Figure AI / Unitree Slide 11 packaging spec — deployment KPI vs cost-access curve
Date: 2026-06-18
Owner: Hugo / Genius Team
Agent: Finance / Charlie AGT-002
Status: SYNTHESIS_CANDIDATE
Visibility: PUBLIC
Output target: slide 11 / /robotics/companies card module
Public-safety: yes. No Hugo private portfolio data. No buy / sell / hold language. No external publishing action.
Source discipline: 🟢 primary / 🟡 secondary / 🟠 estimate / 🔴 guess.
Related source files:
knowledge/robotics-figure-unitree-evidence-v1.mdknowledge/robotics-figure-unitree-warrior-brief-v1.mdknowledge/robotics-figure-unitree-source-to-slide-claim-control-2026-06-17.mdknowledge/robotics-tesla-figure-unitree-public-evidence-ledger-v1.mdknowledge/robotics-morning-review-evidence-upgrade-matrix-2026-06-18.md
0. One-line answer
截至 2026-06-18，Figure AI / Unitree 最适合在 slide 11 或 /robotics/companies 被压缩成一句话：Figure = deployment/manufacturing KPI curve；Unitree = cost-access/developer-platform curve。前者把 humanoid 从 demo 推到 named-customer quantified work；后者把 humanoid body 的实验门槛压到公开可购买价格带。但两者都仍缺 S5 scaled-commercial-economics proof：customer-confirmed ROI/payback、robot count by site、uptime/intervention distribution、repeat order、revenue、gross margin、warranty/support burden。🟢 Figure official posts dated 2025-11-19 / 2026-04-29 / 2026-05-26; 🟢 Unitree official product pages captured 2026-06-10; 🟠 Charlie synthesis as of 2026-06-18.
Slide-safe sentence:
Figure shows the stronger deployment KPI; Unitree shows the stronger cost-access KPI. Neither has yet shown the full economics chain.
Do not use:
Figure has proven humanoid commercialization; Unitree has proven low-cost humanoid adoption.
Reason: both sentences over-upgrade S3/S4 evidence into S5 economics. 🟢/🟠
1. Codex-ready slide module
Slide title
Tesla / Figure / Unitree: evidence curves, not one horse race
If slide 11 is only Figure vs Unitree, use:
Figure vs Unitree: deployment KPI vs cost-access curve
Subtitle
The useful comparison is not “who wins”; it is which bottleneck each company has made measurable.
Layout
Use two main cards plus one bottom guardrail strip.
| Card | Header | 3 quantified anchors | Evidence grade | Signal label | Required caveat |
|---|---|---|---|---|---|
| Figure | Deployment + manufacturing KPI | BMW: 11-month deployment; 10-hour weekday shifts; 90,000+ parts / 1,250+ runtime hours / 30,000+ X3 vehicles. BotQ: 350+ Figure 03 delivered; cadence 1/day -> 1/hour; EOL FPY >80% and battery-line FPY 99.3%. Catalyst: commercial agreement starting at Reno DLC. | 🟢 Figure official BMW post 2025-11-19; 🟢 BotQ post 2026-04-29; 🟢 Catalyst post 2026-05-26 | S4 deployment + manufacturing KPI | Missing robot count by customer, contract value, customer ROI/payback, uptime/intervention distribution, repeat order, revenue/margin, service burden. |
| Unitree | Cost-access + developer-platform KPI | R1 AIR from US$4,900 / R1 from US$5,900; G1 from US$13.5K; H2 US$29,900. H2 Plus references NVIDIA Jetson T5000, FP4 2,070 TFLOPS, Isaac GR00T / TeleOp / Sim. G1-D describes data acquisition, labeling, data-asset management, model training and inference tools. | 🟢 Unitree R1/G1/H2/H2 Plus/G1-D official pages captured 2026-06-10 | S3/S4 cost-access + platform-enablement KPI | Missing shipment scale by product line, industrial uptime, customer mix, gross margin, warranty/support cost, repeat orders, customer ROI. |
| Guardrail | What neither proves | S5 still needs accepted units, repeat deployment, customer-confirmed economics, low intervention, revenue, gross margin and service/warranty quality. | 🟢 absence in reviewed official sources; 🟠 classification | S5 evidence debt | Do not infer failure from missing disclosure; infer only that public evidence is not yet enough for S5. |
Visual recommendation
- Use one axis label: “Evidence type,” not “leaderboard.”
- Figure card color: deployment / factory KPI.
- Unitree card color: cost / developer access.
- Bottom footer must be visible, not hidden in notes.
Footer text:
Evidence map only. No winner ranking. No trade recommendation. Deployment KPI ≠ full economics; low price ≠ reliability or margin.
2. Public site card copy
Figure AI — strongest public deployment KPI in this pair
Figure’s public evidence is unusually concrete for a humanoid company. The BMW deployment disclosure says Figure 02 operated for 11 months on an active production line, ran 10-hour weekday shifts, loaded 90,000+ parts, accumulated 1,250+ runtime hours, and contributed to 30,000+ BMW X3 vehicles. BotQ adds manufacturing evidence: 350+ Figure 03 delivered, production cadence from 1/day to 1/hour, EOL first-pass yield above 80%, battery-line first-pass yield 99.3%, 500+ battery packs and 9,000+ actuators. 🟢 Figure official posts, 2025-11-19 and 2026-04-29.
Why it matters: this is stronger than a demo reel because it has duration, task count, named customer site and production-process metrics. 🟠 Charlie synthesis.
Boundary: it is still not S5 proof because public sources reviewed through 2026-06-18 do not disclose customer-confirmed ROI/payback, contract value, robot count at BMW/Catalyst, uptime/intervention distribution, repeat order, vendor gross margin or service burden. 🟢/🟠
Unitree / 宇树 — strongest public cost-access KPI in this pair
Unitree’s public evidence is different. Its official pages list R1 AIR from US$4,900, R1 from US$5,900, G1 from US$13.5K and H2 at US$29,900, tax/shipping excluded. H2 Plus references NVIDIA Jetson T5000, FP4 2,070 TFLOPS, Isaac GR00T, Isaac TeleOp and Isaac Sim; G1-D describes data acquisition, labeling, data-asset management, model training and inference tooling. 🟢 Unitree official product pages captured 2026-06-10.
Why it matters: low public price anchors and developer/data workflow pages can widen the experiment base for embodied AI and pressure the whole industry cost curve. 🟠 Charlie synthesis.
Boundary: low price does not prove industrial reliability, autonomy, shipment scale, gross margin, warranty/support economics, repeat orders or customer ROI. NVIDIA / Isaac alignment also does not prove Unitree captures the AI/model profit pool. 🟢/🟠
Correct public framing
Figure answers: can a humanoid produce measurable work in a bounded industrial setting? Current answer: S4 evidence, not S5 economics.
Unitree answers: can humanoid hardware become cheap and accessible enough to widen the experiment base? Current answer: S3/S4 evidence, not S5 economics.
The comparison is useful because the two curves test different bottlenecks. It is not a winner ranking. 🟠
3. Claim-control list
| Claim Codex may use | Grade | As-of date | Must cite | Must not infer |
|---|---|---|---|---|
| Figure has stronger deployment KPI than Unitree in the reviewed public set. | 🟢/🟠 | 2026-06-18 | BMW 2025-11-19, BotQ 2026-04-29, Catalyst 2026-05-26 | Not “Figure has proven scaled economics.” |
| Figure BMW metrics include 11 months, 10-hour weekday shifts, 90,000+ parts, 1,250+ runtime hours and 30,000+ X3 vehicles. | 🟢 | 2025-11-19 post, reviewed 2026-06-18 | Figure official BMW post | Not customer ROI, not contract value, not robot count. |
| Figure BotQ metrics include 350+ Figure 03 delivered, 1/day -> 1/hour cadence, >80% EOL FPY and 99.3% battery-line FPY. | 🟢 | 2026-04-29 post, reviewed 2026-06-18 | Figure official BotQ post | Not sell-through, not utilization, not margin. |
| Unitree has stronger cost-access KPI than Figure in the reviewed public set. | 🟢/🟠 | pages captured 2026-06-10, reviewed 2026-06-18 | Unitree R1/G1/H2 official pages | Not reliability, margin, adoption or ROI. |
| Unitree’s developer/data workflow surface is worth tracking. | 🟢/🟠 | pages captured 2026-06-10, reviewed 2026-06-18 | Unitree H2 Plus / G1-D pages | Not proof Unitree owns the AI software layer. |
| Neither company has public S5 scaled-commercial-economics proof in the reviewed source set. | 🟢/🟠 | 2026-06-18 | Absence across reviewed official sources + prior claim-control artifacts | Not proof of failure; only evidence insufficiency. |
4. Signal vs noise
Signal:
- Named customer + duration + task denominator: Figure BMW 11 months, 90,000+ parts, 1,250+ runtime hours. 🟢 / S4.
- Production-process KPI: Figure BotQ 350+ robots, cadence, first-pass yield, battery and actuator counts. 🟢 / S4.
- Named commercial expansion surface: Figure / Catalyst agreement starting at Reno DLC. 🟢 / S3/S4 until count/value/payback are disclosed.
- Public humanoid price anchors: Unitree R1 / G1 / H2. 🟢 / S3/S4.
- Developer/data stack surface: Unitree H2 Plus / G1-D. 🟢 / S3/S4.
Noise unless upgraded:
- Viral videos without duration, repetitions, interventions or site context. 🔴/🟠.
- Customer logos without robot count, contract value, task KPI or repeat deployment. 🟢/🟠.
- Price/spec pages treated as reliability or margin proof. 🟢/🟠.
- Production cadence treated as profitable utilized fleet. 🟢/🟠.
- “winner” framing before S5 economics. 🟠.
5. Acceptance criteria for Codex packaging
Codex packaging is acceptable only if all are true:
- The slide/site module keeps Figure and Unitree as evidence-type comparison, not winner ranking.
- Every numeric claim includes source-grade marker and date anchor in speaker notes or visible source footer.
- The footer says no trade recommendation and warns that deployment KPI / low price are not full economics.
- The module excludes Hugo private portfolio data, private rationale, private channel checks, paid-report excerpts and supplier rumors.
- The Figure card does not say “customer ROI,” “profitable,” “scaled deployment,” or “S5” unless future primary/customer sources disclose the missing economics.
- The Unitree card does not say “shipment scale,” “industrial adoption,” “high margin,” or “reliable paid deployment” unless future primary/customer sources disclose those data.
6. What would change this artifact
Upgrade Figure if public sources disclose at least 3 together: robot count by customer/site, paid contract value or revenue/backlog, repeat order or multi-site expansion, customer-confirmed ROI/payback, uptime/intervention/safety distribution, revenue/gross margin/service burden. 🟢/🟡 needed.
Upgrade Unitree if public sources disclose at least 3 together: humanoid shipments by product line, industrial customer mix, deployment runtime/uptime/intervention, repeat orders, product gross margin, warranty/support cost, paid software/data/tooling revenue. 🟢/🟡 needed.
Downgrade Figure if BMW remains a showcase, Catalyst remains terms-undisclosed, or BotQ output rises without field utilization and repeat-customer evidence. 🟠.
Downgrade Unitree if low-cost humanoids expand demos but not paid reliable deployments, or if support/warranty costs overwhelm low ASP hardware economics. 🟢/🟡/🟠.
7. Source list
Figure AI:
- 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, “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 because terms undisclosed.
- Figure, “Introducing Helix 02: Full-Body Autonomy”, 2026-01-27: https://www.figure.ai/news/helix-02 🟢 for company claim; 🟠 for commercial implication.
Unitree / 宇树:
- Unitree R1 product page: https://www.unitree.com/R1 🟢 captured 2026-06-10 in prior artifacts.
- Unitree G1 product page: https://www.unitree.com/g1/ 🟢 captured 2026-06-10 in prior artifacts.
- Unitree H2 product page: https://www.unitree.com/H2 🟢 captured 2026-06-10 in prior artifacts.
- Unitree H2 Plus product page: https://www.unitree.com/H2plus 🟢 captured 2026-06-10 in prior artifacts.
- Unitree G1-D product page: https://www.unitree.com/G1-D 🟢 captured 2026-06-10 in prior artifacts.
Internal synthesis / claim-control:
knowledge/robotics-figure-unitree-evidence-v1.md🟠knowledge/robotics-figure-unitree-warrior-brief-v1.md🟠knowledge/robotics-figure-unitree-source-to-slide-claim-control-2026-06-17.md🟠- Charlie packaging spec synthesis, 2026-06-18. 🟠
8. Public-safe flag
Public-safe as evidence map / packaging spec.
Exclude:
- Hugo private portfolio data, position weights, watchlist sizing, entry/exit logic or private rationale.
- Any buy / sell / hold language.
- Private channel checks, paid-report excerpts, unverified supplier/customer rumors.
- Any claim that Figure has proven scaled economics.
- Any claim that Unitree low price proves reliability, shipment scale, margin or ROI.