Xiaomi

Phase A — Understand the business

Company Overview

Xiaomi is a Chinese consumer-technology conglomerate built on a deliberate "hardware-at-cost, monetise-the-user" flywheel, now bolting a from-scratch EVEnterprise valueWhat it would cost to buy the whole company: its market value plus its debt, minus the cash you would get with it. Often very different from market cap. business onto the top. Founded 2010 by Lei Jun; IPO'd on HKEX in July 2018 as the exchange's first weighted-voting-rights listing. Global monthly active users reached 750M at end-2025, +7.4% YoY — the installed base that the model exists to harvest.

Four businesses, reported as two segments:

• Smartphone × AIoT (the legacy engine):
• Smartphones — FY2025 revenue RMB 186.4B (−2.8% YoY), 165.2M units, 13.3% global share, #3 for the 5th straight year.
• IoT & lifestyle (appliances, wearables, TVs, the "Mi ecosystem") — FY2025 RMB 123.2B (+18.3%), record 23.1% gross margin. The quiet compounder.
• Internet services (advertising, app store, fintech, cloud) — FY2025 RMB 37.4B (+9.7%) at a 76.5% gross margin. This is the profit annuity — an Apple-Services-lite riding on 750M MAU.
• Smart EV, AI & new initiatives — FY2025 RMB 106.1B (+223.8%), of which EV RMB 103.3B; 411,082 vehicles delivered (+200.4%). Cars: SU7 sedan (vs Model 3) and YU7 SUV (vs Model Y), average selling price >RMB 260k — a genuinely premium mix, not a cheap-car play.

Contract structure: almost entirely transactional hardware sales (no take-or-pay), direct-to-consumer via mi.com + Mi Home stores + carriers; EVs sold direct. The recurring, high-margin layer is Internet services — the only annuity in the model, and structurally the most valuable dollar of revenue Xiaomi earns.

FY2025 group: revenue RMB 457.29B (+24.97%) — first year over RMB 400B; statutory net income RMB 41.64B (+76%); adjusted net profit RMB 39.2B (+44%); R&D RMB 33.1B.

Supply Chain

Names or it didn't happen — the chain, upstream → Xiaomi → customer:

Phones / AIoT (upstream):

• SoCs — Qualcomm (Snapdragon flagship), MediaTek (mid-range), and increasingly Xiaomi's in-house XRING O1 3nm SoC (2025), fabbed at TSMC — a real chokepoint (leading-edge foundry is single-source-equivalent).
• Memory — Samsung, SK Hynix, Micron. This is the acute 2026 chokepoint (see Lens 5): the DRAM/NAND oligopoly is reallocating capacity to AI HBMHigh-bandwidth memoryMemory stacked in layers beside a processor so data reaches it faster. The usual bottleneck in AI chips., and Xiaomi is a price-taker.
• Displays — Samsung Display, BOE, TCL CSOT; camera sensors — Sony, OmniVision.

EV (upstream):

• Batteries — CATL (NMC "Qilin" cell-to-pack, 101 kWh, 800V, for SU7 Max/Ultra) and BYD FinDreams (73.6 kWh LFP blade, 400V, base SU7); the entry SU7 is dual-sourced CATL/BYD to lift capacity.
• Cabin compute — Qualcomm Snapdragon 8295. Motors (HyperEngine) and die-casting are increasingly in-house at Xiaomi's self-built Beijing gigafactory (a deliberate break from the old BAIC contract-manufacturing route).

Downstream: direct retail + online; no customer concentration on the buy side (mass consumer).

Chokepoints ranked: (1) memory (3-supplier oligopoly, AI-starved — live margin damage now); (2) leading-edge foundry (TSMC for XRING); (3) power-cell supply (CATL bargaining power). None are single-point-fatal, but memory is actively compressing earnings today.

Competitive Advantages (moats)

• Brand + price-performance — Xiaomi owns the "flagship specs at 70% of the price" position globally; the moat is cost discipline + scale purchasing, not proprietary technology. Durable in emerging markets, thinner in premium.
• Ecosystem lock-in (AIoT) — the widest smart-home hardware graph in China; switching cost rises with each connected device. Real but soft — consumers multi-home.
• Internet-services annuity — 750M MAU at 76.5% GPM is the closest thing Xiaomi has to a true economic moat: a captive, monetisable audience that phones/IoT feed for free.
• EV demand pull — the scarcest asset in autos: YU7 took 289,000 orders in its first hour, 240,000 locked in 18 hours, waits >10 months. This is Apple-launch-grade pull demand, and it is new moat forming in real time — brand equity transferring from phones to cars.
• Bargaining power: strong over most component suppliers (volume), weak over the memory oligopoly and CATL. Strong over customers in EV (waitlist), weak in phones (commoditised).

Net: the moat is brand + ecosystem + cost engine, now compounded by an EV demand signal most legacy automakers would kill for. It is not a technology moat, and it does not protect margins against exogenous input shocks.

Segments

FY2025 revenue by segment, with trend:

Geographic split not cleanly sourced at segment granularity (n/a), but overseas is a rising share of both phones and IoT; EV is essentially 100% China today (global launch deferred to ~2027).

The mix shift IS the story: Xiaomi is transitioning from a phone company (declining) with an IoT+Services annuity (compounding) into an EV-led growth company — and the market is now pricing the EV leg's execution risk far more than the annuity's stability.

---

Phase B — Measure performance

Earnings Result (Q1 2026, reported ~2026-05-26)

• Revenue RMB 99.1B; adjusted net profit RMB 6.1B; group gross margin 22.0%.
• Smartphone × AIoT RMB 79.3B, GPM 22.5%. Smartphone rev RMB 44.3B (44.7% of group), 33.79M units, ASP a record RMB 1,310 (+8.2% YoY) — deliberate premiumisation. But IDC put Q1'26 units at 33.8M, −19.1% YoY (11.5% share, still #3): Xiaomi is cutting low-end volume on purpose to avoid passing through the memory-cost spike — trading units for margin defence.
• IoT RMB 24.7B, GPM 25.2% (still expanding).
• Smart EV/AI/new RMB 19.9B, GPM 20.1%, operating LOSS −RMB 3.1B (~−$460M, ~−$5,600/car); 80,856 vehicles delivered. Management attributed the reversal to: EV purchase-tax subsidy roll-off, a lower mix of high-margin SU7 Ultra, and rising core-component costs.
• Balance-sheet flags: none acute — net cash, ~0.0 net-debt/EBITDA. FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. historically robust (FY2024 OCF RMB 39.3B, capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. RMB 7.3B, FCF RMB 32.0B); EV capex now rising.
• Vs the company's own history: the single most important line is that the EV segment swung back to a RMB 3.1B operating loss after achieving full-year positive OI (+RMB 0.9B) in 2025 and a first quarterly profit in Q3'25. That reversal — not the phone number — is why the stock is where it is.

Earnings Calls (sentiment trend)

No transcripts on the shelf (web-only). From secondary coverage of the last ~4 calls:

• Tone arc: peak-triumphalist through Q2–Q3 2025 (record revenue, EV profitability, YU7 order records) → defensive/qualifying by Q4'25 and Q1'26 (smartphone margin warnings, EV-margin caveats, explicit talk of price hikes and spec downgrades from memory costs).
• Recurring phrases now: "structural indicators," "operational quality," "premiumisation," "high-end breakthrough," "human × car × home" ecosystem.
• What they stopped saying: the confident EV-margin-expansion narrative of mid-2025 has been replaced by hedged language about "product transition" and "component costs." Management is managing expectations down on near-term EV profitability while holding the long-term volume story.

Comps

Peer table — all multiples `` with source/date, or n/a. Xiaomi is a hybrid (electronics + EV), so it fits no single peer set cleanly; shown against both.

EV/Sales, EV/EBIT, dividend yield, 5-yr avg ROE: n/a — not cleanly sourced for a like-for-like set (do not fabricate). The load-bearing read: Xiaomi trades below profitable-EV peer BYD (26x) and far below Apple (38x), roughly in line with pure-hardware Samsung (14x) — i.e. the market is currently valuing Xiaomi as a hardware company with an EV problem, giving ~zero credit to the EV franchise or the services annuity. That is the whole debate.

Stock-Price Catalysts (5-yr, moves >5%)

The tape trades the EV narrative far more than phones/IoT:

• Jan 2021 — US DoD "Chinese military company" blacklist (−); May 2021 — removed after Xiaomi won in federal court (+).
• 2021 — EV entry announced, Lei Jun pledges $10B ("last major entrepreneurship of my life").
• Mar 2024 — SU7 launch: stock +16% intraday, closed +9%, market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. passed GM and Ford (~$55B at HK$17.34).
• 2024 → H1 2025 — EV ramp + first profitability → re-rating to HK$59.90 peak (~$190B, P/E 40–52x).
• Jun 2025 — YU7 launch, 289k orders/hr (+, demand proof).
• Mar 2025 — fatal SU7 crash, Anhui, 3 deaths; ADAS scrutiny (−). Oct 2025 — Chengdu SU7 Ultra fire, electronic doors failed to open; stock −5.7% (−).
• Q4'25 → 2026 — recalls (116,877 SU7 for ADAS OTA; pressure to recall ~370,000 for door handles), Q1'26 EV back to −RMB 3.1B, phone units −19% on memory costs, analyst downgrades → 52-week low HK$21.30, now HK$25.84.

Pattern: the market reacts to (1) EV deliveries/launches, (2) EV safety events, and (3) EV margin — in that order. Phones and IoT barely move the tape now. Safety and EV margin are the new swing factors; a clean safety resolution and an EV-margin turn are the two catalysts that re-rate the stock.

---

Phase C — Judge people & books

Management

• Lei Jun (Chairman/CEO/founder) — a genuinely elite operator. Serial founder: ran Kingsoft to IPO, sold Joyo.com to Amazon (2004), angel-backed YY/UCWeb, founded Xiaomi 2010, took it to #3 in phones and stood up a from-scratch EV business that hit 400k units and a profit inside two years. Few executives on earth have that range. Vice-Chair Lin Bin.
• Skin in the game: ~24% economic ownership, but ~65% of the vote via the WVR structure (founder super-vote class). Founder-controlled and deeply invested — aligned on upside, entrenched against minority discipline.
• Capital allocation: aggressive buybacks — a HK$20B repurchase authorisation (post-2026 AGM), ~HK$14.6B executed under the prior scheme, plus a fresh HK$4B automatic tranche; heavy EV reinvestment (Beijing Phase II/III + Wuhan plant → 1.2M units capacity by end-2026); negligible dividend. The buyback-into-weakness is a real conviction signal.
• Red flags: (1) WVR entrenchment — minorities cannot force change; (2) promotional style — Nürburgring records, order-number theatrics, keynote hype — which collided with physical-world safety in the fatal-crash episodes and the ADAS marketing scrutiny; (3) the "ship-fast, patch-OTA" software culture applied to a 2-ton object is a governance question, not just an engineering one.
• Archetype: visionary founder-operator at the riskiest possible stage — scaling a safety-critical hardware business at software speed. The thing that makes Xiaomi great (speed, ambition) is the thing that created its biggest tail risk.

Forensic Red Flags

Accounting quality appears clean relative to the sector — no restatements or aggressive-recognition flags surfaced. Watch-items, labelled:

• Segment margin optics — EV gross margin (24.3% FY25) was flattered by SU7 Ultra mix and purchase-tax subsidies, both of which reversed in Q1'26. Read EV operating income, not gross margin.
• Adjusted vs statutory — adjusted net profit (RMB 39.2B) strips SBC and fair-value swings; statutory NI (RMB 41.64B) was higher in FY25 (investment gains), so the adjustment is not obviously flattering — but track the gap.
• Capex ramp — EV plant build (Phase III + Wuhan) will pressure FCF; watch inventory/receivables vs delivery cadence as volume scales.
• Warranty/recall reserves — the live question: are reserves adequate for the ADAS recall (116,877 units) and a potential ~370,000-unit door-handle recall? Under-reserving is the forensic risk to watch.

Regulatory findings (per regulatory/regulatory-findings.md + web):

• SEC/EDGAR: none possible — Xiaomi has no CIK and is not an SEC filer. total_sec_findings: 0.
• US DoD: blacklisted as a "Communist Chinese Military Company" Jan 2021; Xiaomi sued and won — the federal court found the DoD's evidence "insufficient"; removed May 25, 2021. Tail risk: re-designation under any future US-China escalation is a live, unquantifiable overhang for a China-flagship name.
• India (ED): ~US$676M (₹55.51B) frozen since 2022 under FEMA over "royalty" remittances; Xiaomi lost its appeals and the funds remain frozen; a separate ~$72M royalty-tariff dispute is at India's Supreme Court. Material, ring-fenced, and a read on China-tech operating risk in India.
• China auto-safety: SU7 recall of 116,877 units (ADAS OTA fix); regulatory pressure to recall ~370,000 for door-handle design; new national standard GB 48001-2026 mandates mechanical door releases from Jan 1, 2027; ongoing scrutiny of ADAS marketing after two fatal crashes (Anhui Mar'25, Chengdu Oct'25). Product-liability and reputational exposure here is the single most important non-financial risk.
• Verified via SEC EDGAR EFTS (no CIK → n/a), web search, and secondary coverage as of 2026-07-10. No accounting-fraud findings.

---

Phase D — Project & stress-test

Forward Projection (base / bull / bear)

Built bottom-up from FY2025 actuals + guidance; FY = calendar year; current FY = FY2026. Every input -anchored, output . No our model logged (unattended watchlist run).

Anchors: FY25 statutory NI RMB 41.64B, EPS ~RMB 1.50, ~25.69B shares; consensus FY26 revenue RMB 501.1B, EPS RMB 1.13 (cut from 1.56); forward P/E 20.7x > trailing 16.6x → the market expects 2026 earnings to fall.

• FY2026 revenue build: Smartphone×AIoT ~RMB 365–375B (phone units ↓ but ASP ↑ + IoT +~15%); EV ≈ 550,000 units × ~RMB 250k ASP ≈ RMB 137.5B + other ~RMB 4B → Smart-EV/new ~RMB 140B. Group ≈ RMB 505–515B — in line with consensus RMB 501B. ✓ cross-check.
• FY2026 EPS:
• Base ≈ RMB 1.15. Matches consensus RMB 1.13. At HK$25.84 (~RMB 23.7) → ~20.6x (ties to sourced fwd P/E 20.7x ✓).
• Bull ≈ RMB 1.40.
• Bear ≈ RMB 0.85.
• FY2027 base ≈ RMB 1.60 → ~14.8x.
• FY2028 base ≈ RMB 2.10 → ~11.3x.

The math of the thesis: on trailing 2025 it's 16.6x; on 2026 base it's ~21x (E is falling); on a normalised, EV-profitable 2027–28 it's ~11–15x. So the stock is only "cheap" if you believe the EV segment margin turns durably positive within ~18 months. That belief IS the investment.

Bull vs Bear

Bull case. The scarcest asset in autos is genuine pull demand, and Xiaomi has it — 289k YU7 orders in an hour, >10-month waits, at a >RMB 260k ASP. Capacity goes from ~60k/month to 1.2M units/year by end-2026; the order book is the de-risking. Layer on the Internet-services annuity (RMB 37.4B at 76.5% GPM on 750M MAU — a hidden Apple-Services the market is valuing at zero), record-margin IoT (+18%), a net-cash balance sheet, buybacks into the drawdown, and a founder-operator who has executed twice. If EV margin normalises as the subsidy/mix headwinds lap and Wuhan scale kicks in, 2027 EPS inflects and a re-rate from 16x toward the mid-20s BYD multiple implies the analyst HK$40 target (+55%).

Bear case (permanent-impairment vectors). (1) Safety/brand — the door-handle and ADAS failures expose a software-first culture shipping a safety-critical product; a scaled recall, a liability verdict, or a crack in Chinese consumer trust in a hyper-competitive EV market could permanently impair the EV franchise's premium pricing. (2) EV margin structural, not transitory — if the 2025 profit was a subsidy/mix mirage and the China price war caps pricing, EV stays a cash drain and the "cheap 16x" is a value trap. (3) Smartphone in secular decline with an exogenous memory tax (BOM from 10–15% → 30–40% of cost) Xiaomi cannot control — it already cut its 2026 phone forecast by 70M units.

Pre-mortem (18 months out, thesis broke): a second fatal-crash cluster or a forced 370k-unit recall craters SU7/YU7 order velocity; EV margin stays negative through 2026 into 2027; memory costs stay elevated, forcing phone price hikes that bleed share; consensus 2026 EPS RMB 1.13 proves optimistic; the stock re-tests HK$21 and the "value" call dies because E fell faster than the multiple.

Are multiples too high? No — 16.6x trailing on a 25% grower is not demanding; the risk is the denominator falling, not the multiple.

Contrarian view (what the market refuses to see): the market is valuing Xiaomi as "a hardware company with an EV problem" (Samsung-like 14–16x) and giving zero credit to the EV demand franchise or the services annuity. If EV margin merely returns to breakeven-plus — not heroics — the re-rate is large. The bear needs EV to stay loss-making; the bull only needs it to stop bleeding.

Devil's Advocate (short-seller)

• Where the money is really made: the EV gross margin that bulls cite (24.3%) was propped by SU7 Ultra mix + purchase-tax subsidies — both already reversed (Q1'26: −RMB 3.1B OI). Strip the subsidy and the "profitable EV maker" narrative is one quarter old and already broken.
• Concentration risk: EV is ~100% China, in the world's most brutal price war (BYD, and dozens of others). One aggressive BYD price cut and Xiaomi's premium ASP compresses.
• The moat is thinner than it looks: phone "moat" is cost, not tech — and an exogenous memory shock just proved Xiaomi is a price-taker with no pricing power in mid-range. The EV "moat" is brand halo + a novelty order book that has never been tested through a safety scandal at scale.
• Most dangerous competitor bulls underestimate: not Tesla — BYD, which owns the cost curve, the battery (FinDreams supplies Xiaomi), and can out-price the SU7 indefinitely.
• Worst governance: WVR entrenchment (65% vote on 24% economics) + promotional theatrics that materially contributed to the ADAS-marketing safety scrutiny. The order-number PR machine cuts both ways.
• What must hold for today's price: EV margin turns durably positive by 2027, no scaled recall/liability event, memory costs ease, and China EV pricing stabilises. Break any one and the RMB 1.13 EPS is too high.
• If growth disappoints 20–30%: FY26 EPS to ~RMB 0.85 (bear) → the "16x" becomes ~30x on falling earnings → stock re-tests the HK$21 low.
• Single permanent-impairment scenario: a second high-profile fatal-crash cluster + regulator-forced large recall that breaks Chinese consumer trust in Xiaomi's premium-EV safety story. Plausibility: low-to-moderate but non-trivial given two fatal incidents already and a national door-handle standard written partly in response.

Management Questions (ordered by information value)

1. Was FY2025 EV operating profit (+RMB 0.9B) structurally repeatable, or a one-off of SU7 Ultra mix and purchase-tax subsidies now reversing? What is the steady-state EV operating margin at 1M+ units?
2. What is the total reserved and expected cost of the ADAS recall and a potential ~370,000-unit door-handle recall, and are current warranty reserves adequate?
3. At what monthly EV volume does the segment reach sustained positive operating income, and when do you expect to cross it?
4. How much of the Q1'26 phone margin defence (units −19%, ASP +8%) is repeatable vs a one-time response to the memory spike — and what is your memory-cost hedge into 2027?
5. What is the plan if BYD prices aggressively into the SU7/YU7 segments — do you defend ASP or volume?
6. What is the current net-cash position and the peak cumulative EV cash burnCash burnHow much more cash goes out than comes in, per period. The clock on a company with no profits. before self-funding?
7. How do you reconcile the "ship-fast, OTA-patch" software culture with the safety governance a 2-ton product demands post-Chengdu?
8. What is the realistic timeline and margin profile for overseas EV (deferred to ~2027), and how do you clear US/EU regulatory and tariff barriers?
9. What is the durable growth and margin runwayRunwayHow long the cash lasts at the current rate of spending. It shortens the moment spending rises, which is why a figure taken from a quiet quarter flatters. for Internet services (76.5% GPM) as MAU growth slows to ~7%?
10. How dependent is the XRING SoC roadmap on TSMC leading-edge capacity, and what is the contingency under export-control escalation?
11. What is the capital-allocation priority order among EV capex, buybacks, and R&D over the next three years?
12. What is CyberOne's honest commercial timeline — and is it a real business line or an R&D halo?
13. What re-designation risk do you carry from the 2021 US DoD episode, and how are you hardening against it?
14. What is the resolution path and P&L exposure for the frozen ~US$676M in India?
15. What KPI should investors track quarterly to know the EV-margin thesis is working — and what number would tell you it's not?

---