Investment ResearchTopic research, evidence curves, and cycle signals
HomeSign inEN

Research library · updated 2026-06-14 · public

Warrior Brief｜Robotics Q1 Web Page Package

Date: 2026-06-14 Owner: Hugo / Genius Team Research agent: Finance / Charlie AGT-002 Target worker: Warrior / Team Fullstack Status: READY_FOR_WARRIOR Visibility: PUBLIC Output: site Priority: P0 after Q0, or P1 if Q0 still rendering Target route: /robotics/value-stack/

0. Source files

Canonical research package:

  • workspace/finance/knowledge/robotics-q1-value-stack-apple-android-or-something-else-v2.md
  • workspace/finance/knowledge/robotics-q1-ai-model-data-and-smartphone-platform-analog-discussion.md
  • workspace/finance/knowledge/robotics-value-migration-v1.md
  • workspace/finance/knowledge/robotics-foundation-model-layer-evidence-v1.md
  • workspace/finance/knowledge/robotics-oem-make-vs-buy-v1.md
  • workspace/finance/knowledge/robotics-why-now-hardware-cost-catalyst-v1.md

Do not pull private portfolio data. No trade recommendation.

1. Page title and thesis

Title:

Robotics Value Stack: Apple, Android, or Something Else?

Subtitle:

Why humanoid robotics should be analyzed layer by layer, not as one market.

Thesis:

Robotics may become Apple-like, Android-like, or something more industrial. Vertical OEMs may control premium systems; low-cost bodies may expand the installed base; horizontal model/tooling platforms may capture high-margin value; deployment operators may become the hidden bottleneck. The open question is not “who builds the best robot body,” but “who controls the learning loop that makes robots useful across customers and tasks?”

2. Warrior display goal

Make this a public field guide, not a research dump.

The page should help readers understand:

  1. Why “robotics TAM” is the wrong starting point.
  2. Why value capture must be analyzed by layer.
  3. Why AI/model/data could become a large value pool.
  4. Why the smartphone analogy helps but breaks.
  5. Which evidence would upgrade each scenario.

3. Suggested information architecture

Hero

  • Big claim: robotics is a value stack, not one market.
  • Framing: Apple / Android / NVIDIA platform / industrial-service hybrid.
  • Caveat: analogy is a lens, not a forecast.

Section 1 — Six-layer stack

Use cards, not a dense table:

  1. OEM body / full robot platform
  2. AI / model / data / simulation
  3. Components / supply chain
  4. Deployment / integration / RaaS / services
  5. Customer productivity capture
  6. Public-market expression

For each card show:

  • What is sold
  • Evidence today
  • Missing S5 proof

Section 2 — Why AI/model/data matters

Show the learning-loop chain:

Autonomy → real-world data → simulation → deployment tuning → fleet learning → safety validation → workflow integration.

Key public-safe claim:

The model/data layer matters only if it reduces intervention, deployment time, failure rate, and payback time.

Section 3 — Smartphone platform lens

Use four comparison cards:

  1. Apple-style vertical robotics stack

    • Tesla / Figure-like.
    • Closed full-stack premium systems.
    • Watch: margin, repeat deployments, intervention, service attach.

  2. Android-style open embodiment ecosystem

    • Unitree-style low-cost bodies + open/common model/tooling stack.
    • Watch: developer ecosystem, cross-body model portability, paid tooling.

  3. NVIDIA/Google-style enabling platform

    • Compute, simulation, middleware, foundation model tooling.
    • Watch: robotics-specific revenue, OEM attach, developer adoption.

  4. Industrial-service hybrid

    • Integrators/RaaS/operators capture value through deployment and SLA.
    • Watch: payback, support cost, maintenance margin, repeat expansion.

Section 4 — Why the phone analogy breaks

Use strong caveat list:

  • Physical liability
  • Site integration
  • Embodiment diversity
  • Service/maintenance burden
  • Customer ROI gate

Section 5 — Evidence dashboard

Small dashboard:

  • S3/S4 evidence today
  • S5 proof missing
  • Upgrade signals
  • Downgrade signals

Section 6 — Public-market caution

Explain that investable proxies are imperfect:

  • Figure / Unitree / Skild / PI private.
  • Tesla not pure robotics.
  • NVIDIA ecosystem not equal revenue materiality.
  • Leaderdrive has filing-backed component evidence but not named humanoid OEM validation.

4. Must-include evidence cards

Unitree low-cost hardware

  • R1 / R1 AIR about US$4,900-5,900.
  • G1 from US$13.5K.
  • H2 US$29,900.
  • Source grade: 🟢 Unitree official pages captured in local evidence base.
  • Interpretation: S3 cost-access signal, not S5 economics.

Figure vertical integration / manufacturing

  • Figure says critical modules including actuators, batteries, sensors, structures, electronics are designed completely in-house; suppliers support individual components. 🟢
  • BotQ: 350+ Figure 03, 1/day to 1/hour, >80% EOL FPY, 99.3% battery-line FPY, 9,000+ actuators. 🟢
  • Interpretation: Apple-style / vertical-stack evidence, not full economics.

AI/model/data layer

  • PI π0/openpi/π0.5/π0.7: generalist policy and open tooling evidence. 🟢/🟠
  • Skild: US$300m Series A at US$1.5bn; US$1.4bn Series C at >US$14bn; ABB/UR/NVIDIA partnerships; Zebra/Fetch acquisition. 🟢/🟠
  • Interpretation: strong S3/S4 platform formation, not S5 monetization.

Leaderdrive / component layer

  • 2025 revenue RMB 570.714m, +47.31%.
  • Net profit RMB 124.367m, +121.42%.
  • Harmonic reducer sales 425,158, +72.48%.
  • “工业及具身智能机器人零部件” revenue RMB 422.528m, GM 34.88%.
  • Interpretation: filing-backed S4 supplier evidence, not named humanoid customer validation.

Figure BMW deployment

  • 10-hour shifts Monday-Friday.
  • 90,000+ parts loaded.
  • 1,250+ runtime hours.
  • 30,000+ BMW X3 vehicles contributed.
  • Interpretation: deployment/workflow S4 signal, not customer ROI/payback proof.

5. Do-not-include / guardrails

Do not write:

  • “AI/model/data is already the biggest value pool.”
  • “Tesla is Apple of robotics.”
  • “Unitree is Android of robotics.”
  • “NVIDIA is guaranteed to capture robotics economics.”
  • “Leaderdrive is CATL/NVIDIA of robotics.”
  • Buy/sell/hold or position-size language.

Use instead:

  • “could become”
  • “scenario”
  • “evidence today suggests”
  • “missing before S5”
  • “analogy as lens, not forecast”

6. Acceptance criteria

Page is acceptable if:

  1. It opens with value-stack framing, not stock-picking.
  2. It uses Apple/Android as a lens, not deterministic prediction.
  3. It includes the six-layer stack.
  4. It gives AI/model/data the right emphasis without overclaiming.
  5. It preserves S3/S4 vs S5 distinction.
  6. It includes source grades or source-note summaries.
  7. It has a clear “what would prove this” section.
  8. It is readable as a public field guide.

7. Suggested route linkage

This page should follow Q0:

  • Q0 /robotics/cycle-stage-timing/: where are we in the cycle?
  • Q1 /robotics/value-stack/: if the cycle works, where does value go?

Add navigation between the two.