Geek+ (Geekplus)

Phase A — Understand the business

Lens 1 · Company Overview

Geek+ (legal: Beijing Geekplus Technology Co., Ltd.; brand "Geekplus") builds autonomous mobile robots (AMRs) and the AI/software that orchestrates them for warehouse and supply-chain automation. Founded 2015 in Beijing by Yong Zheng (CEO) with Hongbo Li, Xi Chen and Kai Liu. The core product is goods-to-person fulfillment: instead of workers walking the aisles, robots bring shelves, totes, cases or pallets to a stationary picking station. Product families span shelf-to-person (the original "Roboshuttle"/P-series moving-shelf robots), tote-to-person (RoboShuttle, incl. the 2026 V5 with an embedded robot-arm picking station), pallet-to-person, sortation, and — new in 2026 — a general-purpose humanoid, "Gino 1," and an embodied-AI layer branded "Geekplus Brain".

How it makes money: primarily project sales — a customer buys a fleet + the warehouse-management/robot-orchestration software as an integrated solution, installed and commissioned. Increasingly layered on top is a subscription/RaaS motion (robots-as-a-service): management flagged subscription-based service orders up >90% YoY in 2025 and calls it "an important highlight of the order structure". That mix shift matters for Lens 4/11 (recurring, higher-margin revenue vs. lumpy capex sales).

Scale (cumulative, end-2025): >72,000 robots delivered to >40 countries, ~950 end customers including >80 Fortune Global 500 companies; large-customer repurchase rate 78%. Customers cluster in e-commerce, FMCG/retail, 3PL, apparel and now food & beverage (a 2025 new-vertical breakthrough). This is not a science project — it is a shipping, revenue-generating category leader.

Main suppliers: motor/drive, battery (LFP cells), lidar/vision sensors, compute (edge SoCs), and steel/chassis — sourced largely from the Chinese robotics supply chain (see Lens 2). Main competitors: Hai Robotics (Shenzhen, case-handling/ACR specialist — the most dangerous), Locus Robotics (US, goods-to-person), GreyOrange, Exotec (FR, Skypod ASRS-hybrid), AutoStore (NO, cube storage — adjacent), plus the industrial-automation majors (Dematic/KION, Swisslog/KUKA, Honeywell Intelligrated) as integrators.

Lens 2 · Supply Chain

Upstream inputs → Geek+ → end customer, named where sourced:

• Semiconductors / compute: edge SoCs and controllers. Intel is both a historical strategic investor (Intel Capital, ~0.63% stake) and an ecosystem partner. NVIDIA Jetson-class edge compute is the industry-standard AMR brain; specific Geek+ BOM not disclosed → n/a.
• Sensors: lidar (the dominant Chinese suppliers are Hesai and RoboSense), plus cameras/2D-3D vision. Specific vendor not disclosed → n/a, but the Chinese lidar cost curve is a structural tailwind for a China-based AMR maker.
• Batteries: LFP cells — China is the global cost/volume leader (CATL/BYD-class supply). Geek+ benefits from the same battery deflation as EVs.
• Actuation / motors / chassis: Chinese motor and precision-mechanics supply base; steel.
• Geek+ itself: designs the robots + writes the orchestration software/WMS + the new "Geekplus Brain" embodied-AI stack; assembles/integrates and does on-site commissioning.
• Channel → end customer: direct enterprise sales + a systems-integrator channel (Geek+ often sits underneath a Dematic/Swisslog-style integrator or a regional SI in a larger warehouse build). End buyers: e-commerce, 3PL, FMCG, retail, apparel, F&B operators; named marquee logos are guarded, but the customer base includes >80 Fortune Global 500 names.

Chokepoints / single-source risk: (1) Edge compute — if a US export-control regime tightens on advanced edge AI accelerators to Chinese robotics firms (a real tail risk given the "embodied intelligence" pivot toward heavier on-robot inference), the humanoid roadmap is most exposed. (2) The install/commissioning bottleneck is human, not silicon — deploying a fleet in a live warehouse is services-heavy and hard to scale across 40 countries without local integration partners; this is the operational chokepoint that caps growth and pressures margin. Grounding: robotics commercial-layer wiki files (kb/robotics/wiki/*) are empty on disk — this map is web-derived, not compiled.

Lens 3 · Competitive Advantages (moats)

What's genuinely defensible:

1. Market-share scale + reference base. #1 global warehouse-AMR share for 7 consecutive years (Interact Analysis 2025 report); ~9% of the total warehouse-fulfillment AMR market, and by the company's order-fulfillment cut, ~23% of global order-fulfillment mobile robots (≈ the #2 + #3 combined) with ~48.5% of shelf-to-person. >72,000 robots in the field is a data + reliability moat: fleet telemetry trains the orchestration software, and a 78% repurchase rate says the installed base expands land-and-expand rather than churning.
2. Software/orchestration lock-in + switching costs. Once a warehouse is designed around Geek+ robots + its WMS/orchestration layer, ripping it out is a multi-quarter, operationally-risky project. Switching cost is the real moat in this category, more than the hardware.
3. China cost structure exported globally. Building on the Chinese robotics/battery/lidar supply chain while selling into higher-price DM markets is why overseas gross margin is 46.6% vs. a 35.5% blended. That geographic arbitrage is a durable structural edge over Western pure-plays (Locus, Exotec) that build at Western cost.
4. Full-category breadth. Geek+ claims to be the first vendor spanning mobile robots + dedicated arm-picking stations + a general-purpose humanoid under one platform ("Geekplus Brain"). If embodied AI actually collapses these into one orchestration brain, breadth becomes a moat; today it's mostly a marketing claim.

Where the moat is thin:

• Hardware commoditizes. AMRs are increasingly a reference-design + integration game; Chinese peers (Hai Robotics, Quicktron, Libiao) attack on price. Bargaining power over large customers is weak — Fortune-500 buyers dual-source and squeeze.
• Bargaining power: stronger over fragmented Chinese suppliers than over concentrated DM enterprise buyers. Net: Geek+ needs the customer more than the customer needs Geek+ on any single deal, which is why margin, not growth, is the swing variable.

Lens 4 · Segments

Hard limit: no segments.csv on disk (header-only). All figures ``; a clean product-line revenue split is NOT publicly broken out — flagged where missing.

By geography (the disclosed, high-conviction cut):

The geography story is the whole story: a Chinese company that earns three-quarters of revenue outside China, at a materially higher margin abroad. That is unusual for a Chinese hardware name and is the core bull fact. Regional momentum: Americas new orders +50% YoY, ex-China new orders +~40% YoY in 2025.

By product line: Geek+ does not publish a clean revenue split across shelf-to-person / tote-to-person / pallet / sortation / software-subscription → n/a. What is disclosed: subscription-service orders +>90% YoY (order value, not revenue) and new-vertical wins in F&B; the mix is shifting toward higher-margin recurring, which is the right direction. By order structure: FY2025 new orders RMB 4.137bn (+31.7%) — orders growing in line with revenue, a healthy 1.3x book-to-bill-ish signal (orders > revenue).

Trend read: accelerating overseas + margin-up + recurring-up = the highest-quality version of this business the company has ever reported. The deceleration risk is that 31% growth is down from a 45% 2021-24 revenue CAGR / >90% AMR-segment CAGR — the law of large numbers is biting, and the market has noticed (Lens 8/12).

---

Phase B — Measure performance

Lens 5 · Earnings Result (FY2025 annual + H1'25 — via the +private funding/inflection overlay)

FY2025 (year ended Dec 2025), reported 2026-03-31 [all web: PR Newswire]:

• Revenue RMB 3.171bn, +31.6% YoY (FY2024: RMB 2.409bn).
• Gross profit RMB 1.125bn, +34.4%; gross margin 35.5% (margin expanding — GP grew faster than revenue).
• Adjusted net profit RMB 43.8m — the first profitable year in company history (non-IFRS).
• Operating cash flow +RMB 85.7m — first-ever positive OCF.
• New orders RMB 4.137bn, +31.7%.
• Overseas 75.3% of revenue at 46.6% GM.

The critical caveat — GAAP vs. adjusted. The RMB 43.8m profit is adjusted (non-IFRS). On a statutory IFRS basis the company guided FY2025 net loss to RMB 10-30m — i.e. essentially breakeven, still a small loss, a 96-99% YoY improvement. The gap between "first profitable year" (adjusted) and "still a small GAAP loss" is the single most important number-integrity point in this dossier (see Lens 10).

Net-loss history (IFRS) — the burn that got them here:

H1'25 (for the trend): revenue RMB 1.025bn (+31.0% vs RMB 782m H1'24); GP RMB 360m (+43.1%); GM 35.1%; adjusted EBITDA +RMB 11.6m (first positive, vs −RMB 169.8m H1'24); adjusted net loss −RMB 11.9m (narrowed 94%). The H1→FY progression shows the profit inflection landing in 2H.

Balance sheet flags: exact post-IPO cash on hand → n/a cleanly (a pre-IPO snapshot showed only ~RMB 387m cash and negative net current assets of ~RMB(6.45)bn as of Oct 2024, driven by the preferred-share liability that converts to equity on IPO ). The ~HK$2.2-2.5bn net IPO proceeds (Jul 2025) materially recapitalized the balance sheet and the preferred liability converted at listing — so the pre-IPO negative-equity optics are largely resolved. Verify post-IPO net cash from the FY2025 annual report — this is the top open item.

Market reaction: the FY2025 print (+32% orders, first adjusted profit) "failed to impress" — the stock had already de-rated hard; ~+32% order growth was seen as decelerating vs. history. That tells you what's priced in (Lens 8/12): the market wants re-acceleration or margin, not 30% "in-line."

Lens 6 · Earnings Calls / Founder-Investor Sentiment (trend)

No transcripts on the shelf; this is web-derived from results releases + interviews. Tone arc:

• IPO (Jul 2025): triumphal — "world's first listed AMR company," global-logistics-transformation framing, 133x retail oversubscription.
• H1'25 (Aug 2025): "revenue and profit soar, adjusted EBITDA turns positive" — the pivot from a growth story to a path-to-profit story. First time management leads with profitability, not just share.
• FY2025 (Mar 2026): "profitability milestone," and a hard pivot in language to "embodied intelligence" as the strategic core for 2026 — humanoids (Gino 1), Geekplus Brain, "fully automated unmanned warehouses".

What they started saying: "embodied intelligence," "humanoid," "unified platform / Geekplus Brain," "subscription order growth," "profitability." What they stopped emphasizing: pure share-gain / unit-growth bravado — replaced by margin + recurring + AI. Sentiment read: management is deliberately re-basing the narrative from "fastest-growing AMR" to "the AI-native warehouse-automation platform that also makes money." That is the right story for a public company, but the humanoid pivot is a credibility risk — it invites the market to price an unproven, capital-hungry, hype-adjacent option before the core business has durable GAAP profits.

Lens 7 · Comps (Cap table & public/private peers)

Public equity comps (the honest peer set is thin — Geek+ is the only listed pure-play AMR):

Valuation read: at ~4.7x P/S Geek+ screens ~4x the HK Machinery industry average of 1.1x and above an estimated "fair" ~2.5x P/S; on some cuts the market pays up to 6.4x. For a company at ~breakeven GAAP with 35% gross margin and ~31% growth, that is a growth-software multiple on a hardware-heavy P&L. The bull says it deserves a premium as the only listed pure-play with #1 share and a margin inflection; the bear says AMRs are a hardware-integration business that will not sustain a 5x sales multiple once growth normalizes toward 20% (Lens 12/13).

Cap table (+private overlay — post-IPO ownership):

• Founder Yong Zheng — via Tianjin Jizhi Chuangxiang, holds 38.14% of Class A → ~7.19% economic / ~20.50% of voting rights (weighted-voting-rights "-W" structure). Founder retains control on ~7% economics — a WVR governance flag (Lens 9/10).
• Warburg Pincus ~11.86% (led Series B 2017), GGV Capital ~6.19%, Ant Group ~4.93%, Intel ~0.63%. A tier-1 syndicate — the crossover/strategic quality (Warburg, GGV, Ant, Intel) is a real signal of institutional validation.
• Cornerstones at IPO (US$91.3m): Xiong'an Robot (US$41.3m), Arc Avenue (US$25m), Eastspring/Prudential (US$15m), Yige/Zongteng (US$10m) — note state-linked (Xiong'an) + strategic (Zongteng logistics) anchors, a China-policy tell as much as a financial one.

Funding pre-IPO: ~US$530-660m across ~5 rounds (Series B US$60m 2017 Warburg; Series C ~US$200m+ 2020 GGV/D1/Warburg; Series E US$100m Aug 2022).

Lens 8 · Stock-Price Catalysts

2590.HK only has ~12 months of tape (listed Jul 2025), so this is IPO-era, ``:

• 9 Jul 2025 — IPO at HK$16.80, 133x retail oversubscription, 30x institutional; market cap ~HK$21.8-22bn. Strong debut, ran to a 52-wk high ~HK$33.90.
• Aug 2025 — H1 interim (first positive adjusted EBITDA): confirmed the profit-inflection narrative; supported the post-IPO run.
• ~Q4'25-Q1'26 — the de-rate: stock fell to a 52-wk low ~HK$12.59; −44% YTD, −27% in a month, −13.6% in 7 days at the trough. Drivers: (a) HK small-cap-tech risk-off, (b) cornerstone/pre-IPO lock-up expiry approaching (6-month lock-ups from a Jul 2025 IPO expire early 2026; overhang fear), (c) the FY orders print "failing to impress" (decel narrative).
• Jan 2026 — cornerstone hold commitment: Xiong'an Fund and other cornerstones publicly committed to hold into the unlock — a supply-overhang mitigant that the bull case (e.g. Smartkarma's "buying on weakness into lock-up expiry") leans on.
• Jun 2026 — recovery attempt: back to ~HK$13.09 (~HK$27.7bn cap); 14 analysts, 0 sells, avg target HK$33.05 (high HK$54.68 / low HK$21.68) — a Strong Buy consensus that sits ~2.5x above the current price.

Pattern read: this name trades on (1) lock-up/float dynamics, (2) the profitability-vs-growth debate, and (3) HK-tech-sentiment beta far more than on any single customer. The 2.5x gap between price and sell-side target is either a screaming buy or a sell-side that hasn't marked to the decel — the honest answer is both are possible and the lock-up is the near-term arbiter.

---

Phase C — Judge people & books

Lens 9 · Management

CEO / founder — Yong Zheng. Track record: co-founded Geek+ in 2015 and built it from zero to the #1 global warehouse-AMR vendor and the category's first IPO in ~10 years — an unambiguous, quantified operating achievement. Ex-ABB robotics background (industry-native founder, not a finance parachutist). Skin in the game: retains ~7.2% economics but ~20.5% of votes via a weighted-voting-rights structure — high control, modest economic alignment; founder-led at the archetype level (long-term, product-driven, willing to burn to win share). Capital-allocation history: classic land-grab — ~RMB 3.5bn cumulative losses funded ~US$600m of VC to buy #1 share, then IPO'd to recapitalize. The judgment call is whether that share is now monetizable at a profit — the 2025 adjusted-profit + positive-OCF turn is the first evidence it can be. The pivot to burn again on humanoids is the key capital-allocation question (Lens 14): having just reached breakeven, management is signalling a new capital-hungry bet.

Red flags (governance): (1) WVR "-W" structure — founder controls on ~7% economics; standard for HK new-economy listings but a minority-protection concern. (2) Founder + 3 co-founders + heavy China-state/strategic anchor investors (Xiong'an) — a China-policy dependency that cuts both ways (support + geopolitical exposure). (3) No evidence of related-party self-dealing or promotional stock behavior surfaced — but the "embodied intelligence / humanoid" narrative pivot lands squarely in the hottest hype theme of 2026, which warrants skeptical monitoring (is this genuine roadmap or narrative management into the lock-up?). Founder vs professional: founder-operator, industry-native — the right archetype for a category still being invented, but the transition from "grow share at any cost" to "compound profitable FCF" is exactly where founder-CEOs most often stumble.

Lens 10 · Forensic Red Flags

Web-only; no filings on the shelf to tie out. Every figure ``. This lens is deliberately conservative given the sourcing limit.

1. Adjusted vs. GAAP gap (the headline risk). "First profitable year" = adjusted RMB 43.8m; statutory IFRS is still a ~RMB 10-30m loss. The adjustment strips preferred-share fair-value movements (now moot post-conversion) and SBC. Post-IPO, the adjusted/GAAP gap should narrow sharply — FY2026 is the tell: if statutory net income doesn't cross zero in 2026 with the preferred-share noise gone, the "profitability" story is softer than marketed. Watch SBC as a % of revenue.
2. Revenue quality / project-recognition. Solution sales are installed/commissioned over time — revenue-recognition timing (over-time vs point-in-time), contract liabilities/deferred revenue, and trade-receivable aging are the areas to scrutinize in the annual report. Positive tell: management says all trade & bills receivables are expected to be recovered within one year. Negative-history tell: pre-IPO negative net current assets of ~RMB(6.45)bn (Oct 2024) — driven by the preferred liability, but confirm working-capital normalized post-IPO.
3. Inventory / hardware working capital. A hardware business scaling into 40 countries carries inventory + spares risk; no inventory-vs-revenue divergence sourced → n/a — flag for the annual report.
4. Overseas-margin durability. 46.6% overseas GM vs 35.5% blended is the crown jewel — verify it's not flattered by mix/timing and that price competition (Hai Robotics abroad) isn't eroding it. A falling overseas GM would break the whole thesis.
5. Cash burn vs. positive OCF. FY2025 OCF turned +RMB 85.7m, but that's thin; capex + the humanoid R&D ramp could push FCF negative again in 2026. Confirm net cash post-IPO and the FY2026 capex plan.

Regulatory findings (required sub-section) — from regulatory/regulatory-findings.md:

• SEC (EDGAR EFTS: LR + AAER): 0 findings. Geek+ has no CIK — it is a HK (not US) issuer and does not file with the SEC; no EDGAR enforcement search is possible (correctly noted in the on-disk findings file). This is a coverage gap, not a clean bill — SEC has no jurisdiction here.
• Non-SEC / web: search "Geekplus"/"Geek+" (FTC OR DOJ OR CFPB OR "consent decree" OR settlement OR fine OR penalty) enforcement returned no material enforcement actions. Note the relevant regulator for a China-HQ / HK-listed exporter is the HKEX/SFC (disclosure) and US export controls / entity-list risk (given the AI/robotics + China nexus and edge-compute dependency) — neither surfaced a current action, but the export-control tail risk is a live monitoring item, not a settled all-clear.
• Item 3 / Legal Proceedings equivalent: no 10-K exists; the HKEX prospectus/annual-report legal-proceedings section is not on the shelf → n/a; pull it from the FY2025 annual report.
• Net: No material regulatory or legal findings surfaced via SEC EDGAR EFTS (n/a — non-US issuer), web search, and available disclosure as of 2026-07-06 — with the explicit caveat that HKEX/SFC filings and export-control screens were not machine-checked and should be for any real position.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (IPO-readiness lens is moot — already public; run the operating projection instead)

No forecast.ts logged (unattended --watchlist rule + web-only grounding = not conviction-committed). Output ``, arithmetic shown, built off FY2025 actuals.

Base inputs (FY2025 actual ): revenue RMB 3.171bn; GM 35.5%; adjusted net ~RMB 44m; orders RMB 4.137bn (+31.7%, orders > revenue = forward cover).

Revenue paths (RMB bn):

Margin / profit: the swing variable is whether GM holds ~35% while opex (esp. humanoid R&D) scales. Base case: GM ~35-36%, adjusted net margin creeps to ~3-5% by FY27 → **adjusted net ~RMB 150-250m FY27 **; statutory turns clearly positive in FY26 as preferred-share noise is gone. Bear case: GM slips to ~32% on price competition + humanoid opex → back to a small GAAP loss in FY26-27, which would detonate the 5x P/S multiple.

The forecast that actually matters (would log if committed): "Geekplus (2590.HK) FY2026 statutory (IFRS) net income > 0" — p ≈ 0.70. The preferred-share fair-value drag is gone post-conversion and FY2025 was already breakeven-ish; the main threat to a clean GAAP profit is a humanoid-R&D opex surge. This binary (does the "profitability" become GAAP-real in 2026?) is the single most decision-relevant, scoreable question on the name.

Lens 12 · Bull vs Bear

Bull case. Geek+ is the only public pure-play on a >30% CAGR secular wave (warehouse automation is <25% penetrated globally). It has #1 share for 7 years, a 78% repurchase rate, 75% of revenue overseas at 46.6% margin, and just printed its first adjusted profit + first positive operating cash flow — the growth-to-profitability inflection bulls pay up for. The China-cost / DM-price arbitrage is a structural margin edge Western pure-plays can't match. Orders (RMB 4.1bn, >revenue) underwrite forward growth. A tier-1 cap table (Warburg, GGV, Ant, Intel) and a 0-sell / HK$33 avg-target sell-side say the smart money isn't out. If the embodied-AI/humanoid layer (Geekplus Brain + Gino 1) turns real, Geek+ re-rates from "AMR vendor" to "the AI operating system of the automated warehouse" — a much larger TAM. The contrarian bull: the −44% YTD washout + lock-up overhang has already de-rated a franchise leader to a level where the category, not the multiple, is the debate — buying #1 share into a mechanical-selling air pocket.

Bear case (2-3 permanent-impairment risks). (1) AMRs commoditize and the 5x sales multiple is a category error — this is a hardware-integration business (35% GM, services-heavy install), not software; as growth normalizes to 20%, a re-rate to ~2x P/S is ~50% downside independent of execution. (2) Hai Robotics on home turf + price war — a well-funded, IPO-bound specialist attacking the highest-margin overseas deals could compress the 46.6% overseas GM that the whole thesis rests on. (3) The humanoid pivot burns the balance sheet before the core is durably profitable — management just reached breakeven and is signalling a fresh capital-hungry bet into the hottest hype theme; if Gino 1 is a narrative rather than a P&L, 2026 GAAP flips back to loss and credibility breaks.

Pre-mortem (18 months out, thesis broke): it's early 2028. Growth decelerated to ~15% as Hai Robotics + Western integrators split the market; overseas GM slipped from 46.6% to ~40% on price competition; humanoid R&D pushed FY26-27 back into statutory losses; the lock-up unlock + a HK-tech risk-off cycle drained the float; the stock is at HK$8 (~1.5x sales). The "AI warehouse OS" narrative was 3+ years early. Most likely single kill-shot: margin compression from Chinese-peer price competition abroad.

Are multiples too high? Yes on today's economics — 4.7-6.4x P/S at breakeven GAAP prices in flawless execution + the humanoid option. Justifiable only if (a) FY26 GAAP profit is clean, (b) overseas GM holds ~45%+, and (c) growth stays ≥25%. Miss any one and the multiple is indefensible.

Contrarian view (what the market is refusing to see): the market is fixated on the lock-up overhang and the growth deceleration and is discounting the quality of the mix shift — 75% overseas at 46.6% GM + >90% subscription-order growth means the next dollar of revenue is structurally higher-margin and more recurring than the last. If GM inflects up (not just holds) as recurring compounds, the "hardware multiple" bear thesis is wrong and this re-rates as a margin-expansion story. That's the non-consensus upside — but it's an FY26-FY27 show-me, not a today-fact.

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

Dismantling the bull:

• What structurally breaks the model: AMRs become reference-design commodities; the value migrates to (a) the SI/integrator layer above and (b) the edge-AI/foundation-model layer that any warehouse can license — squeezing the robot vendor to a thin-margin box-shipper. Geek+'s "Geekplus Brain" moat claim is unproven and competes with far larger AI labs.
• Revenue concentration: the risk isn't a single customer (base is ~950, diversified) — it's geographic + policy concentration: 75% overseas means a US/EU protectionist turn against Chinese robotics (tariffs, procurement bans, entity-list on edge compute) directly hits the highest-margin revenue. A single geopolitical event can impair 75% of the P&L.
• Weakest moat point: switching costs are real but one-time — they protect the installed base, not new-logo win rates, where Hai Robotics/Locus/Exotec compete on price + performance every deal.
• Most dangerous underestimated competitor: Hai Robotics — Shenzhen ACR/case-handling specialist, same China cost base (so no arbitrage edge for Geek+ against it), reportedly IPO-bound, and technologically differentiated in high-density storage where the market is moving (tote/case > shelf). Bulls treat Geek+'s lead as durable; on home turf against Hai it is not.
• Worst capital-allocation risk: relapsing into burn on humanoids right after reaching breakeven — plus the WVR structure means minority holders can't discipline a founder who chooses a hype-driven, dilutive bet.
• Assumptions that must hold for today's price: ≥25% growth and ~45%+ overseas GM and clean FY26 GAAP profit and no export-control shock and an orderly lock-up. That's five things, several outside management's control.
• If growth disappoints 20-30% (to ~18-22% and GM to ~32%): the P/S compresses from ~5x toward ~2x → ~50-60% downside to ~HK$5-7. This is the base bear outcome, not a tail.
• Single permanent-impairment scenario, plausibility: US/EU restrictions on Chinese warehouse-robotics + edge-AI (procurement bans / entity-list). Plausibility: moderate-and-rising given the 2026 AI-nationalism backdrop and Geek+'s explicit heavy-compute humanoid pivot — this is the scenario a short would build the thesis on.

Lens 14 · Management Questions (ordered by information value)

1. On an IFRS/statutory basis, when do you expect sustained positive net income — and how much FY2026 opex is committed to humanoid/embodied-AI R&D that could delay it? (The adjusted-vs-GAAP gap is the whole integrity question.)
2. What is your post-IPO net cash today, your FY2026 capex plan, and at what point does the humanoid investment turn FCF negative again?
3. Overseas gross margin is 46.6% vs 35.5% blended — is that durable, or is it being competed away (esp. by Hai Robotics) in your fastest-growing overseas deals? Show the trend by half.
4. What share of revenue is now recurring/subscription vs one-time project sales, and what is the gross margin of each? (You disclose order growth but not the revenue/margin split.)
5. Gino 1 and Geekplus Brain: what is the realistic revenue timeline, and are you selling humanoids or piloting them? What's the unit economics vs an AMR fleet?
6. How exposed is your edge-compute and sensor supply chain to US export controls, and what's your mitigation if advanced AI accelerators are restricted to Chinese robotics firms?
7. Who is your single most dangerous competitor by geography, and where are you losing deals today?
8. Customer concentration: what % of revenue is your top customer and top-5? How has that trended? (Not publicly disclosed.)
9. How do you defend new-logo win rates (not just installed-base retention) as AMR hardware commoditizes?
10. What is your through-cycle target operating margin at scale, and what revenue level unlocks it?
11. Post lock-up expiry, what is early-investor and cornerstone intent, and how do you think about the float overhang?
12. The WVR structure gives the founder ~20.5% votes on ~7.2% economics — how should minority holders think about capital-allocation discipline on a bet like humanoids?
13. Where are the Xiong'an/state-linked cornerstone relationships strategic vs financial, and does that create China-policy dependency in your order book?
14. Working capital: how are you managing inventory, spares, and receivables as you scale across 40+ countries?
15. If you had to cut the roadmap to guarantee GAAP profitability, what goes first — geographic expansion, humanoids, or R&D breadth?

---