Robotics OEM Public Evidence Matrix v1

Date: 2026-06-11 Owner: Hugo / Genius Team Agent: Finance / Charlie AGT-002 Status: public-safe research artifact Visibility: PUBLIC Source base: existing robotics knowledge files and cited primary sources

0. One-line answer

截至 2026-06-11，humanoid / embodied robotics 的公开证据已经从“demo 新闻”升级到多条 S4 级别证据曲线：Tesla 披露产线与长期设计产能，Figure / Agility 披露真实客户场景运行指标或 RaaS 商业部署，UBTECH 披露 filing-backed humanoid revenue / units，Unitree 披露低价产品化硬件平台，Apptronik / 1X 披露强合作与产能/预订意图。但这些证据仍没有合并成 S5：scaled commercial economics / repeat-order ROI / margin proof。🟢/🟠 as of 2026-06-11

Public framing: this is an evidence map, not a winner ranking and not a trade recommendation.

1. Core question

现在 public evidence 到底证明了什么？

更具体地说：

1. 哪些公司已经从 demo 进入 deployment / production / revenue evidence？
2. 哪些证据只是 capacity / partnership / low price / preorder，而不是 commercialization proof？
3. Codex 在公开页面或 slide 中应该如何压缩这些公司，而不制造“谁赢了”的误导？

2. Evidence-type matrix

3. Signal / noise compression for public use

Signal

• S4 signal: filing-backed humanoid revenue and unit sales, e.g. UBTECH 2025 annual report disclosing RMB 820.557m full-size embodied humanoid revenue and 1,079 full-size embodied humanoid units sold. 🟢 as of 2025 annual report / disclosed 2026-04-15.
• S4 signal: named customer deployment with operating metrics, e.g. Figure BMW deployment metrics or Agility 100,000+ tote milestone. 🟢 as of Figure 2025-11-19 and Agility 2025-11-20 official posts.
• S4 signal: manufacturing-process metrics, e.g. Figure BotQ >80% EOL first-pass yield and 99.3% battery-line first-pass yield. 🟢 as of Figure 2026-04-29 official post.
• S4 signal: public-company filing or official disclosure of capacity / production-line investment, e.g. Tesla Optimus line disclosures. 🟢 as of Tesla Q1 2026 Form 8-K Exhibit 99.1.
• S3/S4 signal: dramatic product price reduction and developer tooling, e.g. Unitree R1 from US$4,900 and G1 from US$13.5K. 🟢 as of product pages accessed 2026-06-10.

Noise / not enough alone

• Viral demo without real-world runtime, customer, or production metrics. 🔴 if unsupported.
• Customer logo without contract value, task KPI, duration, robot count, or repeat-order evidence. 🟠 if only inferred.
• Designed capacity without actual output / yield / utilization. 🟢 for stated capacity, 🟠 if interpreted as production.
• Low hardware price without margin / reliability / support-cost proof. 🟢 for price, 🟠 for commercial economics inference.
• Preorder or deposit without conversion, delivery, refund, ASP, or retention data. 🟢 for company-disclosed preorder mechanics, 🟠 for revenue implication.

4. Current stage classification

As of 2026-06-11, the public evidence supports:

• Industry stage: Stage 2.5-3 — pilot / early deployment / production-process validation / selective filing-backed commercial application. 🟠 synthesis from public sources.
• Not yet proven: Stage 4-5 scaled deployment / return cycle. Missing evidence includes multi-customer repeat paid deployments, low intervention rates, customer-confirmed ROI/payback, durable gross margin, service-cost burden, and material recurring revenue. 🟠 synthesis from public-source gaps.

5. Public-safe slide compression

Suggested slide title:

“Humanoid robotics is not one race — it is seven evidence curves.”

Suggested subtitle:

“Capacity, deployment KPIs, low-cost hardware, filing-backed revenue, RaaS deployment, partner stack, and home preorders each answer different questions.”

Suggested layout:

1. Left column: Evidence curve

   • Tesla: capacity ambition
   • Figure: deployment + manufacturing KPI
   • Unitree: cost-access curve
   • UBTECH: filing-backed humanoid revenue / units
   • Agility: bounded logistics RaaS deployment
   • Apptronik: partner-rich scaling path
   • 1X: home-humanoid preorder + data loop

2. Middle column: What public evidence supports

   • Use one number per company from the matrix.

3. Right column: Missing before S5

   • Output / utilization / uptime / intervention / ROI / margin / repeat order.

Footer caveat:

“Evidence map, not winner ranking. S4-heavy does not equal S5 scaled commercial economics.”

6. Common misconceptions

1. “Who has the best demo wins.”

   • Correction: demo quality is not enough. The higher-signal evidence is production output, deployment duration, task KPI, customer economics, and margin. 🟠 framework.

2. “Tesla capacity numbers mean Optimus is already mass-produced.”

   • Correction: designed capacity and production-line installation are S4 infrastructure signals, not actual production, yield, or commercial sales. 🟢/🟠 Tesla filing evidence plus interpretation.

3. “Unitree low price means commercialization is solved.”

   • Correction: low price reduces experiment cost, but says little about reliability, uptime, support burden, autonomy, and gross margin. 🟢/🟠 product price plus interpretation.

4. “UBTECH revenue means S5 is already proven.”

   • Correction: filing-backed humanoid revenue and units are strong, but S5 still needs customer ROI/payback, humanoid-specific margin, repeat-order quality, and service burden. 🟢/🟠 filing data plus interpretation.

5. “Partnership stack equals economics.”

   • Correction: Apptronik / 1X partnerships and capacity/preorder disclosures are important S4 signals, but without revenue, margin, utilization, or retention they remain pre-S5. 🟢/🟠.

7. What would change our mind

Upgrade toward S5 if any company discloses:

• Multi-customer paid deployments with robot count, duration, task KPI, and repeat orders. 🟢 if official / filing-backed.
• Customer-confirmed ROI/payback and operational savings. 🟢/🟡 depending on source.
• Uptime, failure rate, intervention rate, and support cost over months, not demos. 🟢 if official customer/company; 🟡 if credible third party.
• Actual production output, yield, utilization, and gross margin. 🟢 if filing or audited disclosure.
• Segment revenue that is material and clearly humanoid / embodied robotics-specific. 🟢 if filing-backed.

Downgrade if:

• Capacity targets slip without explanation or actual output data. 🟢/🟡 depending on source.
• Customer pilots do not renew or become repeat orders. 🟢/🟡.
• Low-cost hardware produces high warranty/support burden or weak autonomy. 🟢/🟡.
• Supplier or OEM filings show revenue growth without cash conversion, margin durability, or customer quality. 🟢/🟠.

8. Think Deeper questions

1. Which evidence curve historically matters first in a new hardware category: price-access, manufacturing yield, customer ROI, or gross margin?
2. Does humanoid robotics commercialize first in bounded industrial/logistics settings, home settings, or internal vertically integrated factories?
3. If low-cost Chinese hardware expands robot experimentation, does value migrate to OEMs, component suppliers, data/model layers, or deployment services?
4. Will public-market investors reward S4 capacity/deployment signals before S5 economics, or demand margin / ROI proof first?

9. Public-safety flag

Public-safe if used as an evidence map.

Do not publish:

• Hugo private portfolio weights or private rationale.
• Buy / sell / hold language for TSLA, 9880.HK, 688017.SH, suppliers, or any public asset.
• Claims that designed capacity equals production.
• Claims that customer logos equal economics.
• Claims that low price equals margin or reliability.
• Claims that private-company funding/valuation proves commercial success.

10. Source list

• Tesla Q1 2025, Q2 2025, Q3 2025, Q4 2025, Q1 2026 Form 8-K Exhibit 99.1 updates. 🟢 Primary. As-of reviewed in existing Tesla Optimus evidence packets through 2026-06-11.
• Figure official BMW deployment post, 2025-11-19. 🟢 Primary.
• Figure BotQ / Figure 03 production post, 2026-04-29. 🟢 Primary.
• Figure Catalyst commercial agreement post, 2026-05-26. 🟢 Primary for announcement; 🟠 for economics because terms undisclosed.
• Unitree R1 / G1 / H2 / H2 Plus / G1-D official product pages, accessed 2026-06-10. 🟢 Primary for product specs/prices.
• UBTECH Robotics Corp Ltd Annual Report 2025, disclosed 2026-04-15. 🟢 Primary filing.
• UBTECH official Walker S / S1 / S2 product pages, accessed 2026-06-11. 🟢 Primary for product claims.
• Agility Robotics GXO multi-year agreement 2024-06-27; Digit 100,000+ totes milestone 2025-11-20; Mercado Libre 2025-12-10; Toyota Motor Manufacturing Canada RaaS 2026-02-19; Schaeffler 2024-11-13. 🟢 Primary company releases.
• Apptronik Mercedes-Benz 2024-03-15; GXO 2024-06-20; NVIDIA 2024-03-18; Google DeepMind 2024-12-19; Jabil 2025-02-25; Series A / Series A-X 2025-02-13 and 2026-02-11 official releases. 🟢 Primary company releases.
• 1X NEO product page and manufacturing page accessed 2026-06-11; “NEO Factory / Building Your NEO” 2026-04-30; NVIDIA collaboration 2025-03-18; World Model Lab 2026-06-04. 🟢 Primary company releases/pages.