Cango

Phase A — Understand the business

Lens 1 · Company Overview

Cango is no longer the company its ticker history describes. From its 2018 NYSE IPO through 2024 it was a Chinese auto-financing transaction platform — matching car buyers, dealers, and financial institutions in the PRC. In November 2024 it launched a Bitcoin-mining business, and over the following ~9 months executed a near-total identity transplant: on 2025-05-27 it divested all PRC operations (now reported as discontinued operations), and reconstituted itself as one of the larger publicly-traded Bitcoin miners in the world ``.

What it does today (mid-2026):

1. Bitcoin self-mining + hashrate leasing — the revenue engine. ~$675.5M of FY2025's $688.1M total revenue came from Bitcoin mining . Operations are described as spanning **40+ sites across North America, the Middle East, South America, and East Africa** .
2. AI / HPC compute (EcoHash) — a 2026 pivot. Subsidiary EcoHash Technology LLC began commercial operations in April 2026, retrofitting Cango's owned 50MW Georgia site into modular, containerized GPU inference capacity, targeting long-tail SME inference demand via an "EcoLink Orchestration Platform" ``.
3. Residual auto-export — a small online international used-car export business (AutoCango.com), now immaterial ``.

The structural truth that frames every other lens: Cango is effectively a Bitcoin-mining proxy for Bitmain, assembled via share issuance to Bitmain-affiliated sellers. Its two most important "customers/suppliers" are related parties: Antpool (Bitmain's pool) took ~100% of mining revenue in 2024 and ~98.57% in 2025; Bitmain affiliates supplied up to 100% of energy/hosting; Antalpha (Bitmain's financing arm) was the dominant lender — ~$557.6M of loans, roughly half of total assets at 2025-12-31 ``. Contract structure is therefore not arm's-length — this is the single most important fact about the business.

Lens 2 · Supply Chain

Map upstream → company → end customer, naming the actual stakeholders. Cango's "supply chain" is unusually short and unusually captive — it runs almost entirely through the Bitmain orbit:

• ASIC hardware (upstream): Bitmain (Antminer rigs). Cango's hashrate was acquired as machines/on-rack hashrate from Bitmain and Bitmain-affiliated sellers — 32 EH/s in the 2024 deal, +18 EH/s from Antalpha in 2025, plus 18 EH/s of machines from Golden TechGen Limited (wholly owned by a former Bitmain CFO, Max Hua; later transferred to Ning Wang / Youngil Kim / Wye Sheng Kong) ``.
• Financing (upstream capital): Antalpha (Bitmain's financing affiliate, owned by Antpool, ultimately controlled by Bitmain chairman Micree Zhan) — ~$557.6M of loans, ~half of total assets ``. Bitcoin-collateralized.
• Energy / hosting: Bitmain-affiliated providers supplied "up to 100%" of energy and hosting costs . Sites span **40+ locations** across N. America, Middle East, S. America, E. Africa — a deliberately distributed, "asset-light, plug-and-play" footprint.
• Mining pool (the buyer of hashrate output): Antpool — ~100% (2024) / ~98.57% (2025) of mining revenue ``.
• End "customer": for mined BTC, the open market (Bitcoin liquidity itself); for EcoHash, prospective AI-inference customers (SMEs) — not yet a material revenue line ``.

Chokepoints / single-source dependencies (severe): hardware (Bitmain), financing (Antalpha), hosting/energy (Bitmain affiliates), and the revenue counterparty (Antpool) are the same controlling group. This is the inverse of supply-chain diversification — a single related party sits on all four nodes. Any repricing or withdrawal by Bitmain/Antalpha hits hardware, capital, power, and revenue simultaneously ``.

Lens 3 · Competitive Advantages (moats)

Bluntly: there is no durable moat. Bitcoin mining is a commodity business — the product (a freshly minted BTC) is identical regardless of who mines it, and competitive position is decided entirely by cost per coin (machine efficiency × power price × uptime). On that axis Cango is a laggard, not a leader:

• Cost position (the only thing that matters): FY2025 cash cost to mine ex-depreciation $79,707/BTC, all-in $97,272/BTC; Q4 2025 worse at $84,552 / $106,251 . With Bitcoin in the **$76k–$80k** range by mid-2026 , Cango's all-in cost is above the Bitcoin price — i.e. it mines at a structural loss before any treasury gains. By contrast, low-cost US miners target sub-$50k all-in. March 2026 progress ($68,215/coin cash cost, −19.3% q/q) is improvement off a bad base ``.
• Scale: peaked at 50 EH/s nameplate (mid-2025), but operational hashrate has since shrunk to 31.67 EH/s (May 2026) as the company decommissions high-cost legacy rigs ``. So even the scale "moat" is contracting by choice.
• Switching costs / network effects / brand: none. Miners have zero customer lock-in.
• The only arguable edges, and they are double-edged: (a) privileged access to Bitmain hardware/financing as the de-facto Bitmain proxy — but that is a dependency, not a moat; (b) a distributed 40-site energy footprint that could, in theory, be repurposed for AI/HPC (the EcoHash thesis) — but every miner is making that pivot, and Cango is sub-scale and capital-starved relative to IREN/Core Scientific/Riot.

Bargaining power: essentially nil over its suppliers (Bitmain controls hardware, power, and the lending), and nil over the BTC price. Cango needs Bitmain far more than Bitmain needs Cango.

Lens 4 · Segments

No segments.csv on disk (web-only). Revenue is effectively single-segment (Bitcoin mining) post-transformation:

• Geography: mining is globally distributed (40+ sites, four continents) but not broken out by region in public releases — n/a. The legacy PRC auto business is discontinued operations (divested 2025-05-27), so historical China revenue is not comparable to today ``.
• Trend: mining revenue grew on hashrate ramp (Q4 2025 revenue +99.6% y/y to $179.5M ), then **reversed sharply** — **Q1 2026 revenue −43%** as BTC price fell and the company shrank the fleet and sold coins . The forward direction is deliberate contraction of mining toward a higher-margin core plus an AI-compute bet — i.e. the "segment" story is mid-pivot and not yet legible.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest prints)

FY2025 (reported 2026-03-16, unaudited):

• Total revenue $688.1M; Bitcoin-mining revenue $675.5M ``.
• Net loss from continuing operations −$452.8M — driven by "non-recurring transformation costs and market-driven fair-value adjustments" (i.e. BTC/asset markdowns) ``.
• Adjusted EBITDA +$24.5M for the year, but Q4 adjusted EBITDA −$156.3M — the exit rate was deeply negative ``.
• BTC mined: 6,594.6 (18.07/day); Q4 1,718.3 (18.68/day) ``.
• Mining cost: $79,707/BTC cash, $97,272/BTC all-in (FY); $84,552 / $106,251 (Q4) ``.

Q4 2025 specifically: revenue $179.5M (+99.6% y/y), net loss −$291.7M vs +$7.7M net income in Q4 2024 ``.

Q1 2026: revenue −43% y/y, net loss −$261.1M — "mainly due to Bitcoin price declines and impairment charges" . **1,266 BTC mined** in Q1 2026 .

Margin / balance-sheet flags: all-in cost > BTC price = structurally loss-making mining; enormous fair-value/impairment swings make GAAP net income nearly meaningless and put the focus on cash and leverage (see Lens 10). Cash/total-assets line items were not disclosed in the results release — n/a.

Market reaction: the tape has been brutal regardless of the operational headlines — stock down >70% YTD 2026, from ~$1.40 in January to ~$0.33 by mid-June 2026 ``. The market is pricing dilution + delisting + related-party + loss-making mining, not the hashrate number.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts/ on disk (web-only). Cango as a 20-F filer reports via 6-K results releases and monthly operational updates rather than a conventional quarterly call cadence, so "sentiment" is read off the management framing in releases, which has shifted clearly across 2025→2026:

• H1 2025 — expansionist / triumphalist: "accelerating strategic transformation," racing past 50 EH/s, "one of the largest Bitcoin miners in the world" ``.
• Late 2025 — long-vision pivot language: introduces the goal to "evolve from a leading Bitcoin miner into a global, distributed AI compute network powered by green energy" ``.
• 2026 — defensive / deleveraging / discipline: the verbs change to "strategically optimizing," "fleet modernization," "prioritize margin and cost-efficiency over scale," "strengthen the balance sheet, reduce leverage, secure liquidity" ``. The phrase that disappeared is growth/scale; the phrase that appeared is survival/efficiency. That tonal inversion — from "we are huge" to "we are getting lean and paying down debt" — is the single clearest sentiment signal, and it is consistent with a balance-sheet rescue, not an expansion.

Lens 7 · Comps

Bitcoin-miner peer set.

• EV/Sales, EV/EBITDA, P/E, ROE for the peer table: n/a (would be fabrication to assert precise multiples). The honest read: EBITDA-negative exit rate and GAAP net losses make P/E and EV/EBITDA meaningless for Cango; the defensible relative metric is EV per EH/s and mNAV (price vs. BTC + cash − debt).
• EV per EH/s sanity check: Cango ~$130M mkt cap / ~31.67 EH/s ≈ ~$4.1M per operating EH/s `` — optically cheap vs. multi-billion peers, but the discount is deserved: above-price mining cost, ~half the balance sheet in Antalpha loans, delisting risk, and ~50% captive voting control. Cheap-on-hashrate is the value trap here.
• Total public-miner market cap ~$88B mid-2026 `` — Cango is a rounding error (~0.15%) of the sector.

Lens 8 · Stock-Price Catalysts (what actually moves CANG)

The >5% moves cluster around dilution / control / solvency events far more than hashrate:

• 2024-11 → 2025-05: pivot announcement + PRC divestiture + Bitmain-hardware deals → large speculative re-rate (the 52-week high reached ~$5.75) ``.
• 2025-06/07: Antalpha/EWCL assume voting control; 146.7M Class A shares issued to mining-asset sellers (Golden TechGen ~19.85%; sellers ~41.38% collectively) — massive dilution overhang ``.
• 2026-02: sold >half its BTC — ~6,451 BTC across Feb–Mar for ~$442M to retire BTC-backed loans → CoinDesk-flagged forced-sale-into-weakness ``.
• 2026-03-10: NYSE sub-$1 non-compliance notice (30-day avg < $1.00); 6-month cure window ``.
• 2026-04: $65M leadership strategic investment + $10M DL Holdings convertible note ($1.62 conversion, due 2028) + EcoHash commercial launch ``.

Pattern: CANG is not a "beat-and-raise" stock. It trades on capital-structure events — who controls it, how much new stock is coming, whether it can pay down Antalpha, and whether it stays listed. BTC price matters (it's levered to it), but idiosyncratic dilution/solvency risk dominates the tape. This is a special-situation security, not an operating-momentum one.

---

Phase C — Judge people & books

Lens 9 · Management

The board and C-suite were replaced wholesale in 2025 as control passed to the Bitmain/Antalpha orbit — this is the central governance fact:

• Co-founders Xiaojun Zhang & Jiayuan Lin resigned and transferred their shares (and voting control) to EWCL (an Antalpha entity) ``.
• Xin Jin — founder & CEO of Antalpha (Bitmain-backed) — appointed Chairman / Non-Executive Director ``. The financier's CEO now chairs the borrower.
• Peng Yu — CEO & Director; Yongyi Zhang — CFO; Chang-Wei Chiu — Director; Simon Ming Yeung Tang — Chief Investment Officer; Jack Jin — CTO of the AI business line (EcoHash) ``.

Assessment against the five tests:

1. Track record: the new team carries digital-asset infrastructure / Antalpha / Bitmain energy-finance pedigree — credible operators of a Bitmain-aligned miner, but with no track record running this public vehicle through a full cycle. Unquantifiable here — n/a on prior P&L delivery.
2. Tenure & skin in the game: brand-new (since mid-2025). "Skin" is real but conflicted — the controlling shareholders (EWCL/Antalpha, Golden TechGen) are Bitmain affiliates; the $65M April-2026 investment came from "members of Company leadership", which deepens insider ownership but also insider control ``.
3. Capital allocation: the defining decisions — issue 146.7M shares to buy Bitmain hashrate at the top, then dump >half the BTC treasury into a downdraft to repay Antalpha loans — have been value-destructive for minority holders (stock −70%+ YTD). Deleveraging is rational given the debt, but the whole arc (lever up via related party → forced de-lever) reflects a vehicle run for the controlling lender's risk, not minority NAV.
4. Red flags: pervasive related-party dealing is the headline — chairman = financier's CEO; lender, hardware vendor, hosting provider, and revenue counterparty all = the same Bitmain group; ~$557.6M of related-party loans . EWCL voting power rose from ~36.7% to **~49.71%** by Feb 2026 .
5. Founder vs. professional manager: neither, exactly — this is a controlled-vehicle management team installed by the controlling strategic (Bitmain/Antalpha). The implication: minority shareholders are along for the ride on Bitmain's strategy, with limited ability to influence outcomes.

Lens 10 · Forensic Red Flags

Acting as a forensic equity analyst. This is a high-risk forensic profile even though there is no SEC enforcement on file.

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. regulatory/regulatory-findings.md (generated 2026-06-17 via SEC EDGAR EFTS, LR + AAER, period 2021-06-17→2026-06-17) reports total_sec_findings: 0 — "No LR found," "No AAER found." ``.
• Non-SEC enforcement (web search — "Cango" (FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine OR penalty) enforcement): no material agency enforcement action surfaced as of 2026-06-17 . (The only "regulatory" event is the **NYSE continued-listing notice**, which is an exchange listing-standard matter, not an enforcement action .)
• 10-K Item 3 (Legal Proceedings): the 20-F equivalent is not on disk (filings=0) — n/a for a verbatim quote; the public 20-F summary flags risk-factor concentration, not active material litigation ``.
• Verification statement: No material enforcement/regulatory finding — verified via SEC EDGAR EFTS (LR, AAER) and web search as of 2026-06-17. The forensic concerns below are accounting/structural, not enforcement.

Accounting & structural risks (the real story):

• Related-party everything (income statement + balance sheet): ~$557.6M Antalpha loans ≈ ~50% of total assets ``; ~98.57%/100% of mining revenue via Antpool; energy/hosting up to 100% via Bitmain. Revenue, COGS, and financing are all transacted with the controlling group — transfer-pricing opacity is structural. This is the textbook profile auditors flag for related-party disclosure scrutiny.
• Revenue recognition / fair value: mining revenue is recognized in BTC, and the income statement is dominated by fair-value remeasurement and impairment of crypto holdings — the −$452.8M FY2025 and −$261.1M Q1 2026 losses are largely non-cash markdowns ``. GAAP earnings are therefore noise; cash and debt are signal.
• Cash flow vs. earnings divergence: the company sold ~6,451 BTC for ~$442M specifically to retire BTC-backed loans, cutting collateralized debt to ~$30.6M ``. Liquidating the treasury to service related-party debt is a going-concern-adjacent signal — operating cash generation alone was insufficient.
• Share-count / dilution: 146.7M Class A shares issued for hashrate (2025) + a $1.62-strike convertible (2028) + leadership equity rounds = persistent dilution of minority holders . Reported shares outstanding diverge across sources (**~268.6M vs ~363.35M**) — that the float itself isn't cleanly sourceable is its own red flag.
• Goodwill/intangibles & "unaudited": FY2025 results are explicitly unaudited; as a recently-transformed 20-F filer the audited financials and auditor identity/PCAOB status warrant verification — n/a here.

Net forensic read: no smoking-gun fraud, but a maximal related-party-concentration + heavy-dilution + treasury-liquidation-to-repay-the-affiliate-lender profile. For a forensic analyst this is a stay-away on the long side until audited financials and arm's-length terms are demonstrable.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

EPS projection for the next three fiscal years (FY2026E–FY2028E). Web-only inputs; output `` with arithmetic. No forecast.ts create (breadth loop).

EPS modeling for Cango is low-confidence by nature — earnings are dominated by non-cash BTC fair-value swings and a moving share count, and the business is mid-pivot. The honest framing is scenario, not point estimate:

• Revenue base: mining revenue is contracting — operational hashrate fell 50→31.67 EH/s, BTC sold down to ~1,057 BTC (Apr 2026), Q1'26 revenue −43% ``. EcoHash AI revenue is pre-material.
• Cost base: all-in cost ~$97k/BTC (FY25) improving toward ~$68k cash cost (Mar'26) but still thin-to-negative vs. ~$76–80k BTC ``.
• Share count: ~268.6M–363.35M and rising with the convertible ``.

• Point EPS for each year: n/a / not reliably modelable — asserting a specific EPS line for a fair-value-dominated, mid-pivot micro-cap would be fabrication. The defensible forward statement is: mining economics are sub-breakeven at current BTC; the equity is a levered BTC + optionality-on-EcoHash call, gated by dilution and delisting.
• The question that actually matters (special-situation framing): not "what's FY27 EPS" but "does the deleveraged balance sheet + $75M of 2026 capital carry it through the NYSE cure period and to EcoHash revenue, without another dilutive raise or a forced reverse-split?" — currently unresolved ``.
• Brier forecast: not logged (watchlist breadth loop — forecast.ts create skipped per instructions).

Lens 12 · Bull vs Bear

Bull case (steelman). Cango is a deleveraged, distressed BTC-mining + AI-compute option trading near a floor. It dumped >half its BTC to crush collateralized debt to ~$30.6M, raised ~$75M in April 2026, and is run by a Bitmain/Antalpha team with privileged access to the cheapest hardware and energy-finance in the industry . At ~$130M market cap on ~31.67 EH/s (~$4.1M/EH/s ), it's optically the cheapest hashrate in the listed universe, and EcoHash gives free optionality on the AI-inference land-grab using an already-built 40-site, grid-connected energy footprint. If BTC re-rates >$100k and EcoHash converts even modest AI revenue, a sub-$1 micro-cap can multiple quickly. It is a high-beta call option on BTC + AI compute.

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

1. Related-party capture (the structural one): Bitmain/Antalpha control ~50% of the vote and sit on every node of the value chain. Minority equity can be diluted, re-priced, or restructured for the controlling lender's benefit with little recourse — a permanent governance discount, not a temporary one.
2. Sub-breakeven mining + commodity economics: all-in cost has run above the BTC price; the only fix is cheaper machines (more Bitmain dependence) or a higher BTC price (not in their control). Shrinking the fleet caps the upside even if BTC rises.
3. Solvency / delisting spiral: sub-$1 since March 2026, >70% YTD decline, treasury already half-liquidated; another leg down in BTC forces either more dilution or a reverse split / delisting — either of which can permanently impair the equity ``.

Pre-mortem (18 months out, thesis broke): BTC chopped sideways/down, EcoHash never reached material AI revenue (sub-scale vs. IREN/CORZ, capital-starved), Cango did one more dilutive raise and a reverse split to keep the NYSE listing, and Antalpha restructured its loans on terms that subordinated minority holders. The stock is a sub-$0.20 zombie or delisted to OTC.

Are multiples too high? On hashrate it looks low, but EBITDA/earnings multiples are meaningless (negative). The right lens is mNAV / solvency, where the discount is warranted, not excessive.

Contrarian view (what the market refuses to see): the bull and bear both treat CANG as a miner. The likeliest real outcome is neither — it becomes a Bitmain-controlled corporate shell whose value accrues to Antalpha (as lender/controller) regardless of the BTC price or EcoHash, while the listed minority equity is the residual, dilutable claim. The market's mistake isn't mispricing the hashrate — it's pricing this as an operating company at all rather than as a controlled financing vehicle.

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

Dismantling the bull case.

• What structurally breaks the money machine: it's already broken — all-in mining cost > BTC price. The "business" only "works" via treasury gains and related-party financing, neither of which is operating earnings.
• Revenue concentration: ~98.57–100% of mining revenue runs through Antpool (related party). If Bitmain repriced pool fees, hosting, or hardware terms — or simply de-prioritized Cango — revenue and margin move at the controller's discretion ``.
• Why the moat is weaker than bulls think: there is no moat — it's a commodity miner and a sub-scale, capital-starved AI-compute also-ran. EcoHash competes with IREN/CORZ/Nebius for the same GPUs, power, and customers, with a fraction of the capital.
• Most dangerous competitor bulls underestimate: not another miner — it's the well-capitalized AI-pivot miners (IREN, Core Scientific) that will win the inference land-grab while Cango is still retrofitting one 50MW site.
• Worst capital-allocation / governance moves: issuing 146.7M shares to buy Bitmain hashrate near the top; liquidating >half the BTC treasury into weakness to repay the affiliate lender; installing the lender's CEO as chairman. The incentives are misaligned with minority holders by construction.
• Assumptions required for today's price: that it survives (stays listed, doesn't dilute to zero) — current price is closer to a solvency option than a going-concern multiple.
• If growth disappoints 20–30%: mining is already shrinking ~37% off peak hashrate; further BTC weakness or EcoHash failure drives another dilutive raise / reverse split.
• Single scenario that permanently impairs: BTC range-bound at $60–80k for 18 months → loss-making mining + no EcoHash traction + a forced dilutive recap on terms set by the controlling lender → minority equity permanently impaired. Plausibility: moderate-to-high.

Lens 14 · Management Questions (ordered by information value)

1. What are the exact terms of the ~$557.6M Antalpha facilities (rate, maturity, collateral, covenants, cross-default), and what is the current outstanding balance after the 2026 BTC-funded repayments?
2. Antpool takes ~99–100% of mining revenue and Bitmain affiliates supply ~100% of energy/hosting — on what terms, and are they arm's-length and benchmarked? What is the contractual repricing risk?
3. What is the definitive, fully-diluted share count today (sources disagree at 268.6M vs 363.35M), and what further issuance is committed (convertible, leadership rounds, earn-outs)?
4. What is your plan and timeline to regain NYSE compliance before the cure window closes — and is a reverse split on the table?
5. EcoHash: what is signed, contracted, or piloted AI-inference revenue and committed customer demand today — in dollars, not capacity?
6. What is the all-in cost per BTC trajectory by quarter, and at what BTC price does the current fleet mine at positive all-in margin?
7. How much GPU capex does the EcoHash roadmap require, and how is it funded without further dilution or related-party debt?
8. Who is the auditor, what is their PCAOB inspection status, and when will audited FY2025/FY2026 financials be available?
9. What governance protections exist for minority shareholders given EWCL/Antalpha control ~50% of the vote? Any independent-committee approval for related-party transactions?
10. What is your liquidity runway at current burn and current BTC price, and what triggers the next capital raise?
11. Is the remaining Bitcoin treasury (~1,057 BTC, Apr'26) a strategic reserve or a liquidity buffer to be sold — and under what rule?
12. What happens to the 40+ mining sites as you shrink the fleet — sold, idled, or converted to AI/HPC, and on what timeline?
13. How do you reconcile "global, distributed AI compute network" ambitions with a market cap (~$130M) a fraction of IREN/CORZ?
14. What is the strategic role of the DL Holdings MOU and the residual AutoCango export business — core, optional, or divestiture candidates?
15. What is the long-term role of Bitmain/Antalpha — is Cango intended to remain a permanent Bitmain proxy, or to diversify away from the group over time?

---