Figure Robotics Company Overview May 2026

Figure AI is a San Jose, California-based AI robotics company founded in 2022 by serial entrepreneur Brett Adcock to build general-purpose humanoid robots that perform human-like tasks in factories, warehouses, and homes.[1]

Founding Story and Leadership

Brett Adcock, who previously co-founded Vettery (acquired by Adecco for $110 million) and Archer Aviation (publicly traded electric air-taxi company), launched Figure AI in early 2022 after identifying labor shortages and the opportunity to create embodied AI systems that operate in human-designed environments. He assembled an elite team with deep expertise from Boston Dynamics, Tesla, Google DeepMind, and Apple, boasting over 100 years of combined AI and humanoid robotics experience.[2]

• Adcock serves as Founder and CEO; the company emphasizes rapid hardware-software integration, drawing on his prior experience scaling complex autonomous systems at Archer.
• The team prioritizes end-to-end neural networks for perception, reasoning, and control rather than relying solely on external large language models.

What this means for competitors: Adcock’s track record in hardware startups and his ability to attract top robotics talent create a high barrier; new entrants must match both execution speed and talent density to challenge Figure’s momentum.

Funding History and Valuation Milestones

Figure has raised approximately $1.9–2.34 billion across rounds, with explosive valuation growth reflecting investor enthusiasm for humanoid robotics.[3]

• Series A (May 2023): $70 million at a $500 million valuation (led by Parkway Venture Capital).
• Series B (February 2024): $675 million at a $2.6 billion valuation, including Microsoft, NVIDIA, OpenAI Startup Fund, Jeff Bezos (Bezos Expeditions), Intel Capital, and others.
• Series C (September 2025): Over $1 billion (some reports cite up to $1.5 billion committed) at a $39 billion post-money valuation—roughly 15× growth in 18 months—led by Parkway Venture Capital with participation from Brookfield Asset Management, NVIDIA, Macquarie Capital, Intel Capital, Align Ventures, LG Technology Ventures, Salesforce, T-Mobile Ventures, Qualcomm Ventures, and others.[4]

Total publicly reported capital raised stands at approximately $1.9–2.34 billion as of late 2025/early 2026.

What this means for competitors: The massive capital infusion funds manufacturing scale-up (e.g., BotQ facility) and AI infrastructure, making it difficult for undercapitalized players to keep pace in data collection, training, and production.

Stated Mission and Positioning in the Humanoid Robotics Market

Figure’s mission is to develop general-purpose humanoid robots that address global labor shortages and perform a wide range of physical tasks in human environments, starting with commercial/industrial settings and expanding to homes. CEO Brett Adcock has stated: “The world was built for humans. So if we can create a robot that interacts with it in the same way, we can automate a huge range of tasks.”[5]

The company positions itself as the market leader in embodied AI and scalable humanoid deployment, emphasizing its proprietary Helix AI platform (for perception, reasoning, and control) and high-volume manufacturing over pure research or narrow-task robots. It has shifted from early OpenAI collaboration toward fully in-house models. In the broader market—projected by Goldman Sachs to reach tens of billions by 2035—Figure stands out for early commercial traction and rapid iteration (Figure 01 → 02 → 03), aiming to ship tens of thousands of units while competitors remain largely in prototype or demonstration phases.[3]

What this means for competitors: Figure’s focus on real-world deployment data and manufacturing scale creates a flywheel (more robots → more data → better models) that is hard to replicate without similar capital and customer access.

Key Partnerships, Deployments, and Commercial Progress

Figure announced a commercial agreement with BMW Manufacturing in January 2024 for staged deployment of humanoid robots in automotive production at the Spartanburg, South Carolina plant.[6]

• An 11-month Figure 02 deployment (2025) involved 10-hour shifts, loading 90,000+ sheet-metal parts, running 1,250+ hours, and contributing to the production of over 30,000 BMW X3 vehicles with high accuracy and reliability.[7]
• Initial OpenAI partnership (2024) focused on specialized AI models for language and task understanding; the collaboration later ended as Figure prioritized proprietary in-house models (e.g., Helix VLA). OpenAI’s startup fund remained an investor.[8]

Additional pilots and a second undisclosed commercial customer have been referenced, with ambitions to ship up to 100,000 robots over four years.

What this means for competitors: Securing and executing on marquee customers like BMW validates the technology in production environments and generates proprietary data advantages; rivals must demonstrate comparable real-world reliability to win similar deals.

Overall Company Profile Summary

As of May 2026, Figure AI stands as one of the most valuable and well-funded humanoid robotics companies globally, with a $39 billion valuation, nearly $2 billion in capital, proven automotive deployments, and a clear roadmap from industrial pilots to home-scale general-purpose robots powered by its Helix AI platform. Its rapid rise is driven by elite leadership, strategic investors, and a deliberate focus on scalable embodied intelligence rather than narrow automation.

For anyone evaluating entry or competition in humanoid robotics, Figure illustrates that success hinges on combining massive capital, top-tier talent, early commercial anchors, and continuous real-world data loops—advantages that compound quickly in this capital-intensive field.

Recent Findings Supplement (May 2026)

Figure AI has accelerated from pilot validation to early commercial production in 2026, with shipments more than doubling month-over-month and real factory shifts now running 20 hours continuously at BMW facilities.[1]

This marks a shift from the 2024–2025 research and demo phase to measurable output scaling, powered by the in-house Helix AI model and a purpose-built BotQ factory.

• April 2026 shipments reached ~240 units (up from ~150 total in 2025), with monthly figures progressing from ~60 in February to ~120 in March to ~240 in April.[2]
• BotQ factory achieved one robot every 90 minutes by April 2026, with initial capacity of 12,000 units per year and plans to scale toward 100,000 robots over four years.[2]
• As of May 2026, Figure robots are completing actual production shifts at BMW manufacturing sites, including 20-hour continuous operations on factory tasks.[1]

For competitors or new entrants, this demonstrates that rapid manufacturing ramp-up and proven multi-hour autonomous operation in high-stakes automotive environments now serve as the primary differentiators—demos alone no longer suffice for credibility.

Figure 03, launched October 9, 2025, represents a complete redesign optimized for both mass production and home/commercial use, retiring Figure 02 after its BMW Spartanburg pilot concluded in November 2025.[2]

The new model incorporates die-cast components, soft textile coverings, and wireless charging to support high-volume output while targeting everyday environments.

• Named TIME Best Invention of 2025 in December 2025.[2]
• In March 2026 controlled trials, Figure robots demonstrated eight distinct autonomous cleaning skills (wiping, sweeping, scrubbing, mopping, vacuuming, dusting, polishing, organizing).[1]
• Displayed at the White House on March 25, 2026, during a summit with representatives from 45 nations.[2]

This positions Figure as the first humanoid company to bridge industrial pilots with consumer-grade form factors, pressuring rivals to accelerate their own manufacturing redesigns or risk being seen as perpetual lab projects.

Figure ended its OpenAI collaboration in early 2025 and now develops its Helix vision-language-action (VLA) model entirely in-house, with CEO Brett Adcock publicly explaining the split in March 2026.[3]

Adcock stated the partnership delivered “very little” value beyond brand association and that OpenAI’s chatbot-oriented techniques did not transfer well to robotics.

• OpenAI’s internal humanoid ambitions created direct competition concerns, including information flow risks.[3]
• Helix now powers all current operations, including the 50-hour nonstop package-sorting demonstrations without teleoperation.

The move underscores a broader industry trend where leading humanoid firms must control their core intelligence stack, making it harder for pure-play robotics companies without deep AI talent to compete on reasoning and adaptability.

BMW’s Spartanburg pilot wrapped in November 2025 after supporting over 30,000 X3 vehicles, handling 90,000+ parts, and logging 1,250 operating hours; the company is now evaluating Figure 03 use cases while expanding its Physical AI program with other suppliers.[2]

BMW’s Leipzig EV plant pilot (announced February 2026) uses Hexagon Robotics’ AEON instead, but Figure remains a participant in BMW’s Center of Competence for Physical AI.[2]

A Brookfield partnership (announced September 2025) provides access to 100,000 residential units and hundreds of millions of square feet of commercial/logistics space for Helix training data under “Project Go-Big.”[2]

These real-world deployments validate the economic case for humanoids in structured industrial settings while highlighting the need for diversified partner ecosystems—new entrants must secure at least one major automotive or logistics anchor customer early to match Figure’s data and credibility advantages.

Figure is in active negotiations for a new $1.5 billion funding round at a $39.5 billion valuation as of May 2026, building on the September 2025 Series C that closed at $39 billion post-money with over $1 billion raised.[1][2]

Total capital raised stands at approximately $1.9 billion. A November 2025 whistleblower lawsuit alleging Figure 02 safety issues was followed by a company counter-suit in January 2026.

Sustained high valuations despite operational ramp-up signal strong investor conviction in humanoid timelines, but also raise the bar for any new competitor to demonstrate comparable capital access or production metrics within the next 12–18 months.

The global humanoid robot market is transitioning from prototype demonstrations to early commercial pilots in 2025–2026, with valuations and production scaling accelerating rapidly due to advances in embodied AI, cheaper actuators/sensors, and labor shortages in manufacturing, logistics, and services. Multiple analyst reports place the 2025–2026 market size in the low-to-mid single-digit billions of USD, with aggressive CAGRs of 30–50%+ projected through the 2030s as units ship in volume and average selling prices decline.[1]

Key data points include:
- MarketsandMarkets: USD 2.92 billion in 2025, rising to USD 15.26 billion by 2030 (CAGR 39.2%).[1]
- Fortune Business Insights: USD 4.89 billion in 2025 and USD 6.24 billion in 2026, reaching USD 165.13 billion by 2034 (CAGR 50.6%).[2]
- Precedence Research: USD 2.16 billion in 2026, expanding to USD 8.78 billion by 2035 (CAGR 16.91%).[3]
- Other forecasts range higher (e.g., ~USD 3–5 billion in 2026 per some aggregators) or more optimistic long-term (IDTechEx ~USD 29.5 billion by 2036).[4]

Growth drivers include real deployments (e.g., BMW, Amazon warehouses, Hyundai factories) validating ROI through uptime, payload, and integration without facility retrofits. Asia Pacific leads in volume (China’s Unitree and AgiBot shipped thousands in 2025), while North America dominates funding and AI software edges. Risks include high initial costs, reliability in unstructured environments, and regulatory/safety hurdles. For new entrants or investors, the window favors those with data moats (fleet learning) or manufacturing scale; pure hardware plays face commoditization from Chinese rivals.

Figure AI has emerged as a valuation and deployment leader among Western humanoid startups by leveraging partnerships (OpenAI, BMW) and its Helix vision-language-action model to move from pilots to near-autonomous operations. In 2025–2026, Figure demonstrated production relevance through an 11-month BMW Spartanburg deployment (90,000+ sheet-metal parts loaded, supporting 30,000+ X3 vehicles) and a landmark 30-hour fully autonomous warehouse shift (38,000+ packages, self-recovery via onboard cameras and Helix-02, no human intervention).[5]

The company introduced Figure 03 in October 2025 for improved natural movement and safety. Production is scaling at its BotQ facility toward 12,000 units annually, with ambitious plans for 100,000 robots deployed over four years and a Robot-as-a-Service model targeting ~USD 20,000 consumer pricing. Funding underpins this: over USD 1 billion committed in the September 2025 Series C at a USD 39 billion post-money valuation (led by Parkway, with NVIDIA, Brookfield, etc.), following a 2024 Series B of USD 675 million at USD 2.6 billion.[6]

This positions Figure to compete on general-purpose intelligence and enterprise trust rather than lowest-cost hardware.

Primary competitors span U.S. AI/software-focused players (Tesla, Figure, 1X, Apptronik) and hardware/manufacturing specialists (Boston Dynamics, Agility, Unitree), with capabilities converging on bipedal mobility, dexterous manipulation, and AI-driven autonomy but diverging sharply on deployment readiness and price.

• Tesla Optimus (Gen 2/3): ~1.73 m tall, ~20 kg payload, highly dexterous hands (22 DOF per hand reported), learns skills via video demonstration using the FSD AI stack. Internal factory testing (hundreds of units in 2026); production ramp starting late 2026 at Fremont (ambitious targets of 50,000–1 million units/year). No third-party sales confirmed yet; consumer/enterprise pricing targeted below USD 20,000 at scale. Tesla-internal resources only (no separate external funding disclosed).[7]
• Boston Dynamics Atlas (electric, 2026 version): 1.5–1.8 m, 56 DOF, 2.3 m reach, lifts 50 kg, IP67-rated, water-resistant, operates −20 °C to 40 °C; supports autonomous, teleop (VR), or tablet control with four-fingered hands. Production began immediately post-CES 2026 reveal; all 2026 units committed to Hyundai RMAC and Google DeepMind pilots, with plans for tens of thousands in Hyundai factories and a 30,000-unit/year factory. Backed by Hyundai (majority owner); no standalone humanoid valuation disclosed.[8]
• Agility Robotics Digit: ~1.75 m (5'9"), 35–50 lb payload, specialized end-effectors for totes/boxes, cameras/LiDAR/sensors, up to 8-hour battery with autonomous docking. First commercially deployed humanoid (Amazon, GXO since 2023; 100,000+ totes moved by late 2025); additional deals with Toyota and Mercado Libre. RoboFab factory in Oregon supports scaling. Total funding ~USD 640 million (USD 400 million Series C in 2025 at ~USD 2.1 billion valuation; Amazon, SoftBank, NVIDIA, DCVC).[9]
• 1X Technologies NEO: General-purpose home/industrial focus; lifts up to 70 kg reported, runs at 6.2 m/s, very quiet (~22 dB). Hayward, CA factory opened 2026 for vertical integration; preorders open at USD 20,000 outright or USD 499/month subscription, with early customer shipments targeted late 2026. Deal for up to 10,000 units to EQT portfolio companies (2026–2030). Total funding >USD 130 million (EQT Ventures, Tiger Global, OpenAI Startup Fund).[10]
• Apptronik Apollo: Optimized for repetitive industrial tasks (lifting, sorting, kitting, material handling); works safely alongside humans. Pilots active since 2024–2025 with Mercedes-Benz, Jabil, GXO; production and deployments scaling in 2026. Total funding ~USD 935 million+ across Series A rounds (2025–2026) at ~USD 5–5.5 billion valuation (Google, Mercedes-Benz, B Capital).[11]
• Unitree (H1/G1): G1 (compact ~1.27–1.32 m, 23–43 DOF, ~USD 16,000) excels in dexterity/research; H1 (full-size ~1.8 m, 3.3 m/s speed record) targets industrial mobility. Commercially available; high-volume production (targeting 20,000+ units in 2026, following 5,000+ shipped in 2025). Strong in China/research/education markets; lower price point drives volume. Limited public funding details (Chinese company with internal scaling).[12]

Deployment status shows a clear split: Agility leads in live commercial logistics hours, Figure and Tesla in high-profile pilots, Boston Dynamics in committed enterprise fleets, while 1X and Apptronik are scaling from pilots, and Unitree dominates unit volume. Real-world validation (e.g., tote throughput, part loading, autonomous recovery) is now the key differentiator over pure demos. By mid-2026, cumulative shipped units remain in the low thousands for Western players versus higher Chinese volumes.

Funding reveals extreme concentration and valuation inflation in 2025–2026, with top U.S. players raising nearly USD 4+ billion combined amid AI hype. Figure’s USD 39 billion valuation dwarfs others, followed by Apptronik (~USD 5 billion), Agility (~USD 2.1 billion). Chinese players like Unitree compete via cost and scale rather than disclosed VC rounds. Total robotics startup funding hit records in 2025, with humanoids capturing a growing share.[5]

This capital fuels production ramps but creates pressure for proven ROI; single-source high valuations are fragile without sustained deployments.

Competitive Comparison Table (as of mid-2026)

For competitors or new entrants, success hinges on closing the gap between pilot hours and scalable, reliable fleets while managing capital burn. Western leaders (Figure, Tesla, Apptronik) hold AI/software advantages for complex tasks, but Unitree’s price aggression and Agility’s early commercial traction highlight that hardware cost and real-world uptime will determine market share as the sector moves from hype to measurable productivity gains. Additional primary-source verification of shipment numbers and long-term reliability data would further strengthen these projections.

Recent Findings Supplement (May 2026)

The global humanoid robot market is shifting from pilot-stage experimentation to early commercial scaling in 2026, with shipments projected at ~90,000 units and revenue estimates ranging from $2.2–5 billion. Bank of America forecasts rapid growth to 1.2 million units shipped annually by 2030 and a cumulative population of 3 billion by 2060 (62% in homes).[1][2] Precedence Research values the 2026 market at $2.16 billion, growing at 16.91% CAGR to $8.78 billion by 2035.[3] Other estimates place 2026 revenue at $4–5 billion.[4]

This acceleration stems from AI advances enabling generalist capabilities, plummeting hardware costs (especially from Chinese players), and real factory/warehouse deployments generating operational data. Investment reached $4.3 billion in 2025 (up sixfold since 2018), with over 50 companies active and 150+ product launches recorded.[1]

Figure AI leads in valuation and AI sophistication but trails in deployed volume. Its $39 billion post-money valuation (September 2025 Series C of >$1 billion, total funding ~$1.9 billion) reflects investor bets on Helix AI for whole-body control and multimodal learning.[5][6] By April 2026, Figure had produced >350 Figure 03 units at its BotQ facility, ramping from 1 robot/day to 1/hour (24× throughput gain in <120 days) with >80% first-pass yield and plans for 12,000 annual capacity.[7]

BMW’s 11-month Figure 02 pilot at Spartanburg produced >30,000 X3 vehicles and logged 1,250 operational hours; scaling to Leipzig is underway.[2] Recent demos show fully autonomous 24–50-hour package sorting with zero teleoperation.[8] This data moat accelerates iteration faster than pure hardware plays.

What this means for competitors: Figure’s software edge and BMW validation set a high bar for industrial AI performance; others must match data volume or risk commoditization on hardware alone.

Tesla Optimus remains in internal R&D, leveraging vertical integration for cost leadership. As of early 2026, Optimus Gen 3 units operate inside Fremont and Giga Texas factories primarily for learning/data collection rather than productive tasks.[9][2] Tesla plans to convert the Fremont Model S/X line into an Optimus production facility targeting up to 1 million units/year capacity, with production starting late 2026 and external sales possibly then (consumer availability targeted 2027+).[2] Target price at scale is <$20,000–30,000.[10]

What this means for competitors: Tesla’s actuator/compute integration could deliver 30–40% cost advantages, pressuring everyone on price once external sales begin.

Boston Dynamics’ production-ready Atlas (Hyundai-owned) is the first to secure committed 2026 industrial fleets. The redesigned electric Atlas (56 DOF, superhuman agility) entered production in January 2026 at CES; all 2026 supply is allocated to Hyundai’s Robotics Metaplant and Google DeepMind, with additional customers in 2027.[11][12] Hyundai plans a new U.S. robotics factory for 30,000 units/year by 2028.[11]

What this means for competitors: Ownership by a global automaker provides guaranteed volume and real-factory integration expertise that pure-play startups lack.

Agility Robotics leads in actual commercial deployments. Digit (~75–100 units installed base) is the most commercially deployed humanoid as of mid-2026, with continuous real-world operation since 2023.[13] New 2025–2026 agreements include Mercado Libre (December 2025) and Toyota Motor Manufacturing Canada (February 2026, expanding to 7–10 units for tote unloading).[13] Total funding ~$640 million (including $400 million Series C in 2025 at ~$2.1 billion valuation); RoboFab capacity is 10,000 units/year.[14]

What this means for competitors: Proven warehouse ROI and operational data give Agility a head start on reliability; others must catch up on real customer contracts.

1X Technologies is first to consumer pre-orders and scaled household production. NEO (home-focused) capacity reached 10,000 units/year at its new California “NEO factory”; pre-orders sold out 10,000 units within days of October 2025 launch, with initial U.S. deliveries targeted late 2026.[15][16] A partnership with EQT targets up to 10,000 units across logistics/manufacturing/healthcare by 2030.[15] Price: $20,000 or $500/month subscription; autonomy improving via world models (teleoperation reduction planned for 2026).[17]

What this means for competitors: Consumer channel opens a massive TAM; competitors without home-specific safety/autonomy features will cede this segment.

Apptronik Apollo is scaling aggressively on massive new capital. Total funding reached ~$935 million+ via Series A (initial $415 million in 2025 + $520 million extension in February 2026) at $5 billion valuation.[18][19] New robot version debuts in 2026; focus is ramping production and expanding retail/manufacturing/logistics pilots (existing partners include Mercedes-Benz, NASA).[18]

What this means for competitors: Top-tier funding validates the category but increases pressure to convert capital into shipped units and revenue.

Unitree dominates on price and volume targets in China. Targeting 20,000 humanoid units in 2026; G1 (~$16,000) is commercially available with 23+ DOF, while new R1 starts at ~$5,900.[20] H1 targets industrial/research use. IPO application accepted March 2026 (target raise ~$608 million).[21] Recent demos include autonomous kung-fu routines and extreme cold-weather testing.

What this means for competitors: Sub-$10k pricing threatens to commoditize hardware; Western players must differentiate on AI/software or lose emerging-market volume.

Competitive Comparison Table (as of May 2026, post-Nov 2025 developments only)

Implications for market entrants or competitors: The window for differentiation is closing fast. Leaders with real deployments (Agility, Figure, Boston Dynamics) or consumer channels (1X, Unitree) are pulling ahead. Pure hardware plays risk being undercut by Unitree’s pricing; software/AI moats (Figure, Tesla, 1X) will determine long-term winners once fleets generate enough data for rapid improvement. Expect consolidation or partnerships as capital requirements rise and production ramps become the new battleground.

Figure AI’s robot lineup has progressed rapidly from the prototype Figure 01 (2023–early 2024) through the production-oriented Figure 02 (late 2024) to the home-focused Figure 03 (October 2025), with the company now operating fleets of Figure 02/03 units in factories and conducting residential pilots.[1]

The core technical shift is the move to a single end-to-end Vision-Language-Action (VLA) neural system—Helix—that maps raw pixels, language, and proprioception directly to whole-body motor commands, replacing hand-engineered control stacks.

Hardware Generations: From Industrial Prototype to Home-Ready Platform

Figure 01 was a proof-of-concept humanoid that performed basic chores (e.g., moving dishes) while relying on external compute and early OpenAI vision-language models. Figure 02 introduced a polished exoskeleton design, 16-DoF human-scale hands, six RGB cameras for 360° coverage, three onboard NVIDIA RTX GPUs, and a 2.25 kWh battery enabling up to 10 hours of operation with 25 kg payload capacity per arm; it entered real factory deployment at BMW.[2]

Figure 03, announced October 2025, is a complete redesign optimized for unpredictable home environments: 9 % lighter, tactile fingertip sensors detecting forces as small as 3 g, embedded palm cameras, 60 % wider field-of-view cameras with double the frame rate and one-quarter the latency, soft multi-density foam and washable textiles for safety, 2 kW wireless inductive charging via foot coils, and a larger 4× more powerful speaker. These changes enable safer, more dexterous interaction in cluttered domestic spaces while supporting continuous operation through periodic docking.[3]

This hardware evolution directly supports longer-horizon autonomy because richer, lower-latency sensing (tactile + palm vision) feeds the neural controller without requiring separate perception pipelines.

• Figure 02 achieved 400 % faster task completion than Figure 01 at BMW and contributed to the production of 30,000 cars.[4]
• Figure 03’s reduced part count and shift to tooled manufacturing (die-casting, injection molding) enable BotQ, Figure’s dedicated high-volume facility targeting 12,000 units/year initially and 100,000 robots over four years.[3]

For any competitor entering the space, matching Figure’s closed-loop hardware–AI co-design (sensors, actuators, and compute purpose-built for the Helix neural stack) is now table stakes; off-the-shelf components will lag in latency and data quality for end-to-end learning.

AI Evolution: OpenAI Collaboration → In-House Helix VLA

Figure initially partnered with OpenAI for speech-to-speech reasoning and early visual-language capabilities on Figure 02. The collaboration ended in early 2025 when Figure cited integration friction and pivoted to fully in-house models after a major internal breakthrough.[5]

Helix, publicly introduced February 2025, is a generalist Vision-Language-Action model that unifies perception, language understanding, and continuous control. It uses a two-system architecture: System 2 (7 B-parameter VLM running at 7–9 Hz) produces a semantic latent vector from images + language; System 1 (80 M-parameter visuomotor transformer at 200 Hz) maps that latent plus raw pixels and state into 35-DoF continuous actions for the upper body. A single set of weights handles all tasks without per-task fine-tuning.[6]

Helix 02 (January 2026) extends this to full-body loco-manipulation with a three-tier hierarchy: System 0 (10 M-parameter neural prior trained on >1,000 hours of retargeted human motion data, running at 1 kHz for balance and contact) provides human-like whole-body priors; System 1 connects every sensor (head cameras, palm cameras, fingertip tactile, proprioception) to every actuator; System 2 handles longer-horizon semantic sequencing. This replaces over 109,000 lines of hand-engineered C++ with a unified neural system.[7]

The mechanism—direct pixels-to-torque via hierarchical latents—eliminates brittle handoffs between perception, planning, and control modules, which is why Figure can demonstrate multi-minute autonomous tasks that previously required resets or teleoperation.

• Helix enables zero-shot generalization to thousands of unseen household objects via natural language (e.g., “pick up the desert item”).
• Multi-robot collaboration runs identical weights on two robots for shared tasks such as collaborative grocery storage.[6]

Publicly Demonstrated Capabilities

Figure 02/Helix units have completed 8-hour autonomous factory shifts sorting packages and 24-hour continuous livestreams at ~21 packages per minute. Helix 02 on Figure 03 hardware has executed a 4-minute, 61-action dishwasher unload/reload cycle across a full kitchen with locomotion, bimanual transfers, whole-body coordination (e.g., hip/foot use), and implicit error recovery—no human intervention. Additional dexterity demos include unscrewing bottle caps with force-regulated tactile grip, extracting single pills from clutter using palm vision, and dispensing precise 5 ml syringe volumes.[7]

Home-oriented demos show autonomous living-room and bedroom tidying: spraying/wiping surfaces, scooping toys, operating a TV remote, rearranging pillows, and reorganizing objects.[8]

These results demonstrate that a single neural VLA, trained on ~500–1,000 hours of high-quality teleoperated human data plus simulation, can already deliver commercially relevant endurance and dexterity in both structured factory and semi-structured home settings.

Key Technical Differentiators vs. Competitors

Figure’s approach is vertically integrated end-to-end learning: one neural network family (Helix) trained on human demonstration data scales across tasks, embodiments, and multi-robot scenarios without task-specific code or robot-specific retraining. In contrast:

• Tesla Optimus emphasizes manufacturing scale and its own factory data loops but has not yet shown equivalent long-horizon VLA autonomy or multi-robot zero-shot collaboration.
• Apptronik (Apollo) leverages NASA-derived hardware heritage and modular designs for logistics but relies more on traditional control stacks.
• Boston Dynamics Atlas excels in dynamic athletic locomotion yet has not demonstrated comparable language-conditioned, data-driven generalization for manipulation sequences.
• 1X prioritizes tendon-driven compliance for home safety but uses a different actuation paradigm and has not published equivalent unified VLA results.[9]

Figure’s data efficiency (<5 % of prior VLA dataset sizes) and onboard embedded-GPU deployment further differentiate it for fleet-scale learning.

Known Limitations and Realistic Outlook

Public demos, while impressive, remain in relatively controlled or semi-structured environments; truly open-ended home chaos at scale is still under validation. Early Figure 02 units required human oversight for edge cases, and full 24/7 home autonomy will depend on continued data scaling and sim-to-real robustness. Production is ramping but costs are estimated at $50k–$100k per unit initially, addressed via robot-as-a-service models. Dexterity continues to advance with tactile sensing, yet critics note that material compliance and high-bandwidth tactile feedback remain active research areas across the industry.[10]

For competitors, the bar has been raised: any viable entrant must now demonstrate comparable end-to-end neural control, multi-hour autonomy, and data-efficient generalization rather than relying on classical robotics pipelines or narrow task specialization. Figure’s trajectory shows that closing the loop between high-quality human data, hierarchical VLAs, and purpose-built hardware can compress years of traditional robotics progress into months of iteration.

Recent Findings Supplement (May 2026)

Figure AI has shifted to fully in-house Helix models post-OpenAI partnership, with Helix 02 (January 2026) introducing a three-tier neural hierarchy for end-to-end whole-body control.[1]

This replaces modular controllers with a unified pixels-to-torque system, enabling longer autonomous tasks on Figure 03 hardware.

• Helix 02 adds System 0 (S0): a 10M-parameter network running at 1 kHz that learns balance, contact, and coordination from >1,000 hours of retargeted human motion data via simulation-trained reinforcement learning.[1]
• System 1 (S1) runs at 200 Hz as a visuomotor policy linking all sensors (head/palm cameras, fingertip tactile, proprioception) to 35+ joint actuators.[1]
• System 2 (S2) handles slower semantic reasoning and goal sequencing at 7–9 Hz.[2]
• New Figure 03 sensors (palm cameras for occluded views, fingertip force sensing down to 3 g) unlock dexterity like unscrewing caps, extracting pills, or precise syringe dispensing.[1]

This architecture means Figure no longer needs external language models for core control; all behaviors emerge from a single learned policy stack. Competitors relying on separate locomotion/manipulation stacks or heavy teleoperation face integration friction that Figure sidesteps through unified training.

Figure 03 production scaled dramatically in early 2026, delivering >350 units and reaching 1 robot/hour output by late April.[3]

The BotQ facility achieved this via dedicated module lines, 150+ networked workstations, >50 inspection points, and >80 functional tests per unit (including stress/burn-in).

• First-pass yield >80% and battery yield at 99.3%; >9,000 actuators produced across SKUs.[3]
• Perception-conditioned S0 now ingests RGB stereo for 3D world models, enabling zero-shot stair traversal and terrain recovery without calibration.[3]
• OTA updates, Fleet Management System, and real-world data loops from the growing fleet accelerate long-tail robustness.[3]

For new entrants, the barrier is no longer just hardware cost but matching Figure’s closed-loop data flywheel: every deployed robot generates training data that directly improves the next fleet’s autonomy.

Helix 02 demonstrated the first multi-humanoid collaborative loco-manipulation from a single shared neural policy in May 2026.[4]

Two robots reset an entire bedroom in under two minutes (doors, hanging clothes, making the bed, trash removal, etc.) while inferring each other’s intent purely from motion, with no shared planner or messaging.

• Tasks include handling deformable objects (comforters) and dynamic re-planning in fast-changing scenes.[4]
• Earlier January 2026 demo: uninterrupted 4-minute dishwasher unload/reload across a full kitchen, using whole-body coordination (hip/foot assists) and implicit error recovery.[1]

This shows Figure’s end-to-end approach scales to multi-agent settings without explicit coordination code—something modular or teleop-heavy competitors have not publicly matched at this horizon length.

In mid-May 2026, Figure 03 robots powered by Helix 02 completed >17-hour fully autonomous warehouse shifts, processing >22,000 packages.[5]

Livestreamed tasks included barcode scanning, box picking/reorientation, and workflow movement at human-level speed, with automatic battery-swap role changes for continuous operation.

• Related 24-hour livestreams reportedly exceeded 30,000 packages sorted.[6]
• No teleoperation; all control via the unified Helix 02 policy.[5]

Endurance at this scale validates the data advantage: fleet hours directly harden policies against edge cases that simulation alone cannot cover.

Figure’s core differentiator remains the shift from hand-engineered controllers (109k+ lines of C++ replaced by S0 neural prior) to a single learned VLA hierarchy.[1]

Unlike many competitors that layer separate vision, planning, and control modules or rely on teleop for complex behaviors, Figure trains one policy stack end-to-end in simulation then transfers zero-shot, now augmented by onboard perception and tactile feedback.

• OpenAI collaboration ended in early 2025; all recent advances are in-house Helix only.[2]

New competitors must either replicate this full-stack neural integration or accept slower iteration when adding hardware features that Figure already folds into its unified training loop.

Public demos still reveal limitations: occasional glitches requiring self-reset in long runs, and the company-run nature of endurance tests means independent verification is limited.[5]

Long-tail failures are addressed via fleet data rather than solved outright, and deformable-object or high-speed multi-robot scenes remain challenging despite progress.

Anyone entering the space should treat these as signals of remaining sim-to-real and robustness gaps—investing in closed-loop real-world data collection at scale is now table stakes rather than optional.

Figure AI’s flagship commercial relationship is its January 2024 milestone-based agreement with BMW Manufacturing, which produced the industry’s first publicly documented, multi-month production-scale deployment of humanoid robots.[1]

In an 11-month run at BMW’s Spartanburg, South Carolina plant (announced November 19, 2025), Figure 02 robots worked daily 10-hour shifts on an active assembly line, loading more than 90,000 sheet-metal parts across 1,250+ runtime hours and contributing directly to the production of over 30,000 BMW X3 vehicles. The core task was precise pick-and-place of sheet-metal components onto welding fixtures (within a 5 mm tolerance in roughly 2 seconds per part), followed by integration with six-axis industrial welders. The robots met an 84-second cycle-time target with >99% placement accuracy per shift and near-zero human interventions.[2]

This deployment moved from off-hours testing to live production, generated reliability data that directly informed the redesign of Figure 03 (e.g., re-architected wrist electronics and thermal management), and validated humanoid viability for repetitive, high-precision material-handling roles that are physically demanding for humans. BMW has not disclosed robot quantities, contract value, or duration; early phases involved small numbers or a single unit before scaling to daily operations.[3]

A second unnamed commercial customer—described by CEO Brett Adcock as “one of the biggest U.S. companies”—was signed by early 2025, with Figure 02 units delivered to a paying client by late 2024, marking Figure’s transition to revenue generation.[4]

Adcock stated that the two customers together create a credible path to shipping 100,000 humanoid robots over the next four years, driven by high-volume potential that accelerates cost reduction and real-world AI data collection. The second customer’s use case is not publicly detailed but aligns with logistics or large-scale industrial operations. No additional named enterprise customers have been announced as of May 2026. A strategic partnership with Brookfield (September 2025) focuses on building diverse real-world training data from residential and commercial properties rather than direct robot deployments.[5]

Figure is primarily targeting automotive manufacturing and logistics/warehousing today, with longer-term expansion into broader industrial, electronics assembly, aerospace, and eventually home/service environments.

In automotive settings, robots perform repetitive, precision material-handling tasks such as sheet-metal loading for welding lines. In logistics demonstrations (including 24–30-hour autonomous runs at Figure’s facilities in 2026), units have sorted and processed 22,000–38,000 packages per shift using onboard vision and the Helix neural network, with automatic recovery from errors and no teleoperation. These point to future warehouse roles in picking, sorting, and palletizing. Public statements and analyst commentary consistently highlight manufacturing and logistics as the near-term beachhead because of structured yet variable environments, high labor costs, and clear ROI from 24/7 operation.[6]

Publicly reported revenue and order-pipeline figures remain sparse and largely speculative, with no official disclosures from Figure or major sell-side analysts.

Secondary sources estimate 2025 pilot-related revenue in the low single-digit to ~$100 million range (the higher figure appears in aggregator analyses and lacks primary corroboration). Shipments in 2025 are estimated around 150 units; 2026 production cadence has reportedly doubled monthly, with one analysis projecting low tens of millions in annualized recurring revenue (ARR) under a hypothetical robots-as-a-service model. Figure closed a Series C round exceeding $1 billion in September 2025 at a $39 billion post-money valuation (total funding ~$1.9–2.3 billion), reflecting investor bets on future scale rather than current financials.[7]

For competitors or new entrants, the BMW results establish a concrete performance benchmark—>99% accuracy, met cycle times, and measurable vehicle output—that any rival must match or exceed in a live factory setting. Success hinges on rapid iteration from pilot data (as Figure did for Figure 03), securing high-volume logistics or manufacturing customers to drive learning loops, and achieving cost curves that support $1,000/month or lower effective pricing at scale. The absence of named follow-on customers beyond BMW and one large undisclosed U.S. firm underscores that the market remains in an early validation phase, where proven uptime, integration ease with existing industrial systems, and data advantages will determine who captures the first meaningful order backlogs.

Recent Findings Supplement (May 2026)

Figure AI's BMW partnership has produced concrete, quantifiable results at Spartanburg but has not yet expanded to new Figure deployments in Europe.[1]

In 2025, Figure 02 humanoids ran 10-hour weekday shifts for 11 months at BMW’s Spartanburg, South Carolina plant. They accumulated ~1,250 operational hours, loaded more than 90,000 sheet metal parts, and contributed to the production of over 30,000 BMW X3 vehicles.[2][3]

BMW publicly referenced this success when announcing its first European humanoid pilot at the Leipzig plant in February 2026. However, the Leipzig tests (initial deployment December 2025, broader tests from April 2026, full pilot summer 2026) use Hexagon Robotics’ wheeled AEON humanoid for high-voltage EV battery assembly and component manufacturing—not Figure robots.[4][5]

This indicates BMW is validating multiple humanoid platforms in parallel rather than committing exclusively to Figure for its European expansion.

Figure AI’s May 2026 livestreamed warehouse demo serves as its strongest public proof point for logistics-scale deployments.[2]

On May 13–14, 2026, multiple Helix-02-powered humanoids performed continuous, fully autonomous package sorting at Figure’s San Jose headquarters. The robots scanned barcodes, picked and oriented boxes, and placed them on a conveyor at near-human speed (~3 seconds per package). One run logged at least 17–30 hours with zero failures reported by the company, processing 22,000–30,000+ packages while robots swapped in from charging stations.[6]

CEO Brett Adcock framed the demo explicitly as evidence that humanoids can sustain full shifts (including 24-hour operations) in real warehouse or factory environments.

Targeted industries remain manufacturing and logistics, with tasks centered on repetitive physical handling rather than high-dexterity or decision-heavy work.[1]

• Automotive manufacturing (BMW): Loading sheet metal parts and supporting assembly-line production.
• Logistics/warehousing: High-volume, repetitive package sorting, barcode scanning, and material movement on conveyors—tasks that directly address labor shortages in fulfillment centers.
• Emerging/home: Figure 03 (launched October 2025) targets household tasks such as loading dishwashers, folding laundry, and general navigation, but these remain demonstration-only with no commercial pilots reported.

No new named commercial customers beyond the established BMW relationship have been announced since late 2025. References to GXO Logistics or Amazon appear only in general industry commentary, not as confirmed Figure deployments.[7]

Public revenue and order-pipeline figures remain minimal and largely estimated; the company’s $39 billion valuation (September 2025 Series C) is driven by investor bets on future scale rather than current bookings.[8]

Analysts estimate 2025 shipments around 150 units and project 2026 annualized recurring revenue (RaaS model) in the low tens of millions even at a ramped production rate of ~240 units per month and hypothetical $1,000/month pricing. No specific order backlog, contract values, or confirmed pilot revenues have been disclosed in recent reporting.

Figure’s strategy prioritizes demonstrating reliability and endurance (via the May 2026 warehouse run) to accelerate enterprise interest in manufacturing and logistics, while BMW’s multi-vendor approach at Leipzig shows that automotive customers are still testing rather than scaling any single humanoid supplier.

For competitors or new entrants, the key takeaway is that public validation still rests on a single high-profile manufacturing pilot (BMW Spartanburg) plus internal proof-of-concept demos, not a broad portfolio of paid, multi-site commercial deployments.[9]

Success metrics emphasized are operational hours logged and packages handled without human intervention, not yet dollars of revenue or units under long-term contract. Companies seeking to compete should focus on matching or exceeding these endurance and autonomy benchmarks in customer-owned facilities rather than headquarters demos.

US regulators rely on the General Duty Clause and evolving consensus standards rather than humanoid-specific rules, forcing integrators like Figure AI to treat every deployment as a custom risk-assessment exercise. OSHA has no dedicated robotics standard, so it enforces workplace safety through Section 5(a)(1) of the OSH Act, which requires employers to keep workplaces “free from recognized hazards.” Inspectors cite failures to follow recognized consensus standards (such as ISO 10218) as evidence of a violation, even without a specific humanoid rule.[1]

• The 2025 updates to ANSI/A3 R15.06-2025 (U.S. adoption of ISO 10218-1/2:2025) integrate collaborative-robot force limits, add explicit cybersecurity risk assessments, and shift emphasis from hardware alone to full application certification (robot + task + workspace + human workflow).[1]
• A new working draft, ISO 25785-1 (expected 2026–2027), specifically targets dynamically stable walking robots; it defines fall-zone calculations based on height, speed, and load and requires “zero-energy” safe poses on power loss. Agility Robotics and Boston Dynamics experts are contributing.[1]
• Compliance documents routinely demanded: site-specific risk assessments, energy-control procedures (with alternatives to full LOTO for powered robots), operator training records, and marked fall zones. One Midwest manufacturer was fined $15,400 per robot ($107,800 for seven) for lacking these.[1]
• UL 3300 is now on OSHA’s NRTL list for consumer/commercial robots; Agility’s Digit has already passed an OSHA-recognized field inspection.[2]

Implication for competitors or new entrants: Any company scaling humanoids must budget for third-party NRTL audits and site-specific re-assessments; skipping them risks OSHA citations, voided insurance, and project delays.

The EU AI Act classifies most industrial and workplace humanoid deployments as high-risk AI systems, triggering mandatory risk management, documentation, human oversight, and post-market monitoring that will not fully apply until 2026–2027. Classification depends on use case, not robot morphology: workplace applications, safety-critical functions, or systems covered by Annex I product legislation (including machinery) qualify as high-risk.[3]

• High-risk obligations for Annex III systems become enforceable August 2026; Annex I-linked systems follow in 2027. Requirements include risk-management systems, high-quality data governance, technical documentation, logging capabilities, human oversight, robustness/cybersecurity, and ongoing monitoring.[4]
• Overlaps with the Machinery Regulation (EU) 2023/1230 (full application January 2027), which mandates CE marking and conformity assessment for physical safety, plus the Cyber Resilience Act and updated Product Liability Directive.[3]
• No dedicated “humanoid” category exists; safe human–robot interaction, emergency stop mechanisms, and handling of autonomous/learning behavior are explicitly called out for high-risk systems.

Implication: European market entry requires parallel conformity assessments under both AI Act and Machinery rules, lengthening certification timelines and raising compliance costs relative to purely U.S. deployments.

Labor unions are already demanding formal agreements before humanoid deployments, while public sentiment remains fragile and easily swayed by visible malfunctions. South Korea’s Hyundai Motor union has publicly warned against factory humanoid use without labor-management consultation.[5] European unions, protected by strong employment laws, are positioned to slow adoption through negotiations or calls for automation taxes.[6]

• Public reactions mix excitement (e.g., Figure AI warehouse livestreams, JAL airport pilots) with skepticism fueled by viral videos of robots freezing, flailing, collapsing, or behaving unpredictably in demos.[7]
• Broader trust issues include over-reliance on robots, privacy concerns from onboard sensors, and discomfort when robots operate near untrained members of the public.

Implication: Large-scale pilots can trigger organized pushback or negative media cycles that stall funding or customer commitments; early transparency and union engagement are now table stakes.

The most concrete documented safety concern tied to Figure AI itself comes from a November 2025 whistleblower lawsuit filed by former principal robotic safety engineer Robert Gruendel. Gruendel alleges he was terminated in September 2025, days after raising detailed concerns that the company’s humanoids could move at “superhuman speed,” apply force “twenty times higher than the threshold of pain,” and fracture a human skull.[8]

• The suit claims one robot malfunctioned and carved a ¼-inch gash into a stainless-steel refrigerator door; Gruendel asserts Figure initially lacked formal safety procedures, incident-reporting systems, or risk-assessment processes.[9]
• Figure AI denies the allegations, stating Gruendel was terminated for poor performance and has countersued; the case remains pending.[10]

Broader real-world signals include navigation failures and user discomfort in Boston Dynamics Spot hospital deployments, erratic demo behavior across multiple humanoid platforms, and historical industrial-robot incidents (e.g., a 2021 Tesla Giga Texas arm pinning an engineer). No mass casualty humanoid events have been publicly reported, but early failures in unstructured environments are already generating negative publicity that can pause programs.[11]

Implication: Even unproven allegations or demo glitches can generate regulatory scrutiny, insurance hikes, and customer hesitation—especially for a high-visibility company like Figure AI.

Four interlocking barriers—safety gaps in existing standards, liability uncertainty, regulatory friction, and socio-political resistance—pose the greatest risk of slowing or derailing Figure AI’s thesis of rapid industrial and commercial scaling. Current standards (ISO 10218, R15.06) were written for fixed manipulators; they do not yet fully address bipedal mobility, high-force bimanual tasks, or fall dynamics in shared human spaces, leaving integrators exposed under OSHA’s General Duty Clause and EU high-risk rules.[12]

• Liability: Autonomous adaptation and learning behaviors challenge traditional product-liability assumptions; courts and insurers will likely demand robust documentation and third-party certification that many early deployments lack.
• Ethical/job-displacement pressure: Union demands and public fear of widespread displacement can translate into political or contractual hurdles, as already seen at Hyundai.
• Adoption friction: High unit costs, reliability shortfalls in unstructured environments, and the need for site-specific infrastructure (floor coatings, fall zones, hot-swap protocols) raise total cost of ownership and slow pilots from proof-of-concept to volume orders.

For any company or investor betting on humanoid platforms, the winning strategy is to treat safety certification and stakeholder engagement as core product features rather than after-the-fact compliance tasks. Early adoption of emerging humanoid-specific drafts (ISO 25785-1), transparent incident reporting, and proactive union dialogue can convert regulatory and social headwinds into competitive moats. Without these, even technically impressive systems risk prolonged pilot purgatory or outright rejection in key markets.

Recent Findings Supplement (May 2026)

US regulatory environment remains anchored in the General Duty Clause with reliance on updated consensus standards. OSHA continues to lack any humanoid- or robotics-specific standard as of May 2026, instead enforcing the General Duty Clause (Section 5(a)(1)) and cross-referencing the newly revised ANSI/A3 R15.06-2025 (Parts 1–3) and ISO 10218-2025 for risk assessments, machine guarding, lockout/tagout (1910.147), and emerging cybersecurity requirements. This creates a de-facto compliance pathway through “recognized hazards” citations when companies deviate from these voluntary standards.[1][2]

• The ANSI/A3 R15.06-2025 revision (published 2025, heavily referenced in 2026 guidance) harmonizes U.S. rules with the latest ISO 10218 updates, adding explicit functional safety, end-effector rules, and cybersecurity provisions—the first major overhaul since 2012.[1]
• Inspectors increasingly expect documented ISO-style risk assessments and manufacturer-specified testing intervals; non-compliance has already triggered citations in warehouse and manufacturing settings.[3]

For companies entering the space, this means proactive adoption of the 2025 standards is now table stakes to avoid OSHA citations and to build defensible safety cases for insurers and customers.

EU AI Act developments introduce phased high-risk obligations that directly impact humanoid deployments. A November 2025 draft “Digital Omnibus” proposal and May 2026 political agreement signal continued refinement, with high-risk workplace AI systems (including locomotion, perception, and task-execution stacks in humanoids) facing obligations from August 2026 onward and full applicability potentially delayed to 2027 under the omnibus mechanism.[4][5]

• Humanoid safety functions (balance control, collision avoidance, language-driven task execution) are explicitly flagged as high-risk in workplace pilots, triggering risk-management, logging, human oversight, and post-market monitoring requirements.[6][7]
• The updated Machinery Regulation and EU Product Liability Directive (transposition by December 2026) treat software/AI updates as products, extending strict liability across the supply chain.[8]

This creates a clear timeline pressure: operators must demonstrate systematic safety cases by Q3 2027 or face market-access barriers in Europe.

Safety certification frameworks are converging around updated ISO/ANSI standards but remain incomplete for bipedal humanoids. No dedicated humanoid safety certification exists; companies must combine ISO 10218, ISO/TS 15066 (collaborative robots), ANSI/A3 R15.06-2025, and emerging ISO 25785-1 elements while navigating CE marking under the revised Machinery Regulation.[9]

• Battery thermal, electromagnetic, and fall-protection protocols are now explicitly called out in 2026 compliance guides as areas where deviation from manufacturer specs can trigger OSHA or EU enforcement.[2]
• Home or unstructured-environment deployments lack tailored rules (ISO 13482 is too general), leaving a regulatory gray zone.[10]

New entrants must budget for multi-standard certification and third-party validation early; reliance on “soft coverings” or internal testing alone is insufficient for commercial scaling.

Labor union resistance is materializing in specific markets, most visibly in Korea. In March 2026, Korean unions publicly blocked Hyundai/Boston Dynamics Atlas deployments, declaring “not a single unit without labor-management agreement” over job-loss fears, forcing initial U.S.-based trials instead.[11]

• European unions (especially in Germany and France) are expected to slow factory rollouts through similar co-determination requirements, while U.S. deployments (BMW Spartanburg, Figure pilots) have faced minimal organized pushback to date.[12]
• Public sentiment remains mixed—viral May 2026 Figure livestreams of 24–48-hour autonomous shifts generated millions of views and positive buzz, yet experts immediately noted the demos were limited to repetitive tasks.[13]

Union opposition creates geographic adoption asymmetry: U.S. and Chinese markets may accelerate while Europe and Korea lag, forcing Figure and peers to prioritize non-union or greenfield sites.

Documented real-world humanoid incidents remain scarce in public records. Searches through May 2026 yield no verified industrial accidents, injuries, or regulatory citations involving Figure AI or comparable commercial humanoids. Recent trials (BMW 11-month Figure 02 deployment accumulating 1,250 hours; JAL May 2026 airport pilots) report only routine pauses or minor task errors in controlled demos, with automatic recovery mechanisms credited for continuity.[14]

• General robotics literature continues to highlight simulation-to-real gaps and recovery failures, but these have not translated into publicized commercial humanoid incidents.[15]
• A December 2025 report of a product-safety executive lawsuit against Figure AI surfaced but lacks public details on outcomes or safety findings.[16]

The absence of high-profile failures is currently a tailwind, but it also means the first serious incident could trigger rapid regulatory tightening.

Key barriers to Figure AI’s growth thesis center on liability uncertainty, certification timelines, and regional labor friction rather than outright bans. Product-liability evolution (EU PLD effective December 2026, U.S. state AI liability proposals) treats AI-driven physical systems as products, exposing manufacturers and deployers to strict liability for software updates or autonomous decisions.[17]

• Regulatory lag—still no dedicated OSHA humanoid rule—creates compliance ambiguity that insurers and large customers will price into contracts.
• Union gatekeeping in key manufacturing regions and the 2026–2027 EU high-risk compliance cliff could delay or fragment scale deployments.
• The positive safety record to date buys time, but any incident will amplify calls for dedicated standards and incident-reporting regimes.

Competitors that front-load ISO/ANSI certification, publish mean-time-between-failure data, and secure union agreements in pilot geographies will gain durable first-mover advantages in the 2026–2028 window.

Figure AI faces substantial headwinds rooted in robotics' long track record of overpromising and underdelivering at scale. Historical failures like Rethink Robotics illustrate how even innovative cobot pioneers can collapse when sales expectations miss and commercialization lags.[1][2]

Rethink closed in 2018 after an acquisition fell through and cash ran out; its Baxter and Sawyer robots, while pioneering safe collaboration, proved too slow and imprecise for demanding industrial use, with sales far below forecasts. Sarcos, after public listing via SPAC, pivoted entirely away from hardware in 2023–2024—suspending exoskeleton and industrial robot commercialization, laying off ~150 employees, and shifting to AI software—because its complex, expensive hardware could not achieve profitable scale or reliable field performance.[3][4]

These cases show the mechanism: high fixed costs for custom actuators, sensors, and integration combine with narrow real-world tolerance to create unsustainable burn rates before recurring revenue materializes. The implication is that Figure’s $39B valuation and aggressive timelines risk repeating the pattern unless it closes the gap between pilot hype and profitable, repeatable deployments faster than predecessors.

Demos continue to outpace reliable, transferable real-world performance. Figure’s recent livestreams of package-sorting shifts (8–67 hours claimed with minimal intervention) look impressive in controlled loops, yet experts and observers note visible pauses, misplaced packages, slower-than-human speeds, and setups where the same items cycle repeatedly rather than handling dynamic warehouse variability.[5][6]

Broader industry analysis highlights the core issue: policies trained on narrow distributions or simulation fail when encountering out-of-distribution objects, lighting changes, or slight environmental perturbations; one 95% reliable step compounds across 100 sequential actions into <1% end-to-end success.[7][8]

Even Figure’s CEO has described an “open-ended state space” and “never-ending problem city.” This gap matters because customers (BMW pilots, logistics trials) quickly discover that lab or short-shift success does not translate to 24/7 uptime with minimal human oversight—precisely the economics required to justify high capital outlays.

Hardware cost and manufacturability remain structural bottlenecks. Figure has publicly detailed redesigning its entire robot from prototype (high part count, CNC-machined components) to more scalable processes, acknowledging that legacy supply chains for actuators, sensors, and batteries are immature.[9]

Current analyst estimates place Figure units at $30,000–$150,000, while average humanoid selling prices in 2024 hovered near $200,000.[10][11] Scaling to thousands of units requires capital-intensive retooling and yield improvements that have historically drained robotics startups before volume economics kick in. The implication is that any delay in hitting sub-$25,000 viable pricing (or proving 90%+ uptime at that cost) erodes the unit economics that justify Figure’s valuation.

Chinese competitors are already shipping at dramatically lower prices and higher volumes. Unitree’s G1/R1 platforms have dropped from ~$85,000 to $5,900–$16,000 in roughly two years through vertical integration (90%+ in-house components) and aggressive iteration.[12][13]

In 2025, Chinese firms (led by Unitree and AgiBot) accounted for ~80% of global humanoid shipments, with Unitree alone reportedly delivering thousands versus Figure’s ~150 units.[14][15]

The mechanism is straightforward: lower labor and component costs, state-supported manufacturing ecosystems, and a willingness to accept lower per-unit margins enable rapid volume that generates real-world data loops unavailable to Western players still focused on high-spec prototypes. This creates a two-tier market where price-sensitive early adopters (factories, logistics) may lock in with Chinese hardware before Figure’s superior AI/software can differentiate at scale.

Specific negative coverage and internal signals have emerged around Figure itself. A November 2025 whistleblower lawsuit by former product-safety head Robert Gruendel alleges the company rushed development, lacked formal safety procedures, and produced robots capable of “superhuman speed” and forces “twenty times higher than the threshold of pain”—including a malfunction that carved a quarter-inch gash in stainless steel—while dismissing safety concerns.[16][17]

Figure denies the claims and has reportedly countersued. Separately, a Fortune article questioned the scale of the BMW partnership (claiming fewer robots and smaller scope than portrayed), prompting Figure to threaten defamation litigation.[18]

WSJ reporting and analyst commentary have also tempered expectations, noting that even humanoid makers themselves view current systems as overhyped for complex industrial or domestic work.[19]

These incidents amplify investor and customer skepticism at a moment when credibility on safety and delivery timelines is critical.

Ranked top risks to Figure AI’s long-term viability (most to least severe, based on recurrence in failed predecessors and current evidence):

1. Failure to cross the “valley of death” from pilot to profitable scaled deployment — sustained high burn without recurring revenue or proven multi-site uptime will mirror Rethink/Sarcos outcomes.
2. Price/volume competition from Chinese players eroding addressable market — early adopters may standardize on sub-$10k hardware before Figure reaches competitive economics.
3. Safety/regulatory or reputational damage from whistleblower allegations or real incidents — high-power humanoids operating near people invite liability and customer hesitation.
4. Persistent sim-to-real and generalization gaps limiting reliability — without dramatic improvements in transfer learning, deployments remain labor-intensive and uneconomic.
5. Manufacturing yield and supply-chain scaling delays — custom actuators and immature component ecosystems could cap production velocity even if AI software succeeds.

These risks are interconnected; progress on any one (e.g., cheaper manufacturing) does not automatically solve the others. Competitors or new entrants must therefore prioritize narrow, high-margin niches with measurable ROI over general-purpose claims, while building robust safety validation and real-world data flywheels from day one.

Recent Findings Supplement (May 2026)

Figure AI faces mounting scrutiny over its BMW partnership and demo credibility, with recent coverage highlighting discrepancies between claims and verified operations. In June 2025, CEO Brett Adcock sidestepped direct questions about whether the BMW relationship had moved beyond a pilot into meaningful commercial revenue during a Bloomberg Tech conference appearance, offering only general comments on operational data collection while providing no contract specifics.[1][1] This followed earlier questions about overstated "fleet" deployments and "end-to-end operations," prompting Adcock to publicly threaten legal action against at least one critical report. The company was simultaneously pursuing a $1.5 billion raise at a $39.5 billion valuation despite limited disclosed revenue.[1]

• As of mid-2025 reporting referenced in 2025 coverage, BMW confirmed only limited single-robot tasks (parts handling in body shop) rather than broad fleet deployment.[2]
• Figure continued releasing videos while avoiding live demos at events where competitors like Agility and Boston Dynamics participated.[1]
• In May 2026, Adcock publicly defended a high-profile test against teleoperation skepticism, underscoring ongoing doubts about autonomous performance.[3]

This pattern suggests investors and partners must demand third-party-verified deployment metrics and revenue attribution rather than relying on founder statements or videos.

Rethink Robotics’ second shutdown in September 2025 exemplifies the recurring “valley of death” for humanoid-adjacent robotics startups, where early promise fails to scale commercially. The company, known for collaborative arms like Baxter and Sawyer, filed bankruptcy again after its 2018 collapse and subsequent acquisition; former CEO Julia Astrid Riemenschneider cited unready products, missed sales targets, and investor pullback.[4] Co-founder Rodney Brooks amplified this in a December 2025 New York Times interview, arguing the current Silicon Valley humanoid craze is “doomed to fail” due to flawed assumptions about rapid generalization from demos.[5]

• Brooks reiterated in 2026 commentary that robotics hardware lacks software’s exponential cost/performance gains, making hype-driven valuations unsustainable.[6]
• Broader 2025–2026 analyses note robotics startups frequently prove lab concepts but collapse on unit economics and reliability at volume.[7]

New entrants must prioritize proven, narrow-use-case deployments with clear payback periods over general-purpose ambitions to avoid Rethink’s fate.

Chinese competitors like Unitree are rapidly closing any perceived technology gap while undercutting on price and achieving manufacturing scale that Western firms have historically struggled to match. In 2025, China captured over 85% of roughly 15,000 global humanoid installations (versus 13% for the U.S.), with Unitree and peers driving the majority.[8] Unitree filed for a Shanghai STAR Market IPO in March 2026 targeting ~$7 billion valuation, reporting 60% gross margins and >5,500 humanoid shipments in 2025 (surpassing all U.S. competitors combined) while targeting 20,000 units in 2026.[9] Base pricing sits at $13,500–$15,400 per unit.[8]

• February 2026 Spring Festival Gala showcased Unitree robots performing fluid kung fu, flips, and fault recovery alongside humans—far more dynamic than prior years’ demos.[10]
• Multiple Chinese firms signaled U.S. market entry after strong CES 2026 showings.[11]

Western startups like Figure must demonstrate not just AI superiority but materially better economics or capabilities to justify premium pricing against this volume-driven threat.

Even Figure’s own April 2026 production update reveals persistent hardware and reliability hurdles that align with historical robotics pitfalls. The company described transitioning from prototype to scalable fleet as a “significant hurdle,” requiring supplier qualification for hundreds of parts, yield and cycle-time improvements, and addressing a “long tail” of edge-case failures only visible after accumulating substantial fleet operating hours.[12]

• Earlier demos drew teleoperation doubts, which the CEO addressed publicly in May 2026.[3]
• Expert commentary continues to flag inference, dexterity, and long-term reliability as open problems limiting use cases.[6]

This implies that pilot successes do not automatically translate to profitable, high-volume deployments without years of iterative, expensive real-world hardening.

Ranked top risks to Figure AI’s long-term viability (based on recency and specificity of evidence):

1. Sustained skepticism around commercial traction and hype vs. reality (e.g., BMW discrepancies and high-valuation raises with minimal revenue disclosure) eroding investor and customer confidence.
2. Aggressive Chinese price and volume competition (Unitree’s 2025–2026 shipments, margins, and IPO path) commoditizing the market before Figure reaches scale.
3. Recurring “valley of death” patterns demonstrated by Rethink’s 2025 shutdown and Brooks’ critiques of humanoid economics.
4. Hardware reliability and manufacturability gaps (long-tail failures, yield issues) delaying profitable deployment despite pilot progress.
5. Demo-to-real-world performance doubts (teleoperation concerns, limited live verification) undermining claims of general-purpose autonomy.