Bitmain

Phase A — Understand the business

Lens 1 · Company Overview

Bitmain Technologies is the dominant designer and seller of application-specific integrated circuit (ASIC) Bitcoin mining machines — the Antminer line — founded in Beijing in 2013 by Micree Zhan (Zhan Ketuan) and Jihan Wu. It is the canonical "sell the pickaxes in a gold rush" business: it does not (primarily) bet on Bitcoin's price, it sells the machines that everyone else uses to bet on it. Three revenue legs:

1. Antminer hardware (the core, ~majority of revenue). SHA-256 ASIC rigs sold to industrial Bitcoin miners. Current flagship is the Antminer S23 Hyd, a water-cooled rig at 580 TH/s, 5,510W, 9.5 J/TH, released January 2026; the prior-gen S21/S21 Pro/S21 XP family remains the volume product. Peak historical market share exceeded 80%; current share is ~35-40% of new ASIC sales.
2. Antpool — the mining pool. Bitmain operates one of the two largest Bitcoin mining pools; this gives it a second, recurring fee stream and deep visibility into customer demand (see Lens 13 for the conflict this creates).
3. Proprietary / captive mining + cloud (via BitFuFu and direct deployment). Bitmain self-mines and deploys its own (or controlled) fleets — court records reveal "at least 16,000 Bitmain-owned or controlled miners" running in the US alone, ~4.2 EH/s, "likely underrepresenting total proprietary US capacity". It also seeded BitFuFu (Nasdaq: FUFU), a cloud-mining vehicle, and signed a 2-year framework to supply it up to 80,000 S-series miners.

Contract structure: mostly purchase orders + multi-quarter framework agreements with large miners (e.g. ~30,000 S21+ units to institutional buyers late-2024/2025; the 80,000-unit BitFuFu frame) plus spot sales. Pricing is brutally cyclical and tied to hashprice — when mining economics weaken, Bitmain discounts hard rather than hold inventory (see Lens 5). Main customers: every major public miner (Riot, MARA, CleanSpark, IREN, Cipher, Hut 8, TeraWulf, Core Scientific, Bitfarms) plus the captive/cloud channel. Main supplier: TSMC (and historically Samsung for some parts). Main competitors: MicroBT (Whatsminer) and Canaan (Avalon) — together the three control ~95-98% of ASIC supply.

Lens 2 · Supply Chain

Named, end-to-end — upstream input → Bitmain → end customer:

• Leading-edge wafers — TSMC (the single chokepoint). Bitmain's most efficient Antminers run on TSMC's N5 (5nm) process; Bitmain has been a TSMC ASIC customer since pre-ordering N5 capacity in 2021. MicroBT has moved to 3nm, and Whatsminer trialed Samsung's 3nm GAA, so Bitmain is arguably one node behind its closest rival on the bleeding edge. This is the binding constraint and the single most important fact in the dossier: TSMC's N5/N3 lines are ~100% booked by AI/smartphone demand into 2026, so Bitmain competes for wafer allocation against Nvidia/Apple/Qualcomm — and the geopolitical lever runs straight through here (see below).
• Chip packaging/test → contract assembly. Final Antminer assembly was offshored years ago — Bitmain and MicroBT assemble "the vast majority of their machines in Malaysia, Thailand, and Indonesia," having moved out of China in 2018 in response to Trump's first-term tariffs. As of January 2026 Bitmain is standing up US assembly (Texas/Florida HQ, ~250 initial workers) to dodge tariffs and customs holds.
• Bitmain → distribution → miners. Direct sales + framework agreements to the public-miner buyer base named in Lens 1, plus its own captive fleet and BitFuFu's cloud channel.
• Chokepoints & single-source dependencies:

1. TSMC leading-edge wafers — no Western alternative at node; Samsung is the only other credible leading-edge foundry and is also export-controlled.
2. US Customs (CBP) — a policy chokepoint inserted into the physical chain: import weight of mining rigs fell ~65% YoY in Jan 2025 as CBP held Bitmain shipments.
3. The Sophgo cutoff — TSMC suspended wafer starts to Bitmain's AI-affiliate Sophgo in Oct 2024 (Entity List Dec 2024), severing the AI-chip leg's supply (Lens 10).

Names present, chokepoints marked — lens passes.

Lens 3 · Competitive Advantages (moats)

Bitmain has a real, multi-layered moat on the supply side and an eroding moat on the demand/geography side:

• Scale + foundry access (durable). Volume buys priority TSMC allocation and amortizes a huge in-house ASIC design team; new entrants can't get leading-edge wafers without Bitmain-scale commitments. This is the classic semiconductor scale moat.
• Brand / installed base (durable). "Antminer" is the industry benchmark; it claims 23 of the top-50 most profitable miners by model (MicroBT 15, Canaan 6). Switching costs are low per-unit but Bitmain's resale liquidity, firmware ecosystem, and repair network are stickiest at industrial scale.
• Vertical integration / demand visibility (double-edged). Owning Antpool + a captive fleet gives Bitmain unmatched real-time read on miner economics — useful for production planning, toxic for customer trust (Lens 13).
• Process/IP (contested). Bitmain leads on efficiency historically but is arguably a node behind MicroBT (5nm vs 3nm) — the moat here is narrowing, not widening.
• Bargaining power: strong over customers (it's the default supplier in a 3-player oligopoly; customers need Bitmain more than any single customer matters), but weak over its key supplier — TSMC has all the leverage, and TSMC answers to US export policy. The decisive moat fact: Bitmain's deepest moat (leading-edge silicon) is precisely the asset a hostile government can sever — and is actively trying to. A moat the US Treasury/Commerce can switch off is not a moat you can underwrite in a Western portfolio.

Lens 4 · Segments

segments.csv is empty (: header-only, no rows) — Bitmain is private and does not report segments. Web-only, all /`` and explicitly unaudited:

Geography: historically China-designed, SE-Asia-assembled, globally sold with the US as the single largest end-market — which is exactly why US policy exposure (below) dominates the thesis. No `` segment figures exist; do not fabricate a split.

Phase B — Measure performance

Lens 5 · Funding & valuation trajectory (+private swap — replaces "Earnings Result")

No SEC financials exist (no CIK, financials.csv empty). Round history and marks, all ``, unaudited:

Key financial datapoints (all dated, all caveated):

• Peak revenue ~$2.5B (2017), ~$2B in Q1 2018. KPMG-audited net income in the 2018 IPO prospectus was $701M — versus a leaked CICC pre-IPO deck claiming $1.25B, a ~$550M overstatement that inflated implied multiples to investors. This is a permanent provenance red flag on any Bitmain "valuation."
• 2018 IPO chatter floated a $40-50B market cap / up to $18B raise — never realized.
• 2022 bear market left Bitmain with a massive S19 inventory backlog it couldn't slow producing without losing TSMC wafer allocation. Reported (unconfirmed) $740M Q3 2018 loss during the prior BCH crash.
• Late-2025 demand crash: hashprice fell from ~$0.12/TH/day (Apr 2024) to ~$0.049 (Apr 2025) to ~$35-38/PH/day (Dec 2025, lowest since early 2023); Bitmain slashed S19/S21 prices to $3-4 per terahash to clear inventory.

Burn signal read: Bitmain is cash-generative at the top of the cycle and bleeds working capital into inventory at the bottom — the price cuts are a classic clear-the-channel move, not necessarily distress, but they confirm the cycle has rolled over. n/a — private, not disclosed for current cash position.

Lens 6 · Founder & company signal (sentiment trend) (+private swap — founder interviews/disclosures, not earnings calls)

No earnings calls (transcripts/ empty). Reading management's public posture over time:

• 2018-2021: consumed by the Wu↔Zhan civil war (Lens 9) — the worst possible distraction during the Sophon AI build-out.
• 2024-2025: posture pivots to (a) AI hedge (Sophon "two more chip generations" announced) and (b) US localization — global business chief Irene Gao publicly framing the US factory as faster delivery/repair for American customers. Tone is offensive-expansion (US HQ, Vietnam, new plants).
• Late-2025 → 2026: posture turns defensive — Bitmain "firmly denied" the Operation Red Sunset allegations, calling remote-control claims "unequivocally false" and distancing itself from Sophgo. The recurring phrase shifted from growth/efficiency to "strictly complies with US laws." When a hardware monopolist's main public message becomes a compliance denial, that is the sentiment tell.

Lens 7 · Cap table & secondary marks (+private swap — replaces "Comps")

Cap table / syndicate quality. Founders dominate the register. As of April 2025: Micree Zhan is the largest shareholder (was ~36% pre-settlement); Jihan Wu fully exited — Zhan bought ~half the company for ~$600M from a founding-shareholder group ("Bitsource," incl. Wu) in the Jan 2021 settlement. External syndicate: Sequoia China, IDG, Coatue, CICC — tier-1 names, but the most recent primary capital is ~2018-2019 vintage. No crossover-fund (Fidelity/T. Rowe/Coatue-style mutual-fund) markup is visible post-2021 — i.e. no IPO-proximity tell. Secondary desks (UpMarket, EquityZen, Hiive) list Bitmain stock, implying some secondary liquidity at the ~$18.78B handle, but these are dealer estimates, not audited marks.

Why no P/E comp table: Bitmain is private — there is no market cap, EV/Sales, EV/EBIT, P/E, or dividend yield to source. n/a — private, not disclosed. The closest public proxies for the ASIC-maker business model (not valuation comps, framing only):

Multiples: n/a (private). Do not fabricate.

Lens 8 · Stock-Price / Value Catalysts (+private — funding/product/regulatory events, no public tape)

No stock, so the analog is "events that re-rated the franchise's perceived value," all ``:

1. 2018 Hong Kong IPO collapse — BTC crash + the CICC-vs-KPMG income discrepancy killed the deal; ~$15B peak mark evaporated.
2. 2019-2021 founder war — Zhan ousted (Oct 2019), regains control (Jan 2021); $600M buyout of Wu's group. Existential distraction.
3. 2021 Nasdaq IPO attempt — also failed.
4. Oct-Dec 2024 — Sophgo / TSMC cutoff → Entity List. The single biggest value-impairing event: severed the AI-chip growth leg's supply (Lens 10).
5. Jan-May 2025 — CBP customs holds + Trump tariffs (rigs hit ~21.6% import cost; weight imports −65% YoY).
6. Jul/Nov 2025 — Senate Intel report + DHS "Operation Red Sunset." Reframed Bitmain from "Chinese vendor" to "national-security threat with potential kill switches." Reportedly cited Bitmain rigs running near a Pentagon-serving Microsoft datacenter and a Wyoming Air Force missile base (NYT).
7. Late 2025 — hashprice crash + aggressive ASIC price cuts ($3-4/TH). The pattern: the market for this franchise reacts overwhelmingly to (a) the Bitcoin cycle/hashprice and (b) US-China policy — not to product cadence. Policy now dominates.

Phase C — Judge people & books

Lens 9 · Management

• Micree Zhan (Zhan Ketuan) — co-founder, controlling shareholder, chairman. Chip engineer (designed Bitmain's early ASICs). Track record: built the dominant Bitcoin-ASIC franchise from 2013; also founded Sophgo (2019, ~22% owned) to commercialize the Sophon AI chips. Skin in the game: maximal — largest holder, bought out his co-founder for ~$600M to take control. Capital allocation: mixed — the AI/Sophon diversification was strategically smart (hedge the mining cycle + a China-AI ban) but the Sophgo Huawei-cutout episode put the whole effort on the Entity List, a self-inflicted capital-destruction event.
• Jihan Wu — co-founder, EXITED. Resigned 2021, no stake as of April 2025; now runs Bitdeer (BTDR) — i.e. a Bitmain co-founder is now a competitor. Founder archetype: Bitmain is a founder-controlled company (Zhan), not professional-manager-run — which at this stage means concentrated decision-making and key-man risk on one engineer-founder who is also tied to the Sophgo problem.
• Irene Gao — global business chief, the public face of the 2025-2026 US-expansion push.
• Red flags: (1) the 2018 CICC-vs-KPMG income discrepancy ($1.25B claimed vs $701M audited) — a governance/disclosure stain; (2) related-party density — Antpool, BitFuFu, Sophgo, captive mining all orbit the same founder/holders, creating self-dealing surface (Lens 13); (3) years-long, litigated boardroom war — the opposite of stable governance; (4) founder's direct entanglement with the Sophgo sanctions matter.

Lens 10 · Forensic Red Flags

No audited statements since the lapsed 2018 prospectus, so forensic work is structural/qualitative — but the red flags here are unusually concrete for a private name:

• The 2018 prospectus discrepancy (the canonical flag): leaked CICC pre-IPO deck showed net income $1.25B vs KPMG-audited $701M — a ~44% overstatement that inflated implied P/E to Preferred-B/B+ investors. Any historical Bitmain profitability claim must be discounted accordingly.
• Inventory accounting risk (structural): Bitmain's core forensic risk is inventory valuation — it cannot slow next-gen production without losing TSMC allocation, so it builds into downturns and then writes down / fire-sells (S19/S21 at $3-4/TH in late 2025). A private company under no obligation to mark inventory honestly + a violently cyclical product = the highest-probability place for hidden losses.
• Related-party / revenue-recognition surface: sales into BitFuFu (framework for 80,000 miners; Bitmain/Antpool are BitFuFu PIPE investors) and into its own captive fleet can flatter "sales" — channel-stuffing into affiliated/controlled buyers is structurally possible and unverifiable from outside.
• US litigation (newly surfaced, 2025) — ``: court records across Orb Energy (TX, Ch.11; Bitmain claims 2,700 Antminers, alleges 99+ BTC diverted), Old Const (TN; 7,143 S21s, ~2.4 EH/s), MakerStar (MS; 6,400 miners, Bitmain awarded $4.5M), and Energy Conversion Group (IA; $4.935M, 5,000 "refurbished" miners alleged misrepresented as cleaned/verified) reveal ≥16,000 Bitmain-owned/controlled US miners and a litigious hosting footprint. The "refurbished miners misrepresented" claim is a product-integrity allegation worth tracking.

Regulatory findings (required sub-section):

• SEC EDGAR (LR + AAER): regulatory/regulatory-findings.md confirms 0 findings — Bitmain has no CIK and is not an SEC filer, so no EDGAR enforcement search is possible.
• Item 3 (Legal Proceedings): n/a — private, no 10-K exists. Never fabricated.
• Non-SEC enforcement / national-security (the material set, all ``):
• US Commerce — Entity List: affiliate Sophgo (Xiamen Sophgo / Xiamen Sophgo Technologies) added December 2024 after TechInsights found a TSMC-made 7nm die inside a Huawei Ascend 910B and identified Sophgo as the suspected cutout; TSMC had already suspended supply to Sophgo in Oct 2024. Sophgo denied a Huawei relationship; Bitmain distanced itself.
• DHS — "Operation Red Sunset" (2025): chip + firmware inspections of Bitmain rigs held at US ports, probing hidden kill switches / remote-access; NSC-level policy discussion spanning Biden→Trump administrations.
• Senate Intelligence Committee (Jul 2025 report): concluded Bitmain machines "can be forced by the PRC to turn over data" and have "the capability to be remotely controlled by Bitmain personnel in China"; flagged "several disturbing vulnerabilities".
• GAIN Act (NDAA amendment, 2025): Senate-passed; prioritizes domestic advanced-chip orders + mandatory export licenses for advanced ICs — projected to raise miner hardware costs and disrupt the Bitmain supply chain.
• Bitmain's response: "strictly complies with US and applicable laws… never engaged in activities that pose risks to US national security"; remote-control claims "unequivocally false".

Bottom line: This is the most regulatorily-exposed name in the crypto-hardware census — the findings are not hypothetical, they are active federal investigations + an existing Entity List designation on the affiliate.

Phase D — Project & stress-test

Lens 11 · IPO-readiness & path-to-tradeable (+private swap — replaces "Forward Projection")

research/private-watch.json does not exist in the research repo (checked MENFEM_RESEARCH_DIR=../menfem-research) — so IPO-readiness is web-grounded, not ledger-grounded. No forecast.ts create (watchlist rule). No dossier-path write-back possible (no private-watch.json entry to update — flagged below as an open item).

Readiness read:

• Stage: mature, profitable-at-cycle-peak monopoly-class private — operationally IPO-able for years (it nearly listed twice).
• IPO readiness: LOW and falling — for any Western exchange. Two failed attempts (HK 2018, Nasdaq 2021). The decisive blocker is now geopolitical, not financial: an Entity-Listed affiliate (Sophgo), an active DHS national-security probe, a Senate Intel report alleging PRC remote-control, and CBP holds make a US listing effectively impossible in the current climate. A Hong Kong relisting is the only plausible venue, and even that is gated by the US-China tape and by getting clean audited financials past the same disclosure scrutiny that sank 2018.
• Catalysts that would unlock an S-1 / tradeable event: (1) resolution/closure of Operation Red Sunset with no adverse finding (low probability of clean resolution near-term); (2) a clean separation from Sophgo + restored TSMC access; (3) a sustained hashprice recovery to refill the order book; (4) a HK (not US) listing track. Estimated window: not within 12 months; 2027+ at the earliest, venue almost certainly Hong Kong. ``.
• Path-to-tradeable for a Western investor today: essentially none that is cleanly investable — secondary shares (UpMarket/EquityZen/Hiive at ~$18.78B) carry both private-market illiquidity and US national-security/sanctions tail risk. The cleanest liquid expressions of the Bitmain thesis are its public peers/competitors — Canaan (CAN), Bitdeer (BTDR) — and its customer base of US miners, not Bitmain itself.

(Open item for the ledger: add a bitmain entry to research/private-watch.json with stage, ipo_readiness: low, catalyst, and this dossier path so privates.ts shows it dossier-warm — could not write-back as the file is absent.)

Lens 12 · Bull vs Bear

Bull case. Bitmain is a structural monopoly on the picks-and-shovels of a $2T+ asset. 35-40% of new ASIC sales, 23 of the top-50 rigs, ~95-98% oligopoly with two friends, the deepest TSMC relationship in the category, a recurring Antpool fee stream, a captive-mining release valve, and an optional AI second act (Sophon). The Bitcoin network hashrate hit all-time highs (~1.15 ZH/s, Oct 2025) — every exahash is its (or a rival's) silicon. At ~$18.78B on ~$2.5B peak revenue it is not obviously expensive for a category monopolist if you ignore the flag risk. The US-factory pivot, if it works, converts the biggest bear point (import/tariff/security risk) into a moat-deepener ("Made-in-USA Antminers" that clear CBP and qualify for US miners).

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

1. Western de-platforming (the kill-shot). Entity-List precedent (Sophgo), an active DHS probe, a Senate Intel "remote-control" finding, and the GAIN Act could legally lock Bitmain out of the US market — its single largest end-market — via import bans or a connected-equipment rule (the Huawei/telecom-gear playbook applied to ASICs). This is permanent impairment of the demand base, not a cycle.
2. Foundry severance. The whole moat rests on TSMC leading-edge access, which the US can pressure (it already cut Sophgo). If Bitmain loses N5/N3 allocation — or is forced one-or-two nodes behind via licensing — MicroBT (already on 3nm) and US-domestic Auradine eat share.
3. Cycle + governance. Hashprice at 2023 lows, fire-sale pricing, chronic inventory overhang, a founder-controlled register with a documented disclosure discrepancy and a litigious US footprint.

Pre-mortem (18 months out, thesis broke): The US issued a connected-equipment / import rule barring new Bitmain ASIC deployments on US soil (Huawei-style); US miners pivoted to MicroBT-via-Malaysia, Auradine, and Bitdeer's SEALMINER; TSMC quietly throttled Bitmain's leading-edge allocation under licensing pressure; the late-2025 fire-sale revealed the inventory hole; and the HK relisting stalled on audited-financial scrutiny. The ~$18.78B secondary mark proved a stale 2026 print with no buyers.

Are multiples too high? Un-assessable (private, no public multiple) — but the secondary mark embeds zero discount for an active national-security probe, which is itself the contrarian signal.

Contrarian view (what the market refuses to see): Bulls treat Bitmain as a cyclical hardware monopoly to be valued on the Bitcoin cycle. The market is mispricing it as a geopolitics asset — its value is now a function of US-China policy, not hashprice. The US factory is a tell that even Bitmain knows its China-made moat is being condemned; the question is whether "Made-in-USA Antminer" is a credible firewall or a fig leaf over an entity Washington has already decided is a threat.

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

Dismantling the bull case:

• Revenue is concentrated in exactly the market that is trying to ban it. The US is the largest end-market and the source of the existential risk. A connected-equipment rule (DHS/Senate Intel are laying the predicate) zeroes the most valuable demand pool overnight. Name the danger bulls underestimate: Auradine (Silicon Valley, explicitly pitching US miners a "non-foreign-ASIC" alternative) + Bitdeer's SEALMINER — both purpose-built to harvest exactly the share a Bitmain ban frees up.
• The moat is owned by someone else. Bitmain doesn't own a fab; it rents the world's most contested one. "Owning" 35-40% share means nothing if TSMC is pressured to throttle you (Sophgo proves Washington will pull that lever) and you're already a node behind MicroBT.
• The worst capital-allocation / governance moves are real, not hypothetical: the Sophgo Huawei-cutout that earned an Entity-List designation; the $1.25B-vs-$701M prospectus discrepancy; a multi-year litigated boardroom war; a dense related-party web (Antpool + BitFuFu + captive mining + Sophgo all founder-adjacent) that makes "sales" partly sales-to-self; and a litigious US hosting footprint with a "refurbished miners misrepresented" claim.
• The vertical-integration conflict: Bitmain competes with its own customers — it self-mines ≥16,000 rigs in the US and runs Antpool. In a hardware shortage it can prioritize its own fleet; customers know this, and it caps trust/pricing power.
• Assumptions that must hold for the ~$18.78B mark: continued US market access, continued TSMC leading-edge allocation, a hashprice recovery, and a viable listing venue. If hardware demand disappoints 20-30% (entirely plausible at 2023-low hashprice + a US import rule), the franchise's earnings power and the secondary mark both crater — there's no recurring SaaS floor, just pool fees.
• Single scenario that permanently impairs the business: a US connected-equipment/import ban on Bitmain ASICs (Huawei-telecom playbook). Plausibility: moderate-to-high and rising given an Entity-Listed affiliate + active DHS probe + bipartisan Senate Intel findings. That is the short.

Lens 14 · Management Questions (ordered by information value)

1. With Sophgo on the Entity List and Operation Red Sunset active, what is your concrete, documented contingency if the US bars import/deployment of Bitmain ASICs — and what % of revenue is US end-market today?
2. Will the US factory perform genuine chip-level manufacturing/assembly that satisfies a "connected-equipment" or domestic-content rule, or is it final assembly that a determined regulator can still treat as Chinese-origin?
3. What is your current, audited TSMC leading-edge (N5/N3) wafer allocation, and is any of it contingent on conditions tied to the Sophgo matter?
4. Will you publish independently audited financials (revenue, net income, inventory, cash) — and reconcile the 2018 CICC-deck vs KPMG net-income discrepancy?
5. What is the carrying value and write-down history of Antminer inventory through the 2022 and 2025 downturns, and how is it marked today?
6. What share of "hardware revenue" is to related/affiliated/controlled buyers (BitFuFu, Antpool-linked entities, your own captive fleet) versus independent third parties?
7. Precisely what remote-management, firmware-update, and telemetry capabilities exist in Antminer firmware, and can a customer fully air-gap a unit?
8. What is the firewall between Bitmain, Sophgo, and Micree Zhan's other vehicles — board, IP, supply, and capital?
9. How do you reconcile competing with your customers (self-mining ≥16,000 US rigs + Antpool) with being their primary supplier?
10. Given MicroBT at 3nm and Samsung-GAA trials, what is your node roadmap and timeline to regain efficiency leadership?
11. What is the real venue and timeline for a listing, and what specifically blocks it today?
12. How exposed is the order book to a sustained sub-$40/PH/day hashprice, and what is your breakeven utilization?
13. What is the resolution status of the US hosting lawsuits (Orb, Old Const, MakerStar, Energy Conversion Group), and is "refurbished as new" a systemic practice?
14. What is current cash, debt, and burn through this downcycle?
15. What is the realistic revenue ceiling for Sophon AI chips if TSMC access is never restored?