MENU

Figure AI | Company Profile

← Back to Drones & Robotics guides

COMPANY PROFILE / DRONES AND ROBOTICS

Figure AI
A private humanoid-robot company that says it will own the robot, the AI, the factory and the training data

Figure AI is a private U.S. company that builds a human-shaped robot (a humanoid) and the AI model that drives it, called Helix. It introduced its third-generation robot, Figure 03, in October 2025, and said in April 2026 that its own factory, BotQ, had shipped more than 350 third-generation robots. Trials at a BMW plant, in a logistics centre and in homes are described, but all of these are the company’s own announcements, and how far each one says the work was autonomous or remotely operated, how many robots took part, for how long, and whether anyone checked it independently differs from post to post. This profile sticks to the company’s website and its SEC Form D, and does not touch valuations or investment views. Check date: 25 October 2026.

Built from announcements on Figure AI’s website (June 2025 to September 2026), its SEC Form D (13 March 2024) and a federal court docket / Checked 25 October 2026

What this article covers
  1. Figure AI in 30 seconds, and where it stands in drones and robotics
  2. The logo and how the name is written
  3. Where the company sits in drones and robotics
  4. Scale
  5. Resources committed to drones and robotics
  6. What it has done so far
  7. What comes next
  8. What the company is aiming for in drones and robotics
  9. Map of partnerships in drones and robotics
  10. Technology, products and services
  11. The risks this company carries
  12. Glossary
  13. References
  14. Claim-to-source audit

1. Figure AI in 30 seconds, and where it stands in drones and robotics

LEGAL NAMEFigure AI Inc.A Delaware corporation, as named in its SEC Form D
LOCATIONCalifornia, U.S.The March 2024 Form D lists Sunnyvale; a 2026 company post refers to its San Jose campus
LISTINGPrivateThe company says any share transfer needs board approval (8 July 2026)
TYPE IN THIS FIELDHumanoids & physical AIAims to own the robot, AI, manufacturing and training data
CORE PRODUCTSFigure 03 and HelixA third-generation humanoid (introduced 9 October 2025) and the AI model that runs it
MANUFACTURINGIts own factory, BotQ29 April 2026: more than 350 third-generation robots shipped; output up from 1 a day to 1 an hour (company claim)
CUSTOMERS (COMPANY ANNOUNCEMENTS)BMW / Catalyst BrandsBMW’s Spartanburg plant in the U.S. and Catalyst’s distribution centre in Reno, Nevada
FUNDING (OFFICIAL)Series C above US$1 billionAnnounced 16 September 2025. Valuation is not covered here
The one point to grasp first

Figure AI says it will own four things: the robot body, its brain (AI), the factory and the training data Sourced. The body is Figure 03, the brain is Helix, the factory is BotQ, and the data comes from Index, an app that collects videos of people at work. At the same time, almost every demonstration or deployment result comes from the company’s own web posts. For example, “contributed to the production of more than 30,000 cars” at BMW is the company’s own tally, and we found no independent check of it. When you read such claims, look at what the company says and what conditions it gives.

2. The logo and how the name is written

A note on names: the legal name is Figure AI Inc. (SEC filing), but the company’s own site often just says “Figure”. Robot generations are written both as “Figure 02” and “F.02”, which mean the same thing. Do not confuse it with other firms with similar names. The logo is the file used in the header of the official site.

Official site: https://www.figure.ai/

3. Where the company sits in drones and robotics

Companies in drones and robotics are easier to understand if you sort them into four types by what they build and who buys it. Figure AI is mainly a humanoid and physical-AI company: it builds human-shaped robots that learn with AI. Physical AI means AI that moves real objects, not just text or images on a screen. Its secondary type is components and enabling software, because it says it designs its own actuators (the parts that move the joints), batteries and sensors Sourced.

The four types of company in drones and robotics Even among “robot companies”, strengths and challenges differ by what is built and who buys it Humanoids & physical AI Industrial & specialist robots Drones & eVTOL Components & enabling software Humanoid, learning robots Factory and medical work Aircraft and operations Parts and AI behind robots Figure AI / Boston Dynamics Unitree / UBTech FANUC / Yaskawa Universal Robots/Intuitive DJI / Skydio Zipline / Joby Harmonic Drive NVIDIA Learning, generality Track record, reliability Airspace uses, regulation Presence in supply chain Scale-up still unproven Adapting to change, price Rules and certification Demand follows shipments Mainly humanoids & physical AI: robot, AI, manufacturing and data are all meant to be in-house Most deployments are company announcements; no independent verification was found
Fig. 1 The four types of company in drones and robotics, and where Figure AI sits. The grouping is ours, and company names are examples from those covered in this series Our calculation. Figure AI’s approach is based on its own announcements, such as the Figure 03 launch post.

The “weakness” row in the figure is a general remark about the type as a whole, not a judgement of Figure AI.

4. Scale

Figure AI is private and does not publish revenue, headcount or profit. This table lists what we could confirm (check date: 25 October 2026).

ItemWhat we could confirmType
FoundingThe SEC Form D gives 2023 as the year of incorporation. The company site says its first robot, F.01, took its first steps in May 2023Sourced
Revenue and profitNot disclosed in the SEC filing or official posts. We do not estimate themSourced
EmployeesNot disclosed. The site describes a team with “over 100 years of combined AI and humanoid experience”, which is not a headcountSourced
Funding (Series B)Form D: offering of US$675,000,000, US$674,981,967 sold, 45 investors. First sale 27 February 2024Sourced
Funding (Series C)Announced 16 September 2025: more than US$1 billion in committed capital, led by Parkway Venture Capital, with Brookfield, NVIDIA, Intel Capital and othersSourced
ValuationIt appears in the company post, but this profile does not cover valuations or investment views, so it is not reproduced–

Amounts are in the original currency (US$). The Series B figure is the total offering stated in the Form D.

A caution

A Form D is a notice to the SEC of a private fundraising. It is not a statement of the company’s finances. All we could confirm about money is the offering size and the number of investors. Revenue and profit are not disclosed, so we cannot give them.

5. Resources committed to drones and robotics

This table shows whether we could confirm each of four resources (people, money, outside investments or acquisitions, and ties with universities or public bodies) in official posts and the SEC filing. “Not stated” is also a result.

ResourceWhat we could confirmType
PeopleHeadcount is not disclosed. The 29 April 2026 post says it built its own service, support and fleet-management systems, with no numbersSourced
Money: manufacturingIts own factory, BotQ. The first-generation line can build up to 12,000 robots a year, with a goal of 100,000 in total over four years (October 2025). In April 2026 it said it had qualified hundreds of suppliers, put more than 50 inspection points in the line, and run 80+ functional tests on every robot (company description)Sourced
Money: computeNscale partnership (3 September 2026): up to 100,000 GPUs, an initial compute commitment of US$3.5 billion with intent to scale beyond US$6 billion, first deployment from the second half of 2027 in TexasSourced Not yet confirmed
Money: dataThrough Index, US$15 million has been paid to data contributors (“Creators”) so far. It said it is committed to spending more than US$1 billion on data and compute over the next 12 months (25 August 2026)Sourced
Outside investments, acquisitions, divestmentsAs a recipient: Series C investors (Brookfield, NVIDIA, Intel Capital and others) and a strategic investment from Nscale. No acquisition or sale of a business appears in the official news list (not confirmed)Sourced
Universities, research bodies, public bodiesThe official posts we read mention no joint research with universities, research institutes, NASA or DARPANot confirmed

The table below lists the main robots, AI models and facilities that the company’s own materials let us confirm (check date: 25 October 2026). All specifications and results are company claims or company tallies.

NameKindMain specifications and results (company claims, with conditions)StageDateReference
Figure 02 (F.02)Second-generation humanoidEleven months at BMW’s Spartanburg plant, 10-hour shifts Monday to Friday, 90,000+ parts loaded, 1,250+ hours of run time, “contributed to” 30,000+ BMW X3s (company tally). The task was loading sheet-metal parts into a welding fixture. Targets: 84-second cycle, placement within 5 mm, zero human interventionsRetired (started November 2025; most of the fleet disposed of by the 30 September 2026 post)19 November 2025Reference 2 https://www.figure.ai/news/production-at-bmw Reference 13 https://www.figure.ai/news/f-02-decommission
Figure 03 (F.03)Third-generation humanoidVersus the previous generation: cameras with twice the frame rate, one-quarter the latency and a 60% wider field of view; 9% less mass; palm cameras; fingertip touch sensors (said to detect 3 g of force); 2 kW wireless charging through coils in the feet; battery said to be certified to UN38.3Shipping and in use (350+ shipped by April 2026)9 October 2025 (launch) / 29 April 2026Reference 1 https://www.figure.ai/news/introducing-figure-03 Reference 4 https://www.figure.ai/news/ramping-figure-03-production
Figure 03 at BMWDemonstration at a customerParts “sequencing”: picking thin parts with both hands and pulling a cart while repositioning its feet, run by Helix 02 (company description). No number of robots, duration or statement on remote operationDemonstration (company announcement)30 June 2026Reference 3 https://www.figure.ai/news/f-03-at-bmw
Helix 02AI model (VLA)One neural network controls the whole body. A four-minute dishwasher task was shown as autonomous, not teleoperated (company claim). S0 is a 10-million-parameter network trained on 1,000+ hours of human motion data and in simulationPublished technical report27 January 2026Reference 5 https://www.figure.ai/news/helix-02
Helix 2.5AI model (VLA)Tested “zero-shot” in 30 unseen Bay Area homes on three tasks (tidying a room, folding towels, making a bed). Success rate 56% with Index pretraining versus 9% without (company’s blind evaluation; no independent verification found)Research result17 September 2026Reference 6 https://www.figure.ai/news/helix-2-5-zero-shot-30-home-generalization
IndexData-collection app264,000+ downloads, 44,000+ weekly users, 16 million+ videos (company figures as of 25 August 2026)In service25 August 2026Reference 7 https://www.figure.ai/news/introducing-index

“Zero-shot” means trying a task in a place where the robot was never trained (see the glossary). Apart from the battery’s UN38.3, we found no regulatory approval or certification.

6. What it has done so far

WhenWhat happenedWhy it matters
May 2023The first-generation F.01 takes its first steps (company site)Starting point of its humanoid work
27 February 2024First sale in the Series B (Form D); US$675 million offeredA public record of fundraising
7 June 2025Helix improvements in parcel sorting: handling time per package cut from about 5.0 s (the post gives both 4.05 s and 4.31 s), barcode-facing success up from about 70% to about 95% (company figures)Shows the learning approach in logistics; conditions are the company’s own
16 September 2025Announces Series C: more than US$1 billion committedSaid to fund manufacturing, compute and data collection
17-18 September 2025Brookfield partnership and Project Go-Big (pretraining on human video)Widens training data from robot operation to human video
9 October 2025Introduces Figure 03; BotQ first-generation line up to 12,000 a yearA third generation redesigned for mass production
19 November 2025Reports eleven months at BMW and starts retiring Figure 02Operating results as tallied by the company
21 November 2025A former employee files suit in federal court (court record). The company filed an answer and counterclaims on 16 January 2026Pending. This profile does not judge the claims
27 January 2026Introduces Helix 02Walking and manipulation under one AI
29 April 2026Says BotQ shipped 350+ third-generation robots and now builds one an hour instead of one a dayThe company’s account of ramping up
26 May 2026Commercial agreement with Catalyst Brands, starting in RenoA deployment agreement (numbers and term not disclosed)
30 June 2026Figure 03 arrives at BMW Spartanburg and demonstrates parts sequencingDemonstration (company announcement)
25 August 2026Launches the Index data-collection appMoves data sourcing in-house
3 September 2026Partnership with Nscale for up to 100,000 GPUsSecuring compute
17 September 2026Introduces Helix 2.5 (tested in 30 homes)Presented as a result of learning from human video
30 September 2026Retirement of Figure 02; mentions F.04Change of generation

7. What comes next

Everything below is a company goal or plan, not something already achieved Not yet confirmed.

BotQ CAPACITYUp to 12,000 a yearFirst-generation line. Goal of 100,000 over four years (October 2025)
COMPUTEUp to 100,000 GPUsNscale partnership; first deployment from the second half of 2027 (3 September 2026)
SPENDING ON DATA AND COMPUTEUS$1 billion+ in 12 monthsCommitment stated in the Index announcement
NEXT ROBOTF.04Mentioned only in the retirement post; no date or specifications
What we could not confirm

None of the following appears in the official posts we read Not yet confirmed.

The sale price of Figure 03; when it will be offered for homes; the number of robots and the period at BMW and Catalyst; any independent performance check after deployment; a third-party safety certification of the whole robot; and the specifications and timing of F.04. Home use is only described as “designed for the home”; we found no start of sales or supply.

8. What the company is aiming for in drones and robotics

This chapter includes readings the company does not state directly. They are inferences from official posts, shown with their basis.

What we readPrimary basisType
It is aiming at both homes and commercial workFigure 03 is said to be designed for Helix (the AI), for the home, and for the world at scale (Reference 1)Sourced
It designs the robot, battery, sensors and electronics in-house and builds in its own factoryThe Figure 03 post says actuators, batteries, sensors, structures and electronics are designed in-house; parts come from suppliers able to meet volumeSourced
It is shifting the centre of learning from data gathered by operating robots to videos of peopleProject Go-Big (September 2025), Index (August 2026), and Helix 2.5, where Index pretraining lifts success from 9% to 56%Inference
It sees its current limit as compute and dataThe Nscale post says the company is “largely bound by data and compute”Sourced (company’s own view)
The number of robots actually working at customers may be small compared with the number builtThe 350+ shipped robots are allocated to internal R&D, data collection, housework and commercial development. Numbers per customer are not disclosedInference
One inference in more detail

The April 2026 post says robots shipped from BotQ are “allocated to internal research and development groups, data collection, efforts for robots to perform end-to-end housework, and commercial use-case development”, and that a larger fleet generates more data for Helix Sourced. So a good share of the shipped robots seems to serve as tools for collecting data and improving the system, not only as workers at customer sites. Together with the lack of per-customer numbers, we read the shipment count as not equal to the number of robots deployed at customers Not yet confirmed.

9. Map of partnerships in drones and robotics

Relationships around Figure AI (only those in official posts) Only relationships stated in official posts and the Form D. Press-only links are excluded (checked 25 October 2026) Figure AI Humanoid robots BMW Group Deployed at Spartanburg (2025 on) Catalyst Brands Deployment deal, Reno centre Brookfield Data and deployment; investor Nscale Supplies up to 100,000 GPUs NVIDIA GPUs in robots; investor Series C investors Led by Parkway, and others Deployment / joint work Compute and technology supply Investment
Fig. 2 Source: Figure AI’s official posts (September 2025 to September 2026). The BMW and Catalyst deployments are company announcements; numbers and terms are not disclosed. NVIDIA is a Series C investor, and the Nscale post describes training on NVIDIA GPUs, checking in NVIDIA Isaac Sim and running on NVIDIA GPUs in the robots. Investment links are dotted.

10. Technology, products and services

Figure AI’s products are easiest to follow as four layers: the body (Figure 03), the brain (Helix), the factory (BotQ) and the data (Index).

LayerWhat it isWhat the company says
BodyFigure 03Cameras with twice the frame rate, one-quarter the latency and a 60% wider field of view than before; a camera in each palm; fingertip touch sensors that detect 3 g of force; 10 Gbps millimetre-wave data offload; 2 kW wireless charging through the feet; 9% less mass than Figure 02
Brain (AI)Helix 02 / Helix 2.5Three layers, S2, S1 and S0 (see the illustration below). S0 is said to replace 109,504 lines of hand-written C++ with one neural network
FactoryBotQBuilds key parts in-house rather than through contract manufacturers. Its own manufacturing execution system links 150+ workstations. Battery line first-pass yield 99.3%; 9,000+ actuators built in 10+ types (April 2026)
DataIndexAn app where people film their own tasks at home or work. Five-stage screening: filtering, fraud review, de-duplication, rebalancing and annotation. Per 1,000 hours: 373 tasks, 1,146 objects, 116 environments
OperationsFleet management and over-the-air updatesA system that tracks robot health and location, wireless software updates, and processes for fleet-wide upgrades and recall campaigns
The three layers of Helix (company description) The upper the layer, the slower it thinks; the lower, the faster it moves the body System 2 (S2) Understands, sets goals Reads the scene and spoken instructions, and turns a large order such as “walk to the dishwasher and open it” into signals for the layer below System 1 (S1) 200 times a second Takes in head and palm cameras, fingertip touch and body state, and produces target positions for every joint System 0 (S0) 1,000 times a second Keeps balance and contact in check and sends force commands to joints. Trained on 1,000+ hours of human motion and in simulation (10M parameters)
Illustration The three layers of Helix 02. Source: Figure AI’s technical post of 27 January 2026. The figures are the company’s own; no independent verification was found.
A materials engineer’s view: where softness, heat and batteries matter

Hands and skin: the Figure 03 fingertips use softer material to enlarge the contact area and give steadier grips. The company says commercial tactile sensors could not stand up to real use, so it built its own, guided by durability, long-term reliability and fidelity Sourced. Such sensors often pair a soft layer with an element that turns deformation into an electrical signal, but Figure has not published its structure or materials.

Heat in the forearm: the biggest hardware failure point at BMW was the Figure 02 forearm. It packs three joints and wiring into a small space and has tight limits on heat. For Figure 03 the company removed the forearm relay board and moving cables, so each wrist’s motor controller talks directly to the main computer Sourced.

Soft cover and battery: Figure 03 is covered in multi-density foam to protect against pinch points and in washable textiles that can be swapped without tools; cut-resistant fabrics are an option. The battery has four layers of protection (battery management system, cell, interconnect and pack) and is said to be certified to the UN transport test standard UN38.3 Sourced (UN38.3 is a shipping safety test, not a safety certification of the robot).

Parts made for volume: Figure 02 was mostly machined, while Figure 03 relies far more on die-casting, injection moulding and stamping. These need money for tooling up front, but the cost per unit falls as volumes grow.

11. The risks this company carries

RiskWhat it isType
Results are hard to check from outsideThe BMW work, the 350+ shipments and Helix performance are all company announcements or tallies. We found no independent verification or customer confirmationSourced
Hardware reliabilityThe company says the forearm was the top failure point on Figure 02. It redesigned it for Figure 03, but has not published long-term failure rates for Figure 03Sourced
Safety around peopleFigure 03 has home-oriented safety features (foam, textiles, battery protection), but we found no third-party safety certification of the whole robot. Safety standards for humanoids are still being drafted (see our explainer “How to read robot safety standards”)Inference
Dependence on compute and dataThe company itself says it is currently “largely bound by data and compute”. The compute plan starts from the second half of 2027Sourced Not yet confirmed
Handling of collected dataIndex gathers videos filmed by ordinary people at home and work. The posts we read do not describe consent or privacy for people and places that appear in themSourced (not stated)
LitigationA former employee filed suit in federal court on 21 November 2025; the company filed an answer and counterclaims on 16 January 2026 (court record). It is pending, and this profile does not judge either sideSourced
Uncertain plans12,000 a year, 100,000 in four years and home supply are all company goalsNot yet confirmed

12. Glossary

Humanoid
A robot shaped like a person (head, torso, two arms, two legs), so it can use spaces and tools made for people.
Physical AI
AI that works on real objects, not only on text or images on a screen.
VLA (vision-language-action model)
An AI that turns camera images and spoken instructions directly into robot movements. Helix is one.
Zero-shot
Trying a task in a place the system was never trained in.
Actuator
The part that moves a joint (typically a motor plus a gearbox).
Tactile sensor
A sensor on a fingertip or similar that feels the size of a touch force or slipping.
UN38.3
The United Nations test standard for shipping lithium batteries. It is not a safety certification of a robot.
Loco-manipulation
Walking and handling objects at the same time.
Sequencing
In a factory, picking parts and laying them out in the order they will be assembled.
Teleoperation
A person driving the robot remotely. Telling it apart from autonomy is key when reading demonstrations.
Form D
A notice to the U.S. SEC of a private fundraising. It does not contain the company’s financial statements.
Vertical integration
Owning the steps needed for a product, such as parts, assembly and software.
A humanoid illustrates handling parts on a shelf. The scene connects object perception and hand control with workflows using existing shelves and carts.
Fig. 3. Concept illustration generated with ImageGen. A humanoid illustrates handling parts on a shelf. The scene connects object perception and hand control with workflows using existing shelves and carts. Not an actual product, facility or measurement. Cutaways, colours and arrows explain the concept; proportions are illustrative.

13. References

  1. Figure AI Introducing Figure 03 (9 October 2025). Design, cameras, touch sensing, charging and manufacturing (BotQ) of Figure 03. https://www.figure.ai/news/introducing-figure-03
  2. Figure AI F.02 Contributed to the Production of 30,000 Cars at BMW (19 November 2025). Eleven months of work at BMW and the forearm failures. https://www.figure.ai/news/production-at-bmw
  3. Figure AI F.03 Arrives at BMW (30 June 2026). Figure 03 demonstrates parts sequencing at BMW. https://www.figure.ai/news/f-03-at-bmw
  4. Figure AI Ramping Figure 03 Production (29 April 2026). BotQ shipments, production rate and yields. https://www.figure.ai/news/ramping-figure-03-production
  5. Figure AI Introducing Helix 02: Full-Body Autonomy (27 January 2026). The S0, S1 and S2 layers. https://www.figure.ai/news/helix-02
  6. Figure AI Helix 2.5: Zero-Shot 30-Home Generalization (17 September 2026). Evaluation in 30 homes. https://www.figure.ai/news/helix-2-5-zero-shot-30-home-generalization
  7. Figure AI Introducing Index: Building The World’s Largest and Most Diverse Physical Dataset (25 August 2026). The app that collects human video. https://www.figure.ai/news/introducing-index
  8. Figure AI Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation (16 September 2025). The funding announcement (the valuation is not used here). https://www.figure.ai/news/series-c
  9. Figure AI Figure Announces Strategic Partnership with Brookfield (17 September 2025). https://www.figure.ai/news/figure-announces-strategic-partnership-with-brookfield
  10. Figure AI Figure Signs Agreement with Catalyst Brands to Scale Humanoid Operations (26 May 2026). https://www.figure.ai/news/figure-signs-agreement-with-catalyst-brands
  11. Figure AI Figure and Nscale Sign Strategic Partnership For Up to 100,000 GPUs on the NVIDIA Vera Rubin Platform (3 September 2026). Compute agreement. https://www.figure.ai/news/figure-and-nscale-sign-strategic-partnership
  12. Figure AI Scaling Helix: a New State of the Art in Humanoid Logistics (7 June 2025). Work in parcel logistics. https://www.figure.ai/news/scaling-helix-logistics
  13. Figure AI F.02 Decommission (30 September 2026). Retirement of Figure 02. https://www.figure.ai/news/f-02-decommission
  14. Figure AI Notice Regarding Unauthorized Attempts to Sell Figure Stock (8 July 2026). Notice on share-transfer limits. https://www.figure.ai/news/notice-regarding-unauthorized-attempts-to-sell-figure-stock
  15. Figure AI Project Go-Big: Internet-Scale Humanoid Pretraining and Direct Human-to-Robot Transfer (18 September 2025). Pretraining on human video. https://www.figure.ai/news/project-go-big
  16. Figure AI Company page on the official site (describes F.01 to F.03). https://www.figure.ai/company
  17. Figure AI Official news list (used to check dates and titles). https://www.figure.ai/news
  18. U.S. SEC Form D of Figure AI Inc. (filed 13 March 2024). Series B offering amount, amount sold and number of investors. https://www.sec.gov/Archives/edgar/data/2014185/000149315224009735/primary_doc.xml
  19. U.S. SEC EDGAR filing index (CIK 0002014185). Used to check the filer name, state of incorporation and address. https://data.sec.gov/submissions/CIK0002014185.json
  20. U.S. District Court (N.D. Cal.) U.S. District Court, Northern District of California: docket of case 5:25-cv-10094 (as mirrored by CourtListener). Used only to check the filing, answer and counterclaim dates. https://www.courtlistener.com/docket/71948915/gruendel-v-figure-ai-inc/

14. Claim-to-source audit

Claim in the textTypeReference
Legal name Figure AI Inc., Delaware corporation, 2023 incorporation, Series B offering amount, amount sold and investorsSourcedReference 18 https://www.sec.gov/Archives/edgar/data/2014185/000149315224009735/primary_doc.xml Reference 19 https://data.sec.gov/submissions/CIK0002014185.json
The first-generation F.01 took its first steps in May 2023SourcedReference 16 https://www.figure.ai/company
Figure 03 launch (9 October 2025) and its specifications (cameras, touch, charging, UN38.3, mass, manufacturing method)SourcedReference 1 https://www.figure.ai/news/introducing-figure-03
BotQ first-generation line up to 12,000 a year, goal of 100,000 in four yearsNot yet confirmedReference 1 https://www.figure.ai/news/introducing-figure-03
BotQ shipped 350+ third-generation robots; 1 a day to 1 an hour; yields and inspection figuresSourcedReference 4 https://www.figure.ai/news/ramping-figure-03-production
Eleven months of Figure 02 at BMW (90,000+ parts, 1,250+ hours, 30,000+ X3s) and forearm failuresSourcedReference 2 https://www.figure.ai/news/production-at-bmw
Retirement of Figure 02 (November 2025; post of 30 September 2026)SourcedReference 2 https://www.figure.ai/news/production-at-bmw Reference 13 https://www.figure.ai/news/f-02-decommission
Figure 03 parts-sequencing demonstration at BMWSourcedReference 3 https://www.figure.ai/news/f-03-at-bmw
Helix 02 layers and figures (200 Hz, 1 kHz, 10M parameters, 1,000+ hours, 109,504 lines)SourcedReference 5 https://www.figure.ai/news/helix-02
Helix 2.5 evaluation in 30 homes (9% and 56%)SourcedReference 6 https://www.figure.ai/news/helix-2-5-zero-shot-30-home-generalization
Index figures (264,000 downloads, 44,000 weekly users, 16M videos, US$15M paid, US$1B+ in 12 months)SourcedReference 7 https://www.figure.ai/news/introducing-index
Series C above US$1 billion led by Parkway; investors include Brookfield and NVIDIASourcedReference 8 https://www.figure.ai/news/series-c
Brookfield partnership (17 September 2025) and Project Go-BigSourcedReference 9 https://www.figure.ai/news/figure-announces-strategic-partnership-with-brookfield Reference 15 https://www.figure.ai/news/project-go-big
Commercial agreement with Catalyst Brands (starting in Reno)SourcedReference 10 https://www.figure.ai/news/figure-signs-agreement-with-catalyst-brands
Nscale partnership (up to 100,000 GPUs, US$3.5B, intent above US$6B, second half of 2027)SourcedReference 11 https://www.figure.ai/news/figure-and-nscale-sign-strategic-partnership
Helix logistics improvements (from about 5.0 s; barcode-facing about 70% to about 95%) and the post’s own 4.05 s / 4.31 s inconsistencySourcedReference 12 https://www.figure.ai/news/scaling-helix-logistics
The company says share transfers need board approvalSourcedReference 14 https://www.figure.ai/news/notice-regarding-unauthorized-attempts-to-sell-figure-stock
Suit by a former employee (21 November 2025); answer and counterclaims (16 January 2026)SourcedReference 20 https://www.courtlistener.com/docket/71948915/gruendel-v-figure-ai-inc/
Treatment of headcount, revenue, profit and valuationNot yet confirmedHeadcount, revenue and profit are not disclosed and not estimated. Valuation is outside this article.
Figure 03 price, home launch date, robot counts and periods at BMW/Catalyst, independent checks, whole-robot third-party safety certificationNot yet confirmedNot in the official posts we read. We do not claim they do not exist Reference 17 https://www.figure.ai/news
Ties with universities, research bodies or public bodies; acquisitions or divestmentsNot yet confirmedNot in the official posts we read Reference 17 https://www.figure.ai/news
Humanoids & physical AI as the primary typeOur calculationThis article’s own classification
The reading that learning is shifting to human videoInferenceThis article’s interpretation (basis: References 6, 7 and 15)
The reading that shipments are not the same as customer deploymentsInferenceThis article’s interpretation (basis: Reference 4)
Safety standards for humanoids are still being drafted and whole-robot third-party certification is unconfirmedInferenceThis article’s interpretation (see our explainer on robot safety standards)

Last updated 25 October 2026 / Troy Technical
Figures in this article come from announcements on Figure AI’s website, its SEC Form D and a federal court docket. No research-firm estimates or press-based figures have been used. Demonstrations, deployments, shipments and production capacity are company claims for which we found no independent verification. Passages marked “Inference” are this article’s interpretation of primary sources, not statements made by the company. Valuations and investment views are not covered.

🌐 Japanese