Figure AI / Unitree Morning Review Evidence Audit v1

Date: 2026-06-13 Owner: Hugo / Genius Team Agent: Finance / Charlie AGT-002 Status: public-safe research artifact for morning review and Codex packaging Visibility: PUBLIC Related files:

• knowledge/robotics-research-harness.md
• knowledge/robotics-figure-unitree-evidence-v1.md
• knowledge/robotics-figure-ai-deep-dive-v1.md
• knowledge/robotics-unitree-deep-dive-v1.md
• knowledge/robotics-oem-three-evidence-curves-public-brief-v1.md

Public-safety: yes. No Hugo private portfolio data. No trade recommendation. No external publishing action.

0. One-line answer

截至 2026-06-13，Figure AI 和 Unitree / 宇树不应被压缩成“谁赢”的排序。更准确的公开表达是：Figure 正在提供更强的客户现场 / 制造 KPI 证据；Unitree 正在提供更强的硬件成本 / 开发者平台证据。两者都还缺 S5 scaled commercial economics：客户确认 ROI/payback、repeat orders、长期 uptime / intervention rate、合同金额、收入、毛利率和服务成本。🟢/🟠 synthesis from primary official sources checked in prior artifacts through 2026-06-10; reviewed for morning handoff 2026-06-13.

1. Why this artifact exists

The existing knowledge base already has thick evidence. The remaining morning-review risk is compression error:

1. Saying Figure has “proved commercialization” because BMW KPI exists.
2. Saying Unitree “wins” because its humanoid price points are low.
3. Treating deployment KPI and low-cost hardware as the same evidence type.
4. Converting a research framework into a company ranking.

This page gives Codex a compact, public-safe evidence audit for slide 11 or /robotics/companies.

2. Current evidence audit

3. Signal vs noise

Signal

• Figure BMW operating metrics: S4. Named customer + active assembly-line context + 1,250+ runtime hours + 90,000+ parts loaded is materially stronger than demo evidence. 🟢 Figure official post, 2025-11-19; checked in robotics-figure-ai-deep-dive-v1.md.
• Figure BotQ manufacturing metrics: S4. 350+ Figure 03 delivered, 1/day to 1/hour cadence, >80% EOL first-pass yield, 99.3% battery-line first-pass yield, and 9,000+ actuators show production-process disclosure, not only design intent. 🟢 Figure official post, 2026-04-29.
• Figure Catalyst agreement: S3/S4. Named commercial agreement and starting location are stronger than a generic logo, but volume/value/KPI are undisclosed. 🟢 for announcement / 🟠 for economics, Figure official post, 2026-05-26.
• Unitree R1/G1/H2 prices: S3. US$4,900 / US$13.5K / US$29,900 price anchors lower the experimentation-cost threshold. 🟢 Unitree official pages accessed 2026-06-10.
• Unitree H2 Plus / G1-D workflow signals: S3/S4. Jetson / Isaac / TeleOp / Sim and data / training workflow references matter if they become a standard developer surface. 🟢 for product-page claim / 🟠 for commercialization inference.

Noise unless upgraded

• Viral humanoid videos without runtime, reset count, failure rate, intervention rate, or customer context. 🔴/🟠
• Customer logos without robot count, value, timeline, task KPI, or repeat-order evidence. 🟢 for logo if official / 🟠 for economics.
• Low price treated as reliability, gross-margin, shipment-scale, or customer-ROI proof. 🟠
• Manufacturing output treated as economically utilized fleet. 🟢 for production claim / 🟠 for utilization inference.
• AI-stack integration treated as generalized autonomy or commercial moat proof. 🟢 for integration claim / 🟠 for value-capture inference.

4. Public-safe slide compression

Slide title

Figure vs Unitree: deployment proof vs cost-access proof

Subtitle

Figure answers “can a humanoid run measurable work at a customer site?” Unitree answers “how cheap and accessible can the embodiment become?” Neither yet answers full S5 economics.

Three-card layout

1) Figure = deployment KPI
BMW: 11 months, 1,250+ runtime hours, 90,000+ parts, 30,000+ X3 contribution.
Signal: S4 customer-site evidence.
Missing: contract value, robot count, ROI/payback, intervention rate, repeat orders.

2) Figure = manufacturing KPI
BotQ: 350+ Figure 03, 1/day -> 1/hour, >80% EOL FPY, 99.3% battery-line FPY.
Signal: S4 production-process evidence.
Missing: sell-through, utilization, revenue, gross margin, field failure rate.

3) Unitree = cost-access platform
R1 from $4.9k, G1 from $13.5k, H2 $29.9k; H2 Plus / G1-D dev/data workflow.
Signal: S3/S4 experiment-base expansion.
Missing: shipments, uptime, customer KPI, margin, support cost, ROI/payback.

Required footer:

Evidence map only. No winner ranking. No trade recommendation. Deployment KPI ≠ full economics; low price ≠ reliability or margin.

5. Public site section draft

Figure 和 Unitree：两条证据曲线，不是一个排行榜

Figure 和 Unitree 的价值在于它们分别把 humanoid robotics 的不同问题变得可量化。

Figure 的强项是客户现场和制造 KPI。截至 2026-06-13 早晨 review，最强公开锚点仍是 BMW Spartanburg：11 个月部署、10-hour shifts Monday-Friday、90,000+ parts loaded、1,250+ runtime hours、contribution to 30,000+ BMW X3 vehicles。BotQ 则给出制造侧指标：350+ Figure 03 delivered、production cadence from 1/day to 1/hour、>80% end-of-line first-pass yield、99.3% battery-line first-pass yield、9,000+ actuators。🟢 Figure official posts dated 2025-11-19 and 2026-04-29.

Unitree 的强项是成本曲线和开发者可获得性。官方产品页披露 R1 AIR from US$4,900 / R1 from US$5,900、G1 from US$13.5K、H2 US$29,900；H2 Plus 和 G1-D 则把 Jetson / Isaac / TeleOp / Sim / data acquisition / model training workflow 放进产品叙事。🟢 Unitree official product pages accessed 2026-06-10.

这两类证据都比“demo video”更有研究价值，但它们仍不是 S5。下一阶段真正改变结论的，不是更炫的视频，而是客户确认的 repeat deployment、ROI/payback、低 intervention rate、长期 uptime、shipment scale、收入、毛利率和服务成本。

6. What would change our mind

Upgrade Figure toward S5 if:

• BMW, Catalyst, or a third named customer confirms robot count, contract value, repeat orders, uptime, intervention rate, customer ROI/payback, or task-level cost savings. 🟢
• Figure links BotQ output to contracted deployments, revenue/backlog, fleet utilization, gross margin, or field failure rates. 🟢/🟠
• Multiple customer sites show a repeatable deployment playbook rather than one-off supported demos. 🟢/🟠

Downgrade Figure if:

• BMW remains the only deeply quantified deployment after another 6-12 months. 🟠
• Catalyst remains an announcement without robot count, task KPI, or repeat deployment. 🟠
• Production cadence rises faster than field utilization, creating inventory / demo-fleet risk. 🟠

Upgrade Unitree toward S5 if:

• Unitree discloses humanoid shipment volume by product line, customer mix, repeat-order rate, or gross margin. 🟢
• Named industrial customers disclose runtime, uptime, robot count, intervention rate, support burden, and payback. 🟢
• H2 Plus / G1-D becomes a measurable developer standard with active developer count, dataset scale, model ecosystem traction, or paid software/service metrics. 🟢/🟠

Downgrade Unitree if:

• Low-cost humanoids remain mainly demo / education / research devices without reliable industrial tasks. 🟠
• Support, warranty, maintenance, or safety cost makes headline pricing economically misleading. 🟠
• Value migrates to NVIDIA, open-source models, integrators, or customers while Unitree captures mostly commoditized hardware margin. 🟠

7. Common misconceptions

1. “Figure 有 BMW KPI，所以 humanoid 商业化已经解决。”

   • Correction: BMW KPI 是强 S4 deployment evidence，但不是 contract economics、repeat order、customer ROI、intervention rate 或 gross margin。🟢/🟠

2. “Unitree 便宜，所以 Unitree 一定赢。”

   • Correction: 低价降低实验门槛，也可能压缩硬件利润池；value capture 还取决于 shipment quality、software/data、deployment service、support cost 和 gross margin。🟠

3. “制造产出等于商业部署。”

   • Correction: 350+ robots / 1-hour cadence 是 manufacturing evidence，不等于 economically utilized fleet。🟢/🟠

4. “开发者工具链等于 AI moat。”

   • Correction: H2 Plus / G1-D 的 workflow 说明 Unitree 重视数据和模型循环，但 moat 需要 adoption、data scale、model performance、paid usage 或 switching cost。🟢/🟠

8. Think Deeper questions

1. If Figure discloses repeat orders before customer ROI, should that be enough to upgrade from S4 to near-S5?
2. If Unitree’s low price expands the developer base but hardware margins collapse, where does value migrate: data, model layer, integration, service, or customer workflows?
3. Which is more likely to lead the sector into S5: Figure-style customer KPI compounding, or Unitree-style shipment / platform scale?
4. What public metric would falsify the bullish reading faster: high intervention rate, weak repeat orders, warranty burden, or lack of gross margin disclosure?

9. Source list

Primary / official sources:

• Figure, “F.02 Contributed to the Production of 30,000 Cars at BMW”, 2025-11-19, checked in prior artifact 2026-06-10: https://www.figure.ai/news/production-at-bmw 🟢
• Figure, “Ramping Figure 03 Production”, 2026-04-29, checked in prior artifact 2026-06-10: https://www.figure.ai/news/ramping-figure-03-production 🟢
• Figure, “Figure Signs Agreement with Catalyst Brands to Scale Humanoid Operations”, 2026-05-26, checked in prior artifact 2026-06-10: https://www.figure.ai/news/figure-signs-agreement-with-catalyst-brands 🟢 for announcement / 🟠 for economics because value, robot count, ROI, and margin are undisclosed.
• Unitree R1 product page, accessed 2026-06-10 in prior artifact: https://www.unitree.com/R1 🟢
• Unitree G1 product page, accessed 2026-06-10 in prior artifact: https://www.unitree.com/g1/ 🟢
• Unitree H2 product page, accessed 2026-06-10 in prior artifact: https://www.unitree.com/H2 🟢
• Unitree H2 Plus product page, accessed 2026-06-10 in prior artifact: https://www.unitree.com/H2plus 🟢
• Unitree G1-D product page, accessed 2026-06-10 in prior artifact: https://www.unitree.com/G1-D 🟢

Internal synthesis cross-references:

• knowledge/robotics-figure-unitree-evidence-v1.md, dated 2026-06-10. 🟠 internal synthesis from primary sources.
• knowledge/robotics-figure-ai-deep-dive-v1.md, dated 2026-06-10. 🟠 internal synthesis from primary sources.
• knowledge/robotics-unitree-deep-dive-v1.md, dated 2026-06-10. 🟠 internal synthesis from primary sources.
• knowledge/robotics-oem-three-evidence-curves-public-brief-v1.md, dated 2026-06-11. 🟠 internal synthesis from primary sources.

10. Public-safe flag

Safe to publish if used as written:

• No Hugo private portfolio weights or private rationale.
• No buy / sell / hold language.
• No paid-report excerpts.
• No private-channel claims.
• Includes source grades and as-of dates for material claims.

Do not publish:

• “Figure wins” / “Unitree wins” ranking.
• “Figure has proven scaled humanoid economics.”
• “Unitree low price proves industrial reliability or margin.”
• Any trade recommendation or implication.