Robotics Demand-Side S4-to-S5 Evidence Gate

Date: 2026-06-17 Owner: Finance / Charlie AGT-002 Status: SYNTHESIS_CANDIDATE Visibility: PUBLIC Output intent: morning-review appendix / future /robotics/signals/ or /robotics/why-now/ module after Hugo review Public-safety: public-source evidence map only; no trade recommendation; no Hugo private portfolio data; no private channel checks; no paid-report excerpts.

One-line answer

Robotics 的下一条高价值证据线，不是更多 demo，而是需求侧是否开始形成可复用的采购、付款、部署、复购和经济性披露链；截至 2026-06-17，公开证据已经支持 S4 demand-side validation，但仍不足以支持 S5 scaled-commercial-economics proof。🟢/🟠

Core question

哪些客户侧证据能把 humanoid / embodied robotics 从“有兴趣、有试点、有预算”推进到“有可重复商业经济性”？

Working answer: 用三层需求侧证据门，而不是单一客户 logo。

1. Procurement/spec gate: 买方是否把预算、数量、交付、验收、SDK、保修、国产化或场景要求写进公开采购文件？
2. Payment/deployment gate: 试点是否转成 commercial agreement / RaaS / bounded workflow throughput？
3. Economics/replication gate: 是否披露 robot count、robot-hours、uptime/intervention、ROI/payback、repeat order、vendor revenue/gross margin/service burden？

目前公开证据在第 1–2 层明显增强，但第 3 层仍缺口很大。🟢/🟠

Why this artifact matters for morning review

已有 artifact 已经覆盖 Tesla capacity intent、Figure deployment KPI、Unitree cost-access、Leaderdrive filing-backed supplier signal、RaaS customer-payment gate、public procurement gate 和 customer-side deployment funnel。这个文件的增量是把它们压成一个明早 review 可用的需求侧判断框架：

• 公司侧说“我们能造/能演示”不够；客户侧是否把机器人写进采购、付款、验收和扩张流程，才是 S4→S5 的关键前置信号。🟠
• 需求侧证据可以防止把 TAM、政策口号、customer logo、single pilot 或 funding round 误读成商业化。🟠
• 对 Codex/Warrior 来说，这可以变成一张 slide 或一个 site sidebar："Customer demand is becoming observable; economics are still unpaid evidence debt." 🟠

Evidence ladder

Signal vs noise

Signal

• Public procurement with explicit budget, unit count, delivery window, warranty, SDK/API, acceptance criteria and buyer type. 🟢
• Customer pilot converting into commercial/RaaS agreement, especially after a prior proof-of-concept. 🟢
• Throughput numerator plus denominator: not just “100,000 totes,” but robot count, robot-hours, uptime and intervention rate. 🟢/🟠
• Same-customer expansion after acceptance: one workflow/site → multiple workflows/sites → renewal. 🟢
• Customer-side evaluation infrastructure: center of competence, multi-vendor comparison, pilot-to-rollout process. 🟢
• Vendor financial disclosure tied to deliveries: robot revenue, gross margin, warranty/service burden, backlog/RPO or RaaS unit economics. 🟢

Noise unless upgraded

• Customer logo without contract scope, robot count, task definition or economics. 🟠
• Procurement budget without award, delivery, acceptance or outcome. 🟢/🟠
• Technical specs such as TOPS/DOF/runtime presented as productivity proof. 🟢/🟠
• “RaaS” label without price, utilization, renewal and vendor margin. 🟠
• Policy target treated as delivered deployment. 🟢/🟠
• One-off education/lab purchase treated as industrial labor substitution. 🟢/🟠

Stage classification

Current classification: S4 demand-side validation, not S5 economics.

Why S4:

• Government/policy sources are defining scene funnels, timelines and scenario targets. 🟢
• Public tenders show budgets, specs, unit counts, training, delivery, warranty and developer openness. 🟢
• GXO/Agility shows pilot-to-RaaS conversion and 100,000+ tote throughput numerator. 🟢
• BMW shows a large industrial customer moving from one deployment example toward second-plant / second-vendor evaluation and a Physical AI center of competence. 🟢

Why not S5:

• Reviewed public humanoid sources still do not disclose the full denominator: robot count by customer/site, robot-hours, uptime distribution, intervention minutes, safety events, maintenance hours, service cost, RaaS price, renewal, ROI/payback, vendor robot revenue, gross margin or cash conversion. 🟠
• Procurement and RaaS evidence are stronger than demos, but they can still represent learning budgets, pilots, early deployments or subsidized category construction. 🟠
• Visible procurement unit counts are still small in examples: one unit, four units, one training room or one batch; this is not yet fleet-scale economics. 🟢/🟠

What would change our mind

Upgrade toward S5 if primary/customer/filing sources disclose at least three of these:

1. Robot count by customer/site and utilization robot-hours across multiple months. 🟢
2. Uptime, intervention rate, safety incidents and maintenance/service hours. 🟢
3. Customer ROI/payback tied to throughput, labor capacity, quality, injury reduction, downtime or scrap reduction. 🟢
4. Repeat purchase or renewal after measured acceptance, not just first deployment. 🟢
5. Same workflow replicated across multiple sites or independent customers with comparable metrics. 🟢
6. Vendor robot revenue, gross margin, RaaS price/billing unit, backlog/RPO, warranty/service burden and cash conversion. 🟢
7. Policy scene-summary reports with accepted units, active scenarios, deployment duration and performance metrics by 2026-11/12. 🟢/🟠

Downgrade if:

1. 2026–2027 procurement remains concentrated in display, education, lab and demonstration platforms. 🟢/🟠
2. Commercial/RaaS announcements do not produce robot count, utilization, throughput or renewal within 12–24 months. 🟢/🟠
3. Customers keep building evaluation centers but avoid fleet purchase/renewal. 🟢/🟠
4. Specs emphasize DOF/TOPS/appearance while omitting uptime, safety, maintenance, intervention and ROI. 🟢
5. Service, maintenance, teleoperation or safety burden absorbs the economic value of robot labor. 🟠

Public-safe slide / site draft

Title: Customer demand is becoming visible — economics are still the missing denominator

Five cards:

1. Policy funnel

   • MIIT/SASAC 2026 action: ≥20 key scenes per selected province; ≥10 per central enterprise; 100+ high-value scenarios; “万台级” landing target.
   • Source: MIIT/SASAC notice via Shenzhen Software Industry Association, 2026-06-10 🟢.

2. Procurement specs

   • Suzhou award: 4 humanoids at RMB 469,999 each + RMB 120,000 feature development; total RMB 1,999,996.
   • Specs: 47 DOF, 1.5h runtime, 275 TOPS, SDK/API/simulation files.
   • Source: Suzhou Public Resources Trading Platform, 2026-05-20 🟢.

3. Payment conversion

   • GXO/Agility: proof-of-concept 2023-12-06 → multi-year RaaS/commercial deployment 2024-06-27.
   • Agility: 100,000+ totes moved by 2025-11-20.
   • Source: GXO / Agility 🟢.

4. Customer replication

   • Agility: Mercado Libre commercial agreement 2025-12-10; Toyota Motor Manufacturing Canada RaaS after successful pilot 2026-02-19.
   • BMW: Leipzig / Hexagon AEON pilot process and Physical AI Center of Competence in 2026.
   • Source: Agility / BMW 🟢.

5. Missing S5 denominator

   • Robot count, robot-hours, uptime/intervention, safety, maintenance, ROI/payback, renewal, vendor revenue/GM/service burden.
   • Source: Charlie evidence-gate synthesis, 2026-06-17 🟠.

Footer: Evidence map only. No winner ranking. No trade recommendation. Procurement/RaaS/customer logos are S4 until utilization, ROI, repeat order and unit economics are disclosed.

Common misconceptions

1. “Procurement equals commercialization.” Correction: procurement proves budgeted demand and specifications; commercialization needs delivery, acceptance, repeat use and economics. 🟢/🟠

2. “RaaS equals proven unit economics.” Correction: RaaS proves a payment structure; unit economics need price, utilization, renewal, service burden and gross margin. 🟢/🟠

3. “Customer logo means winner.” Correction: customer logo is a starting point; the signal is whether the customer expands, renews and discloses operational outcomes. 🟠

4. “DOF/TOPS/runtime specs prove productivity.” Correction: specs prove buyer requirements and hardware configuration, not uptime, intervention, throughput or payback. 🟢/🟠

5. “Policy target means delivered fleet.” Correction: policy target creates a funnel and deadline; delivered deployment needs accepted units and measured scenario outcomes. 🟢/🟠

Think Deeper questions

• The first S5 proof will likely come from which source: customer acceptance report, vendor filing, RaaS renewal, or policy scene-summary report?
• If customers evaluate multiple humanoid vendors in parallel, does value migrate from robot bodies toward fleet orchestration, integration, maintenance and service operations?
• Will education/lab procurement become a real data/talent flywheel, or remain a small learning budget?
• What denominator should public research demand after every throughput headline: robot count, robot-hours, intervention, service cost or ROI?
• Does customer-side procurement standardization commoditize robot OEMs, or accelerate category adoption by making tests comparable?

Source list

• MIIT / SASAC, “工业和信息化部办公厅 国务院国资委办公厅关于联合开展2026年度人形机器人与具身智能实景实训专项行动的通知,” 工信厅联科函〔2026〕256号, dated 2026-06-03; reposted by Shenzhen Software Industry Association 2026-06-10. 🟢 https://ssia.org.cn/page74?article_id=2726
• Suzhou Public Resources Trading Platform, “苏州市具身智能机器人综合创新中心人形机器人采购项目中标结果公告,” 2026-05-20, project SZQY2026-GQ-WZ-G-011. 🟢 https://ggzy.suzhou.gov.cn/jyxx/003034/003034003/20260520/987cb1ce-ab04-46d9-a445-3ff2df9d3bc5.html
• Beijing Public Resources Trading Service Platform, “（教育教学改善）具身智能数据采集与人形机器人装配技术实训中心建设公开招标公告,” 2026-05-14, project S110000C005106773001. 🟢 https://ggzyfw.beijing.gov.cn/jyxxcggg/20260514/5531110.html
• GXO, “GXO conducting industry-leading pilot of human-centric robot,” 2023-12-06. 🟢 https://gxo.com/news_article/gxo-conducting-industry-leading-pilot-of-human-centric-robot/
• GXO, “GXO signs industry-first multi-year agreement with Agility Robotics,” 2024-06-27. 🟢 https://gxo.com/news_article/gxo-signs-industry-first-multi-year-agreement-with-agility-robotics/
• Agility Robotics, “Digit Moves Over 100,000 Totes in Commercial Deployment,” 2025-11-20. 🟢 https://www.agilityrobotics.com/content/digit-moves-over-100k-totes
• Agility Robotics, “Mercado Libre and Agility Robotics Announce Commercial Agreement to Deploy Humanoid Robots,” 2025-12-10. 🟢 https://www.agilityrobotics.com/content/mercado-libre-and-agility-robotics-announce-commercial-agreement
• Agility Robotics, “Agility Robotics Announces Commercial Agreement with Toyota Motor Manufacturing Canada,” 2026-02-19. 🟢 https://www.agilityrobotics.com/content/agility-robotics-announces-commercial-agreement-with-toyota-motor-manufacturing-canada
• BMW Group PressClub, “BMW Group to deploy humanoid robots in production in Germany for the first time,” 2026. 🟢 https://www.press.bmwgroup.com/global/article/detail/T0455864EN/bmw-group-to-deploy-humanoid-robots-in-production-in-germany-for-the-first-time?language=en
• Figure AI, “F.02 Contributed to the Production of 30,000 Cars at BMW,” 2025-11-19. 🟢 https://www.figure.ai/news/production-at-bmw
• Charlie synthesis across procurement / RaaS / customer deployment artifacts, 2026-06-17. 🟠

Public-safety flag

PUBLIC-safe as a research framework and evidence-gate memo. Do not include Hugo portfolio weights, trade rationale, private channel checks, paid-report excerpts, rumors, buy/sell/hold language, tax/legal context, or named supplier winner inference. Do not imply procurement, RaaS, customer logo, throughput numerator, policy target or pilot equals S5 economics unless future primary sources disclose utilization, ROI, repeat order and financial materiality.