronic-device/IC design; led the team that built one of the first ASIC mining machines; Beihang University (BS electronic info, MS software eng, PhD coursework). The genuine technical founder of the category. Track record: built Canaan from inception to a top-3 global rig maker and through three brutal BTC winters without bankruptcy — real survival skill — but also presided over cumulative net losses of ~$660M and a stock down ~84% off recent highs.
• James Jin Cheng — CFO (since Aug 2021, age 48). 20+ yrs finance (Zhaopin VP, Lenovo, Nokia, Mars); Shanghai Jiao Tong, Peking MBA, Fordham MSc Global Finance, CIMA/IPA fellow. Conventional, credentialed finance operator; the architect of the heavy capital-markets program.
• Tenure & skin in the game — the key governance fact: Zhang holds only 6.2% economic but 31.9% voting via 311.6M Class B shares (15 votes each) held by Flueqel Ltd. (BVI trust, Zhang beneficiary); directors+officers as a group = 6.3% economic / 32.0% voting. So the founder controls the company through supervoting while owning a small slice of the economics — low economic alignment, high control. Years of dilution have washed out his economic stake while his votes persist.
• Capital-allocation history — the weakest dimension. The pattern is dilute-to-survive: FY2025 financing inflow $197.9M from convertible preferred ($99.7M) + registered direct ($70.7M) + ATM ($49.5M); share count went from ~3.2B (end-2023) to ~10.1B Class A (end-2025) and 11.5B+ by the 20-F date. Simultaneously management ran a token $30M buyback (9.4M ADS at avg $0.73, ~$23.1M left) — buying back a rounding error while issuing hundreds of millions of shares, which is value-incoherent. ROE/ROIC are deeply negative throughout. The newer moves (Cipher consolidation, power, treasury) are a genuine attempt to find a higher-ROIC model — but were also funded with stock ($39.75M all-share Cipher deal at $0.7394).
• Red flags: related-party transactions reported as "None" (a positive); no dividends ever; SBC is large but declining ($42.1M → $30.9M → $22.8M). The promotional tilt (treasury/AI/"gigawatt-scale" framing while the core shrinks) is the main behavioral flag.
• Archetype: technical founder-operator with iron survival instincts and a weak capital-allocation record, now controlling via supervoting a company he owns little of. For this stage (sub-scale, sub-$1, cash-hungry) that means: aligned on survival and control, less aligned on per-share value.

Lens 10 · Forensic Red Flags

Acting forensically against the FY2025 statements, the flags, ranked:

1. Inventory build into a falling market — the #1 flag. Inventories +91% to $180.8M (from $94.6M) while BTC and hashprice were rolling over into 2026. Canaan's own history is a write-down machine: $190.2M (2023), $100.6M (2024), $18.6M (2025) in inventory/prepayment write-downs. The low 2025 write-down is what created the gross profit; a renewed BTC slide (BTC ~$63.8K in June 2026, below production cost ) sets up the next write-down. Gross profit quality is hostage to BTC.
2. Depreciation-life change flatters FY2026 — classic. In Q1 2026 Canaan extended mining-equipment useful life from 1.5 to 2.0 years, an estimate change that reduces FY2026 depreciation by ~$5.8M. Small in absolute terms but textbook earnings-management optics, applied right as it chases profitability.
3. Earnings vs cash-flow divergence is enormous. Net loss $(210.3)M but operating cash burn $(261.1)M — cash burn exceeds the accounting loss, and the company stays solvent only via $197.9M of financing. Investing cash is positive ($56.7M) largely from selling crypto ($93.1M proceeds) — i.e. liquidating the treasury to fund operations even while marketing treasury accumulation. That tension is real.
4. Receivables jumped 12.7× to $19.3M (from $1.5M) — small in dollars but a >1,000% increase worth watching for revenue-pull-forward as the model shifts toward larger US orders.
5. Crypto/financing fair-value noise dominates the P&L — preferred-share and derivative remeasurement drove ~$95M of the net loss. Non-cash and non-operating, but it makes the headline number nearly uninterpretable without the bridge, and it reflects a genuinely complex, dilution-heavy capital structure.
6. SBC still meaningful ($22.8M) but declining and a shrinking % of revenue — a positive trend, not a flag.

Mitigants: US GAAP, audited by KPMG Huazhen; clean audit opinion; management concluded disclosure controls AND ICFR were effective as of 2025-12-31, with KPMG attestation — no material weakness. Audit fees $1.83M; accounting basis is GAAP not IFRS.

Regulatory findings (required):

• SEC Litigation Releases / AAERs: None. 0 LR, 0 AAER naming Canaan, 2021-06-18 → 2026-06-18, verified via EDGAR EFTS.
• 10-K Item 3 / 20-F Legal Proceedings: Canaan states it is "currently not a party to … any legal, arbitral or administrative proceedings … likely to have a material and adverse effect". No material litigation disclosed.
• Non-SEC enforcement (web): No material FTC/DOJ/CFTC enforcement, consent decree, or penalty surfaced for Canaan in web search. The structural regulatory exposures are disclosure risks, not actions: (a) HFCAA — auditor KPMG Huazhen is China-based (PCAOB ID 1186); ADSs could face delisting if the PCAOB loses inspection access; (b) PRC risk — possible PRC "resident-enterprise" tax exposure (25%) and China's crypto-mining ban shaping the ex-China pivot; (c) Nasdaq minimum-bid deficiency (not an enforcement action but a live listing risk — Lens 13).
• Conclusion: No material regulatory or legal enforcement findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 20-F Legal Proceedings as of 2026-06-18. The live regulatory risk is delisting (HFCAA + bid-price), not misconduct.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Built bottom-up from FY2025 actuals; all outputs `` with arithmetic shown. Because Canaan is barely gross-profitable and net losses are dominated by non-cash crypto/financing marks, the decision variable is operating loss + cash runway, not EPS — but the skill asks for the three-year EPS path, so:

Anchors (FY2025 actual): revenue $529.7M; gross margin 7.8%; opex $153.4M; operating loss $(112.2)M; net loss $(210.3)M; net loss/ADS ≈ $(0.45). Diluted shares ~7.04B weighted (10.1B+ issued).

Demand backdrop: BTC ~$63.8K, below est. production cost; hashprice ~$32, down 66% from Oct-2025 peak; difficulty just cut 10%; capacity migrating to AI/HPC. This is a trough demand environment for new rigs but a relief for efficient self-miners.

Arithmetic example (base FY2026): $430M × 8% GM = $34.4M gross profit − ~$140M opex (held near flat) = ~$(106)M; less ~$36M lower non-op crypto/financing drag than 2025 → op loss ~$(70)M; ÷ ~11B ADS-equivalent (÷15) → ~$(0.22)/ADS.

The number that actually decides it — runway: cash $80.8M + ~$148M crypto treasury (Q1'26) ≈ ~$229M of liquidity against ~$200–260M annual burn. That is ~1 year before the next raise. The base case requires continued capital-market access; the bear case is a dilution spiral or a covenant/BTC-pledge squeeze. Per-ADS EPS is the wrong scoreboard — dilution is.

(Per --watchlist rules, no Brier forecast logged this run — forecast.ts create skipped.)

Lens 12 · Bull vs Bear

Bull case. Canaan is a genuine top-3 ASIC designer that (a) just proved it can make gross profit ($41.2M, 7.8%) and operating leverage (opex flat as revenue doubled); (b) shipped the A16/A16XP at 12.8 J/TH air-cooled — ahead of Bitmain's S21 XP (13.5 J/TH) on paper, i.e. back at the technology frontier; (c) is vertically integrating into sub-3¢/kWh ERCOT power via the Cipher JV consolidation (120 MW, 4.4 EH/s), structurally lifting mining-segment ROIC; (d) holds a ~$148M crypto treasury as embedded BTC/ETH optionality; and (e) has AI/HPC optionality (Black Pearl conversion, Tether modular-mining partnership) on the most valuable resource in the cycle — contracted power. At ~0.5× book and a large fraction of net-cash-plus-treasury, the downside looks "priced." Consensus PT $2.95–3.26 implies multi-bagger upside.

Bear case (2–3 permanent-impairment risks).

1. Dilution is the business model. Share count ~3.2B → 11.5B+ in two years; equity funded $1.18B to hold $437M of book; the cash-burning core cannot self-fund. Even a successful operational turn can leave the per-ADS holder underwater as the float keeps growing. This is the structural killer.
2. Margin is a BTC derivative, not a franchise. The entire 2025 gross profit rests on write-downs falling to $18.6M. BTC at ~$63.8K (below production cost) + hashprice down 66% is precisely the setup that brought back $100M+ write-downs before.
3. Delisting. Sub-$1 since Dec 2025; Nasdaq cure deadline 2026-07-13; a reverse split is the probable "fix," which solves the symptom not the cause and often precedes further weakness.

Pre-mortem (18 months out, thesis broke): BTC chopped in the $50–70K range, new-rig demand stayed soft (Q1'26 lumpiness was the norm), Canaan took another inventory write-down that erased the gross-margin story, it executed a reverse split to hold the listing, and it raised equity again at a depressed price — net-net the operational turn was real but the per-ADS value was diluted and de-rated away.

Are multiples too high? No — they're low (sub-book). The risk here isn't an expensive multiple compressing; it's that book value and share count themselves erode via losses + issuance. The market is right to price going-concern risk; bulls are right that the optionality is cheap. Both can be true.

Contrarian view (what the market refuses to see): the market treats Canaan as a dying sub-$1 BTC proxy, but it is quietly assembling the one asset that matters in 2026 — cheap contracted power feeding both mining and AI/HPC — at fire-sale valuations, while back at product parity with Bitmain. If BTC merely stabilizes and one AI-HPC colocation deal lands, the power+compute story re-rates a name priced for death. The catch: you must survive the dilution to collect it.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration & lumpiness: Q1 2026 revenue $62.7M (product $42.9M) was the completion of "final deliveries under a major U.S. order" — i.e. a single large order can dominate a quarter, and when it ends, revenue falls off a cliff. This is project-style lumpiness dressed as a product business. What's the encore order?
• The moat is weaker than bulls think: Bitmain out-spends, out-scales, and out-ships Canaan; "parity on paper" with the S21 XP is not the same as winning sockets when Bitmain has the balance sheet to discount and the brand miners default to. Canaan has been #3 for its entire life.
• Most dangerous competitor bulls underestimate: not Bitmain — it's the AI/HPC operators themselves bidding power away from mining. The same "redeployment of power to AI/HPC" Canaan touts as its opportunity is the force destroying its core customers' economics (miners unplugging rigs). Canaan is selling rigs into a shrinking-power-for-mining world.
• Worst capital-allocation moves: the dilute-while-buying-back incoherence; funding acquisitions and operations with sub-$1 stock (Cipher at $0.7394) — issuing the most shares at the lowest prices is the cardinal value-destroying sin; and a crypto treasury that is simultaneously being sold to fund opex ($93.1M crypto sold in 2025) — "treasury accumulation" is partly a narrative over a liquidation.
• What must hold for $0.34: that Canaan survives without a catastrophic dilution event, regains the listing, and keeps write-downs low — three things, all BTC-and-capital-market-dependent.
• Growth disappoints by 20–30%: at ~$370–430M revenue with 8% GM, gross profit is ~$30–34M against ~$150M opex — the operating loss widens materially and runway shortens to <1 year, forcing dilution at depressed prices. There's little margin of safety in the operating model.
• Single scenario that permanently impairs: a sustained BTC bear ($40–55K) for 12+ months — it simultaneously (a) kills new-rig demand, (b) forces a fresh inventory write-down erasing gross profit, (c) crushes the self-mining segment and the treasury mark, and (d) shuts the equity window. That is the trifecta that ended weaker miners in prior cycles, and it is plausible, not tail — BTC is already below production cost as of June 2026.

Lens 14 · Management Questions (ordered by information value)

1. What is the funding plan to the next cash-flow-positive quarter — how many more shares will you issue, and at what point do you choose dilution over a reverse-split-then-raise vs. a strategic investor?
2. With BTC below network production cost (June 2026), what is your inventory write-down exposure for FY2026, and how much of the 2025 gross profit reverses if BTC averages $55K?
3. The Q1'26 revenue rested on completing one large US order — what is the contracted/backlog order book for the next 2–3 quarters by customer and EH/s?
4. On the Nasdaq bid-price deficiency (cure by 2026-07-13) — will you effect a reverse split, and what's the ratio and timing?
5. How do you reconcile marketing a growing BTC/ETH treasury while selling $93M of crypto to fund operations — is the treasury a conviction position or a liquidity buffer?
6. The Cipher JV gives you 120 MW at sub-3¢/kWh — what is the path from 4.4 EH/s of consolidated mining to the "gigawatt-scale" pipeline, and how is it financed without further equity?
7. On AI/HPC: is the Black Pearl conversion a Canaan-owned data center or a colocation lease, and do you have a signed compute tenant — or is this optionality only?
8. You hold 6.2% economic / 31.9% voting — how do you think about minority-shareholder alignment when control rests on supervoting shares whose economics have been diluted away?
9. Foundry concentration is two partners — name the single-source risk: what is your second-source plan and have you secured A16XP wafer allocation for 2026–27 volumes?
10. The Tether modular-mining partnership has no timeline — what is the revenue model and exclusivity, and when does it ship?
11. Mining-equipment useful life was extended from 1.5 to 2.0 years in Q1'26 — what operational evidence supports that, given prior fleets impaired on a 1.5-yr basis?
12. What gross margin can the machine business sustain through-cycle (not just at the 2025 ASP peak), and what is the structural floor?
13. How exposed are you to US tariffs/sanctions on China-origin mining hardware, and what % of 2026 volume will be US-assembled?
14. What is the HFCAA contingency if PCAOB inspection access to KPMG Huazhen lapses — do you have an auditor-transition plan?
15. What single operating metric should shareholders hold you accountable to over the next 12 months — EH/s self-mined, $/TH machine margin, or cash runway?

---