Q5 China vs US Robotics Path Comparison: Deployment Machine vs Frontier Stack v1

Date: 2026-06-15 Owner: Hugo / Genius Team Agent: Finance / Charlie AGT-002 Status: SYNTHESIS_CANDIDATE Visibility: PUBLIC Primary use: Q5 research mainline, /robotics/us-china-paths, China/US field-guide module, slide compression source

Related files:

• knowledge/robotics-research-map.md
• knowledge/robotics-mainline-q0-q5-evidence-matrix-v1.md
• knowledge/robotics-china-policy-deployment-machine-v1.md
• knowledge/robotics-four-deep-questions-us-china-tesla-figure-leaderdrive-v1.md
• knowledge/robotics-q0-cycle-stage-and-investment-timing-v2.md
• knowledge/robotics-q1-value-stack-apple-android-or-something-else-v2.md
• knowledge/robotics-q2-tesla-optimus-apple-style-candidate-still-missing-s5-proof-v2.md
• knowledge/robotics-q3-figure-deployment-engineering-bottleneck-v2.md
• knowledge/robotics-leaderdrive-s4-to-s5-evidence-debt-v1.md

Public-safety note:

• Public-safe: yes.
• No Hugo private portfolio data.
• No trade recommendation.
• No country-winner rhetoric.
• No private channel checks, paid-report excerpts, or supplier rumors.

0. One-line answer

截至 2026-06-15，Q5 的正确问题不是“中美谁赢机器人”，而是两条路径谁更先把 S4 evidence 转成 S5 economics：中国路径更像 deployment machine + supplier flywheel，美国路径更像 frontier humanoid + AI/model/compute stack。中国有更大的工业机器人部署基数、国内供应商份额提升和 2026 实景实训政策漏斗；美国有 Tesla/Figure/Apptronik/Agility/NVIDIA/模型层和 mega-round frontier concentration。两边都还没有公开证明 broad S5 humanoid economics。🟢 IFR / A3 / USCC / MIIT / Figure / Tesla sources; 🟠 Charlie synthesis.

1. Core question

Q5 应该放在 Q0-Q4 之后，因为 China vs US 是综合题，不是起点。

The core question:

Does China's deployment flywheel convert into humanoid economics faster than the US frontier AI/model stack converts into deployable fleets?

Sub-questions:

1. China 的工业机器人 deployment base 是否能转化成 humanoid/embodied-AI real-scene deployment？
2. 中国 2026 MIIT/SASAC 实景实训机制能否把 policy target 变成 accepted units、routine deployment、ROI/payback 和 repeat orders？
3. US 的 frontier capital / AI / model / compute stack 是否能把 Tesla/Figure-style evidence 推到 multi-customer S5 economics？
4. 哪一层最终捕获价值：OEM body、model/data、components、deployment service、customer productivity，还是 public-market proxy？

2. Path comparison matrix

3. What China proves

3.1 Deployment scale is already real in industrial robotics

China installed 295,000 industrial robots in 2024, accounting for 54% of global demand, with operational stock around 2.027m robots and domestic supplier share reaching 57% from 47% in 2023. 🟢 IFR China release / World Robotics 2025 references.

Interpretation:

• This is real deployment scale, not a policy slogan. 🟢
• It creates a large base of factories, integrators, components, maintenance, and customer-process knowledge. 🟠
• It does not prove humanoid robots have crossed the S5 economics gate. 🟠

3.2 China is building an evidence-production machine

The 2026 MIIT/SASAC real-scene training action creates a dated deployment funnel:

Policy objective
→ real-scene training spaces
→ user unit + OEM + model + component consortium
→ task skill packages
→ validation procedure and report
→ routine deployment in same / adjacent scenarios
→ cross-region / cross-industry replication
→ only then: economics, revenue, margin, repeat orders

Key anchors from the existing China policy artifact:

• Notice dated 2026-06-03 and published 2026-06-08. 🟢
• Work-plan submission deadline 2026-06-30. 🟢
• Effectiveness summary deadline 2026-11-30. 🟢
• End-2026 target for representative scenarios to complete validation and routine deployment. 🟢
• 100+ high-value application scenarios. 🟢
• “万台级规模落地能力” as landing-capability target, not delivered-unit proof. 🟢/🟠
• Provincial regions select at least 20 key scenario units covering at least two domains; relevant central SOEs select at least 10 key scenarios. 🟢
• RaaS / utility-based payment / operating-lease language. 🟢

Interpretation:

• China may have an edge in the “missing middle”: scene access, user units, validation, and replication. 🟠
• The policy itself is S4 infrastructure, not S5 economics. It does not disclose delivered units, uptime, intervention rate, ROI/payback, RaaS pricing, vendor revenue, gross margin, or repeat orders. 🟠

3.3 Supplier evidence can enter filings before OEM economics

Leaderdrive / 绿的谐波 illustrates the China supplier-flywheel path:

• 2025 revenue RMB 570.714m, +47.31% YoY. 🟢
• 2025 net profit RMB 124.367m, +121.42% YoY. 🟢
• “工业及具身智能机器人零部件” revenue RMB 422.528m, +52.61%, gross margin 34.88%. 🟢
• Harmonic reducer sales 425,158 units, +72.48%. 🟢
• 2026 Q1 revenue RMB 140.134m, +42.96%; net profit RMB 32.634m, +61.17%. 🟢

Interpretation:

• This shows supply-chain acceleration can be filing-backed even while humanoid OEM economics remain private or undisclosed. 🟠
• It does not prove named humanoid OEM design win, humanoid-only revenue, contract value, BOM attach, or customer ROI/payback. 🟢/🟠

4. What the US proves

4.1 Frontier humanoid evidence is stronger in named private/public stacks

US evidence is not primarily industrial robot installation scale. It is frontier-company and model-stack concentration:

• Tesla Optimus has SEC-filed S4 manufacturing-infrastructure / designed-capacity evidence, including Fremont and Texas capacity-intent anchors in existing Tesla artifacts. 🟢/🟠
• Figure has BMW customer-site deployment KPI and BotQ manufacturing-process KPI. 🟢
• Figure raised US$675m at US$2.6bn valuation in 2024 with investors/partners including Microsoft, OpenAI Startup Fund, NVIDIA and others. 🟢 Figure / PRNewswire source in existing artifacts.
• NVIDIA / Isaac GR00T and related model/data/simulation infrastructure provide a strong AI/model layer signal. 🟢/🟠

Interpretation:

• The US path is frontier-stack-heavy: AI/model/compute, private capital, and a few high-signal humanoid companies. 🟠
• This does not equal broad deployment economics. Tesla still lacks actual Optimus output/economics; Figure still lacks robot count, ROI/payback, intervention distribution, revenue, and margin. 🟢/🟠

4.2 North America industrial automation is not the same scale as China

A3 reported 31,311 North American robot orders in 2024, valued at US$1.963bn, with +0.5% units and +0.1% value vs 2023. 🟢 A3, URL returned HTTP 200 and keyword anchors matched on 2026-06-15.

Interpretation:

• North America remains a meaningful industrial automation market. 🟢
• But its current industrial-robot order scale is much smaller and flatter than China’s factory-robot installation base. 🟢/🟠
• This does not decide humanoid outcomes because frontier humanoid evidence and industrial automation base are different layers. 🟠

5. Why neither side has proven S5

China is not S5 because:

• industrial robot installations are not humanoid accepted units;
• policy targets are not delivered robots;
• scenario lists are not repeat paid deployments;
• supplier revenue is not named humanoid OEM economics;
• domestic supplier share is not customer ROI/payback;
• “万台级” landing capability is not audited unit deployment or revenue.

US is not S5 because:

• Tesla designed capacity is not actual production or robot economics;
• Figure BMW KPI is not customer ROI/payback or repeat order;
• Figure BotQ manufacturing output is not economically utilized customer fleet;
• frontier funding is not product-market fit;
• model/data stack evidence is not robot-level gross margin or service burden;
• multi-customer paid deployment economics remain sparse in public sources.

6. What would change our mind

China upgrade toward S5 if:

1. 2026 MIIT/SASAC program publishes named scenario outcomes with delivered/accepted robot units, routine deployment data, uptime/intervention, and validation reports. 🟢 needed.
2. SOEs or industrial customers disclose paid deployments, RaaS pricing, utilization, ROI/payback, and repeat expansion. 🟢 needed.
3. China humanoid OEMs disclose revenue, backlog, margin, service cost, or repeat orders across multiple customers. 🟢 needed.
4. Supplier filings disclose named humanoid customers or humanoid-specific revenue/margin with durable gross margin. 🟢 needed.
5. Deployment playbooks replicate across provinces and scenarios without heavy subsidy dependence. 🟢/🟠 needed.

US upgrade toward S5 if:

1. Tesla discloses actual Optimus output, deployed robot count, factory productivity, intervention rate, ROI/payback, revenue or margin. 🟢 needed.
2. Figure / Apptronik / Agility / other US humanoid companies disclose multi-customer paid deployments with robot count, uptime/intervention, ROI/payback, repeat orders, and service economics. 🟢/🟡 needed.
3. Model/data layers measurably reduce deployment time, intervention rate, or task-engineering cost in customer environments. 🟢/🟠 needed.
4. The US frontier stack creates repeatable deployment playbooks rather than bespoke pilots. 🟢/🟠 needed.

7. Public synthesis: not horse race, but conversion test

The public page should not say “China wins” or “US wins.” It should say:

China has the stronger deployment machine; the US has the stronger frontier stack. The investment question is which path converts first from S4 evidence into S5 economics.

This frame preserves the real difference:

• China: deployment scale, customer access, policy-to-scene machinery, domestic suppliers, industrial base.
• US: frontier humanoid companies, AI/model/compute, mega-round private capital, vertical-stack ambition.

It also preserves the real uncertainty:

• China may create many scenarios but still fail to prove repeatable economics.
• The US may build better frontier robots but fail to deploy cheaply enough or broadly enough.
• Value may migrate away from OEM bodies into model/data, components, deployment service, safety/certification, or customer productivity.

8. Signal vs noise

Signal

• Industrial robot installations, operational stock, domestic supplier share. 🟢
• Dated policy deployment funnel with deadlines, scenario requirements, validation and routine-deployment targets. 🟢
• Named-customer deployment KPI with runtime/task/intervention context. 🟢
• Filing-backed supplier revenue, unit shipment, margin, inventory. 🟢
• SEC-filed capacity/manufacturing infrastructure evidence. 🟢
• Model/data/simulation tooling that measurably reduces deployment cost or intervention. 🟢/🟠

Noise unless upgraded

• “China has more industrial robots, so humanoids are solved.” 🟠
• “US has better AI, so deployment economics are solved.” 🟠
• “Policy target equals delivered units.” 🔴/🟠
• “Mega-round equals product-market fit.” 🔴/🟠
• “Customer logo equals ROI/payback.” 🔴/🟠
• “Supplier product fit equals named OEM exposure.” 🔴/🟠
• “Country comparison equals stock recommendation.” 🔴

9. Common misconceptions

1. Misconception: “China’s industrial robot scale proves humanoid commercialization.”

   • Correction: It proves deployment base and industrial automation depth; humanoid S5 still needs accepted units, uptime/intervention, ROI/payback, revenue, margin, and repeat orders. 🟢/🟠

2. Misconception: “US frontier AI/model strength automatically wins robotics.”

   • Correction: AI/model stack matters only if it lowers intervention, deployment time, failure rate, and payback in real customer sites. 🟠

3. Misconception: “Policy support means every Chinese vendor benefits.”

   • Correction: Policy creates a funnel; winners require validation, safety, reliability, customer economics, and repeat deployment. 🟠

4. Misconception: “Tesla/Figure evidence means US already has S5.”

   • Correction: Tesla is S4 capacity intent; Figure is S4 deployment/manufacturing KPI. Both still lack full economics. 🟢/🟠

5. Misconception: “China vs US should be the opening question.”

   • Correction: It should come after cycle stage, value layer, Tesla/Figure evidence, and supplier evidence. Otherwise geography turns into narrative before evidence. 🟠

10. Think Deeper questions

1. Which converts first into S5: China’s real-scene deployment funnel or the US frontier humanoid/model stack?
2. Does the bottleneck move from robot body to deployment engineering, and if so which country has the better deployment engineering machine?
3. If China wins deployment scale but US wins model/data layer, where does value capture land?
4. If US companies remain private longer, which public-market layers can reflect their progress without forcing bad proxies?
5. Could China’s policy-driven deployments produce data faster but economics slower?
6. Could US frontier companies achieve better autonomy but slower adoption because customer-site integration is harder?
7. Which evidence would matter most in 2026-2027: accepted units, robot revenue, customer ROI, intervention rate, repeat order, or supplier humanoid revenue split?

11. Public-safe site draft

China vs US robotics: deployment machine vs frontier stack

The robotics race is not one race.

China’s advantage is deployment machinery. In 2024, China installed 295,000 industrial robots, about 54% of global demand, and its operational stock exceeded 2 million units. Domestic suppliers reached 57% share in China, up from 47% in 2023. On top of that, the 2026 MIIT/SASAC real-scene training action tries to turn humanoid and embodied-AI progress into a deployment funnel: scene lists, user units, OEM/model/component consortia, validation reports, routine deployment, and replication. 🟢

The US advantage is frontier stack concentration. Tesla has the strongest filing-backed capacity-intent curve; Figure has the clearest customer-site humanoid deployment KPI and BotQ manufacturing KPI; NVIDIA and related model/data/simulation infrastructure give the US a strong AI/model layer. 🟢/🟠

Neither path has proven S5 yet. China’s industrial robot scale is not humanoid unit economics. US frontier demos, capacity and funding are not repeat paid deployments. The next important question is conversion: which system turns S4 evidence into S5 economics first?

Footer: Evidence map only. No country winner ranking. No trade recommendation. Industrial robot scale ≠ humanoid economics; frontier AI ≠ deployment economics.

12. Slide-ready compression

Title: China vs US robotics: deployment machine vs frontier stack

Three-card layout:

1. China: deployment machine

   • 295,000 industrial robot installations; 54% global demand; 2.027m operational stock; 57% domestic supplier share. 🟢
   • 2026 MIIT/SASAC real-scene action: 100+ scenarios, end-2026 routine deployment target, “万台级” landing capability target. 🟢/🟠

2. US: frontier stack

   • Tesla capacity-intent filings; Figure BMW + BotQ KPI; Figure US$675m Series B at US$2.6bn; NVIDIA/model/data stack. 🟢/🟠
   • Stronger frontier concentration, not broad S5 proof.

3. Conversion test

   • China must prove accepted units, utilization, ROI/payback and repeat deployment.
   • US must prove multi-customer paid deployments, low intervention and economics.
   • Neither gets S5 credit from demos, policy targets, mega-rounds, or logos alone.

Footer: Evidence map only. No country winner ranking. No trade recommendation. Deployment scale ≠ humanoid economics; frontier stack ≠ customer ROI.

13. Source list

Primary / official sources:

• IFR, “China Tops World Record of 2 Million Factory Robots,” 2025-09-25. URL reached HTTP 200 on 2026-06-15; PDF text keywords not extracted in current environment, so numeric anchors rely on existing IFR-derived artifacts. 🟢 https://ifr.org/downloads/press_docs/2025-09-25-IFR_press_release_China_in_English.pdf
• IFR / World Robotics 2025 references in existing China policy artifact for global installations, China installations, operational stock and supplier share. 🟢
• A3, “North American Robotics Market Holds Steady in 2024 Amid Sectoral Variability,” 2025-02-17. URL returned HTTP 200 on 2026-06-15; keyword anchors 31,311, 1.963, 0.5%, 0.1% matched. 🟢 https://www.automate.org/market-intelligence/insights/a3-reports-north-american-robotics-market-holds-steady-in-2024-amid-sectoral-variability
• U.S.-China Economic and Security Review Commission, “Humanoid Robots,” 2024-10-10. URL reached HTTP 200 on 2026-06-15; PDF text keywords not extracted in current environment, so claims rely on existing USCC-derived artifacts. 🟢 https://www.uscc.gov/sites/default/files/2024-10/Humanoid_Robots.pdf
• MIIT / SASAC, “两部门关于联合开展2026年度人形机器人与具身智能实景实训专项行动的通知,” dated 2026-06-03, published 2026-06-08. URL returned HTTP 200 on 2026-06-15; keyword anchors 100, 万台, 2026, 场景 matched. 🟢 https://www.miit.gov.cn/zwgk/zcwj/wjfb/tz/art/2026/art_f291ccd3da4c47ce95741de63cc088e6.html
• Tesla Q1 2025 to Q1 2026 SEC Form 8-K Exhibit 99.1 sequence, cross-referenced in Tesla Q2 artifacts. 🟢
• Figure official BMW / BotQ / Catalyst / Helix sources, cross-referenced in Q3 artifacts and rechecked 2026-06-15. 🟢
• 绿的谐波 2025 annual report / 2026 Q1 report / abnormal-trading announcement, cross-referenced in Q4 artifacts. 🟢

Secondary / synthesis:

• F-Prime Capital, “State of Robotics 2025,” for Western robotics/AV funding context. 🟡
• Crunchbase News China robotics funding discussion in existing four-deep-questions artifact; use only as secondary lead until category definitions are rechecked. 🟡
• Charlie calculations and synthesis: China/US industrial installation ratio, path comparison, S4/S5 classification. 🟠

14. Public-safety flag

Public-safe: yes, if used as an evidence map and not as a country-winner or stock-picking page.

Do not include:

• Hugo private portfolio data.
• Trade recommendations.
• Buy / sell / hold language.
• Paid-report excerpts.
• Private channel checks.
• Country triumphalism.
• Claims that policy targets are delivered units.
• Claims that industrial robot scale proves humanoid economics.
• Claims that US frontier stack proves customer ROI.