Figure AI history / product / deployment timeline v1

Date: 2026-06-14 Owner: Finance / Charlie AGT-002 Status: public-safe source-backed research artifact Visibility: PUBLIC Target use: morning review, /robotics/companies, Figure section inside Tesla / Figure / Unitree module Public-safety: no Hugo portfolio data, no trade recommendation, no private channel checks, no paid-report excerpts

0. One-line answer

截至 2026-06-14，Figure AI 最适合被讲成一条“公司形成 -> 产品代际 -> 客户现场 KPI -> 制造爬坡 -> 第二客户场景”的证据曲线，而不是一个简单 winner ranking。最强公开证据仍是 BMW 11 个月现场部署、1,250+ runtime hours、90,000+ parts loaded、30,000+ BMW X3 contribution，以及 BotQ 350+ Figure 03、1/day 到 1/hour cadence、>80% EOL first-pass yield、99.3% battery-line first-pass yield。🟢 Figure official posts dated 2025-11-19 and 2026-04-29; 🟠 Charlie stage classification.

当前判断：Figure 已经越过 demo-only 叙事，进入 S4 customer-site deployment KPI + manufacturing KPI evidence；但截至 2026-06-14，公开证据仍未披露 robot count by customer、contract value、customer-confirmed ROI/payback、repeat order、long-duration uptime/intervention distribution、robot revenue、gross margin 或 service cost，因此不是 S5 scaled-commercial-economics proof。🟢/🟠

1. Core question

Figure 的研究问题不是“它是不是最强 humanoid 公司”，而是：

1. 它是否把 humanoid 从 demo 推到了可测量的客户现场工作？
2. 它的产品代际是否从 Figure 01/02 走向可制造、可测试、可维护的 Figure 03 fleet？
3. 它的制造爬坡是否对应真实客户利用率，而不是 build-ahead / demo-fleet risk？
4. 它的 Helix / fleet / data loop 是否能降低部署成本和 intervention rate？
5. 公开证据什么时候足够把 S4 升级为 S5？

2. Timeline: what Figure proves by stage

3. Product evolution reading

Figure 02: deployment proof vehicle

Figure 02 is currently most valuable as the BMW proof vehicle, not because public sources reveal a full cost model. The BMW disclosure turns Figure 02 into a measured deployment artifact: duration, shifts, runtime hours, part count, and vehicle-production context are disclosed. 🟢 Figure official post, 2025-11-19.

Stage label: S4 deployment KPI.

Missing before S5: customer-confirmed ROI/payback, robot count, intervention distribution, repeat deployment, contract value, and maintenance economics. 🟢/🟠 as of 2026-06-14.

Helix 02: autonomy claim, not deployment economics

Helix 02 matters because Figure claims a fully autonomous, no-teleop 4-minute household task and a learned whole-body controller trained on 1,000+ hours of human motion data. 🟢 Figure official post, 2026-01-27.

But this is still S3 unless the model reduces intervention or deployment cost in BMW / Catalyst-style sites. A 4-minute demo is not the same evidence unit as a 10-hour weekday industrial shift. 🟢/🟠

Figure 03 / BotQ: manufacturing and data-fleet bridge

Figure 03 / BotQ is the key bridge between “a robot can work” and “a fleet can be produced, tested, updated, and learned from.” The important numbers are not just 350+ robots; they are cadence, yield, actuator/battery output, inspection points, functional tests, OTA and fleet feedback. 🟢 Figure official post, 2026-04-29.

Stage label: S4 manufacturing-process evidence.

Open risk: manufacturing cadence can outrun economically validated customer deployment. If 350+ robots are not linked to contracted utilization, revenue/backlog, repeat orders, or data productivity, the signal remains S4 rather than S5. 🟢/🟠

4. Figure's evidence curve vs Tesla / Unitree

Figure should not be ranked against Tesla and Unitree on one score. It answers a different bottleneck.

Source basis: Tesla / Figure / Unitree public evidence ledger reviewed 2026-06-13; Figure official sources reviewed through 2026-06-14. 🟠 internal synthesis from 🟢 primary sources.

5. Signal vs noise

Signal

• Named customer + quantified runtime / task / shift metrics: BMW 1,250+ runtime hours, 90,000+ parts loaded, 10-hour shifts Monday-Friday. 🟢 S4.
• Manufacturing process metrics: 350+ robots, cadence from 1/day to 1/hour, >80% EOL first-pass yield, 99.3% battery-line first-pass yield, 9,000+ actuators. 🟢 S4.
• Fleet management language tied to health/status/location, OTA, and field-failure feedback. 🟢/🟠 S4 candidate because it may reduce deployment friction.
• Second named customer scenario with “commercial agreement” language and starting site. 🟢/🟠 S3/S4.

Noise unless upgraded

• Autonomy demos without long-duration field success distribution. 🟢 for claim / 🟠 for economics.
• Commercial agreement without robot count, contract value, deployment KPI, repeat order, or margin. 🟢 for announcement / 🟠 for economics.
• Production output without sell-through, utilization, revenue, or customer economics. 🟢 for production process / 🟠 for commercialization.
• “Figure has BMW, therefore humanoids are solved.” 🟠 overread.

6. What would change our mind

Upgrade toward S5 if primary or customer-side sources disclose:

1. BMW repeat deployment or second BMW site with robot count, uptime, intervention rate, and customer-confirmed ROI/payback. 🟢
2. Catalyst follow-through with deployed robot count, task KPI, deployment duration, contract economics, or repeat expansion beyond Reno. 🟢
3. BotQ output tied to contracted deployments, backlog, revenue, gross margin, field failure rate, and service cost. 🟢/🟠
4. Helix / fleet system shown to reduce intervention rate, deployment setup time, cycle time, or total cost in paid customer environments. 🟢/🟡.
5. Multi-customer deployment table showing repeatable playbook across manufacturing and logistics rather than one bespoke cell. 🟢/🟠

Downgrade if:

1. BMW remains the only deeply quantified deployment after another 6-12 months. 🟠
2. Catalyst remains a named agreement without robot count, task KPI, or follow-through. 🟠
3. Production cadence rises while field utilization, customer adoption, or fleet reliability stay undisclosed. 🟠
4. Helix remains demo-rich but does not lower intervention or deployment cost in real sites. 🟠
5. Support / maintenance burden offsets hardware or deployment gross margin. 🟠

7. Common misconceptions

1. Misconception: “Figure 有 BMW KPI，所以已经证明 scaled commercialization。”

   • Correction: BMW KPI 是强 S4，但 S5 还需要客户确认 ROI/payback、repeat order、intervention rate、robot count、contract value 和 margin。🟢/🟠

2. Misconception: “350+ Figure 03 delivered 就等于客户规模化部署。”

   • Correction: manufacturing output 是生产过程证据；商业化还需要 sell-through、utilization、customer economics 和 gross margin。🟢/🟠

3. Misconception: “Helix 无遥操作 demo 证明现场 autonomy 已经解决。”

   • Correction: 4-minute no-teleop demo 是技术证据；现场 autonomy 需要长时段、多任务、多客户、低 intervention rate。🟢/🟠

4. Misconception: “Catalyst commercial agreement 等于确定收入质量。”

   • Correction: commercial agreement 比 logo/MOU 强，但没有 robot count、contract value、deployment KPI 或 repeat order 时，仍不能当作 S5 economics。🟢/🟠

8. Think Deeper questions

1. Figure 的真正瓶颈在 robot body、AI/autonomy、制造、deployment service、还是 customer workflow integration？
2. 如果 BMW KPI 是 company-disclosed 而非 customer ROI disclosure，应该如何折扣证据强度？
3. BotQ 的 1 robot/hour cadence 如果成立，下一步最应该看 sell-through、field utilization、failure rate，还是 gross margin？
4. Figure 的 fleet feedback loop 会成为 data moat，还是只是 humanoid deployment 的基本运维成本？
5. 对 public-market 研究来说，Figure 私有化状态下的 S4 证据会迁移到哪些公开 value layer：compute、components、integrators、customers，还是没有干净 proxy？

9. Public-safe site draft section

Figure AI: from demo company to measurable deployment curve

Figure 的公开证据价值不在于“它赢了”，而在于它把 humanoid robotics 的两个关键问题变得可测量。

第一是客户现场。Figure 披露 BMW Spartanburg 的 11 个月 Figure 02 部署：10-hour shifts Monday-Friday、90,000+ parts loaded、1,250+ runtime hours，并贡献到 30,000+ BMW X3 vehicles。这个证据比 demo video 强，因为它有 named customer、真实生产线、运行时长和任务数量。

第二是制造爬坡。Figure 披露 BotQ delivered 350+ Figure 03 robots，产线 cadence 从 1/day 提到 1/hour，并给出 >80% EOL first-pass yield、99.3% battery-line first-pass yield、9,000+ actuators、80+ functional tests per robot。这个证据说明 humanoid 公司开始披露生产过程 KPI，而不只是未来产能愿景。

但结论必须克制：Figure 仍未公开 robot count by customer、contract value、customer ROI/payback、repeat order、uptime/intervention distribution、robot revenue、gross margin 或 service cost。因此 Figure 当前是 S4 deployment + manufacturing evidence，不是 S5 scaled-commercial-economics proof。

Footer: Evidence map only. No winner ranking. No trade recommendation. Deployment KPI and manufacturing KPI are stronger than demos, but they are not full customer economics.

10. Slide-ready compression

Title: Figure AI: the clearest deployment curve, still pre-S5 economics

Three-card layout:

1. Customer-site KPI

   • BMW: 11 months, 10-hour shifts Monday-Friday, 90,000+ parts, 1,250+ runtime hours, 30,000+ X3 contribution. 🟢
   • Signal: S4 deployment KPI.

2. Manufacturing KPI

   • BotQ: 350+ Figure 03, 1/day -> 1/hour, >80% EOL FPY, 99.3% battery-line FPY, 9,000+ actuators. 🟢
   • Signal: S4 production-process evidence.

3. Missing before S5

   • Robot count by customer, contract value, customer ROI/payback, repeat order, uptime/intervention, revenue, gross margin, service cost. 🟢/🟠

Footer: Evidence map only. No trade recommendation. BMW KPI ≠ full economics; production cadence ≠ utilized fleet.

11. Source list

• Figure, “F.02 Contributed to the Production of 30,000 Cars at BMW”, 2025-11-19: https://www.figure.ai/news/production-at-bmw 🟢
• Figure, “Introducing Helix 02: Full-Body Autonomy”, 2026-01-27: https://www.figure.ai/news/helix-02 🟢
• Figure, “Ramping Figure 03 Production”, 2026-04-29: https://www.figure.ai/news/ramping-figure-03-production 🟢
• Figure, “Figure Signs Agreement with Catalyst Brands to Scale Humanoid Operations”, 2026-05-26: https://www.figure.ai/news/figure-signs-agreement-with-catalyst-brands 🟢 for announcement / 🟠 for economics undisclosed
• Internal cross-reference: knowledge/robotics-figure-ai-deep-dive-v1.md, dated 2026-06-10. 🟠 synthesis from primary Figure sources.
• Internal cross-reference: knowledge/robotics-figure-unitree-morning-review-evidence-audit-v1.md, dated 2026-06-13. 🟠 synthesis from primary Figure / Unitree sources.
• Internal cross-reference: knowledge/robotics-tesla-figure-unitree-public-evidence-ledger-v1.md, dated 2026-06-13. 🟠 synthesis from primary Tesla / Figure / Unitree sources.

12. Public-safety flag

PUBLIC-safe if used as an evidence map. Do not include Hugo private portfolio data, private channel checks, paid-report excerpts, supplier rumors, or buy / sell / hold language. Do not imply Figure has proven scaled humanoid economics unless future primary/customer-side sources disclose customer ROI/payback, repeat orders, robot count, intervention rate, revenue, margin, and service-cost evidence.