Figure AI

COMPELLING CPS 57

Developer of commercially autonomous humanoid robots powered by AI.

San Jose, CA, United States·Founded 2022·~120 emp·PRIVATE ↓ JSON ↓ MD
Researched 2026-02-16 ● Current
Figure AI — robotics.press intelligence card

Figure AI is among the most credible and heavily capitalized humanoid robotics startups globally, with rare evidence of sustained industrial deployment (BMW Spartanburg: 90,000+ parts, 1,250+ runtime hours, >99% placement accuracy) and a deliberate manufacturing scale-up strategy (BotQ). However, the $39B valuation is extreme relative to near-zero confirmed revenue, and the gap between pilot deployments and scaled, profitable commercial adoption remains vast—making this a high-conviction technology bet with outsized execution risk.

Moat NARROW

- Helix Vision-Language-Action AI stack with proprietary training data from real industrial deployments (BMW, logistics pilots) creating a data flywheel advantage - BotQ vertically integrated manufacturing facility with dedicated reliability engineering and supply chain structured for scale—a capital-intensive barrier to entry - Field deployment learnings from 1,250+ hours at BMW feeding directly into hardware redesign (F.03), creating an iterative advantage competitors without equivalent deployment data cannot replicate - Strategic investor ecosystem (NVIDIA, Microsoft, OpenAI) providing preferential access to compute infrastructure, cloud platforms, and AI model partnerships - Rapid hardware iteration capability (3 generations in ~24 months) enabled by ~$1.9B in funding and founder-led decision-making speed

Management STRONG

CEO Brett Adcock brings a proven founder-operator track record from Vettery (acquired) and Archer Aviation (public), with demonstrated ability to raise capital at scale ($1.9B cumulative) and attract talent from Boston Dynamics, Tesla, Apple, and Google. The aggressive iteration cadence (F.01 to F.03 in ~24 months) and willingness to publicly share deployment failure modes and learnings signal engineering discipline. However, Contrary Research has flagged concerns about AI bench depth relative to ambitions and the presence of multiple Archer/Vettery colleagues in growth roles at a pre-revenue stage.

Financials OPAQUE
Bull Case

Demonstrated real-world industrial deployment at BMW Spartanburg with quantified metrics: 90,000+ parts loaded, 37-second cycle times, >99% placement accuracy, and 1,250+ runtime hours over ~11 months—rare among humanoid robotics peers.

Helix AI stack showing measurable throughput improvements in logistics (handling time reduced to ~4.05s/package, ~95% barcode orientation accuracy) with 'Sport Mode' achieving 20-50% speed increases, connecting AI R&D to commercially relevant KPIs.

BotQ manufacturing facility targeting 12,000 units/year with vertical integration, dedicated reliability engineering, accelerated lifecycle testing, and a supply chain structured for 100,000 robots—an unusually mature manufacturing posture for a startup of this age.

Exceptionally strong strategic investor base (Microsoft, NVIDIA, OpenAI Startup Fund, Bezos, Intel Capital, Qualcomm, Salesforce) providing not just ~$1.9B in capital but ecosystem access to compute, cloud, and AI infrastructure critical for embodied AI development.

Rapid hardware iteration cycle (F.01 → F.02 → F.03 in ~24 months) with explicit field-to-factory feedback loops—e.g., F.02 forearm failure modes at BMW directly drove F.03 electronics re-architecture, demonstrating disciplined engineering learning.

Helix 02 (Jan 2026) represents a unified whole-body controller replacing 100,000+ lines of hand-coded control with learned models, trained on 1,000+ hours of human motion data—a potential step-function in autonomy if validated beyond lab demos.

Bear Case

$39B post-money valuation with no confirmed revenue figures and only pilot-stage deployments; BMW publicly stated no current robots in-plant and no timetable for reintroduction, indicating the gap between trial success and commercial adoption.

Unit economics remain unproven: the frequently cited ~$1,000/month RaaS price is unconfirmed by the company and likely underestimates full-cycle costs given current component prices, maintenance requirements, and early-stage reliability profiles.

Field reliability concerns persist—F.02's forearm subsystem was the top hardware failure point at BMW due to thermal constraints and dynamic cabling, and while F.03 addresses this, new failure modes will inevitably emerge at scale.

Manufacturing scale claims (100,000 robots in ~4 years) are extraordinarily ambitious and require proving yield rates, supplier reliability, after-sales service infrastructure, and cost-downs that have not yet been demonstrated.

Competitive field is well-funded and advancing rapidly (Agility Robotics, Apptronik, Unitree, Tesla Optimus); without sustained multi-site revenue-generating deployments, competitive moats remain theoretical rather than proven.

Questions raised about depth of AI research bench relative to ambitions (Contrary Research), and early commercialization team composition skews toward founder's prior network rather than domain-specific robotics go-to-market expertise.

Key Risks

Valuation-to-revenue disconnect: $39B valuation with no confirmed revenue creates extreme expectations for near-term commercial scaling that may not materialize

Reliability and safety: Humanoid robots in industrial and especially home environments face stringent uptime, safety certification, and liability requirements that remain unproven at scale

Manufacturing execution risk: Scaling BotQ from NPI to consistent high-yield throughput while managing 3 dozen unique commodities and specialized processes is a multi-year challenge

Customer concentration and pipeline conversion: Over-reliance on BMW pilot (now concluded) and exploratory UPS talks; no confirmed multi-site, multi-year commercial contracts disclosed

Technology generalization: Helix 02's long-horizon autonomy demonstrated only in controlled lab settings; real-world generalization across diverse, unstructured environments remains the core unsolved problem in embodied AI

Competitive dynamics: Tesla Optimus, Agility Robotics, and well-funded Chinese competitors (Unitree) could achieve comparable performance at lower cost or with stronger distribution channels

Catalysts

BotQ production ramp and first public throughput/yield metrics in 2026, validating manufacturing scalability claims

Conversion of UPS exploratory talks or other logistics partnerships into confirmed multi-site pilot or commercial deployments

BMW or other automotive OEM announcing expanded deployment or renewed contract with Figure robots, signaling industrial adoption beyond single-trial validation

Helix 02 demonstrated in sustained, intervention-free industrial operations (not just lab demos), with published uptime and autonomy metrics

Potential IPO preparation (TechMarketBriefs pre-IPO analysis suggests 2026 timeline), which would force financial disclosure and provide a market-based valuation test

Irreplaceability 3
Market Weight
Tech Differentiation
Operational Deployment
Strategic Momentum
Ecosystem Influence
Coverage Necessity
Fin. Valuation
Fin. Revenue
TypeDeep Research
Published2026-02-16
Length5,402 words · 22 min read
Sources298 sources cited

Generated by automated research. Cross-reference with primary sources before investment decisions.

BotQ Fixed · FIELDED · Launched 2025
└─ High-volume manufacturing facility for humanoid robot production with vertical integration of critical modules and digital manufacturing backbone. Targets mass production scaling with reliability engineering and accelerated lifecycle testing. BotQ is intended to eventually use Figure's own humanoid robots inside the assembly line for material handling and component transport, increasing flexibility and accelerating capacity ramp without rigid conveyor investments. Manufacturing processes for F.03 components include die-casting, injection molding, and stamping to reduce per-part cycle times versus prior CNC-heavy approaches.
Figure 01 Fixed · PROTOTYPE · Launched 2023
└─ Foundational humanoid robot prototype establishing the bipedal form factor and early manipulation capabilities. Demonstrates basic locomotion and dexterity foundations. Figure 01 is the foundational prototype from which Figure 02 and Figure 03 were iterated. Public footage and details emerged during 2023 following the company's exit from stealth in early 2023.
Figure 03 Fixed · PROTOTYPE · Launched 2025
└─ Home-oriented and crossover commercial humanoid robot redesigned for mass manufacturing and reliability. Features embedded palm cameras and fingertip tactile sensors for complex manipulation in diverse environments. F.03's electronics redesign was directly informed by failure modes discovered during the BMW Spartanburg deployment of F.02, specifically eliminating the forearm distribution PCB and dynamic cabling to simplify thermal management and improve reliability. The shift to die-casting, injection molding, and stamping is intended to dramatically reduce part cycle times and cost per unit consistent with automotive-like scaling logic. F.03 is underpinned by the Helix AI stack.
Helix Software · FIELDED · Launched 2024
└─ Vision-Language-Action (VLA) AI stack designed to unify perception, language understanding, and learned control from pixels to torques. Core software enabling autonomous task execution across humanoid platforms. Helix enhancements introduced in early 2025 for logistics include implicit stereo depth estimation, multi-scale visual representation (preserving fine object details and broader scene context), learned visual proprioception for self-calibration, and Sport Mode (test-time trajectory resampling). The system handles diverse package types including rigid boxes, padded envelopes, and deformable poly bags. The company claims improvements generalize across Helix use cases beyond logistics.
Figure 02 Fixed · LIMITED · Launched 2024
└─ Workforce-oriented humanoid robot refined for industrial tasks with integrated battery, multi-camera vision, improved compute, and advanced hand dexterity. Deployed in automotive manufacturing environments. Figure 02 was deployed at BMW Group Plant Spartanburg for an approximately 11-month period, eventually running 10-hour weekday shifts. A two-week pilot in August 2024 preceded the extended deployment. The forearm subsystem — tightly packaged, thermally constrained, with a microcontroller-based distribution PCB and dynamic cabling — was identified as the top hardware failure point, directly informing the F.03 redesign. BMW later clarified there were no currently active Figure robots at the plant and no established timetable for reintroduction.
Helix 02 Software · PROTOTYPE · Launched 2026
└─ Unified whole-body controller introduced January 2026 replacing extensive hand-tuned control code. Enables long-horizon autonomous task execution with hierarchical reasoning and learned balance control. Helix 02 introduces System 0, a learned whole-body balance and coordination layer trained on 1,000+ hours of human motion data plus sim-to-real reinforcement learning, which replaces over 100,000 lines of hand-tuned control code present in the prior Helix architecture. The unified controller spans from pixel-level perception to full-body joint torques. Long-horizon autonomy was demonstrated via a multi-minute dishwasher unload/reload sequence executed without resets.
Jeff Bezos Investor via Bezos Expeditions
Brett Adcock CEO
Till Reuter CEO of KUKA Robotics (Series A participant/investor)
Figure AI Press Contact
Combat Support L1
Mission planning L3 · C2 / Fleet Management
C2 / Fleet Management L2 · Autonomy & Software
Load carrying L3 · Logistics
Multi-sensor fusion L3 · Visual Detection
AI / Analytics L2 · Autonomy & Software
Data fusion L3 · AI / Analytics
Navigation L2 · Autonomy & Software
Computer vision L3 · AI / Analytics
Logistics L2 · Combat Support
Autonomy & Software L1
Detection L1
Obstacle avoidance L3 · Navigation
Predictive maintenance L3 · AI / Analytics
Visual Detection L2 · Detection

News & Analysis

5