Figure AI Explained: Brett Adcock’s $39B Humanoid Robot (2026)

Brett Adcock has founded one logistics company (Vettery, sold for $100M), one electric-aircraft company (Archer Aviation, took public), and now Figure AI — the humanoid robotics company that’s currently leading the U.S. private market by valuation. He started Figure in 2022 with the explicit goal of building a general-purpose humanoid robot that could do useful work in factories, warehouses, and eventually homes. Three years later Figure has raised more capital than any pure-play humanoid company on earth and has a verifiable production-scale customer deployment at BMW.

This is the explainer. What Figure actually builds, what’s real versus marketing, who’s putting money in, and whether the company can deliver on the scale it’s selling.

What is Figure AI in plain English?

Figure AI is a Sunnyvale, California-based company that designs and builds humanoid robots — machines shaped like people, with two arms, two legs, hands, and cameras for eyes — that are intended to do physical work that’s currently done by human employees. Their main customer profile is industrial: car factories, warehouses, logistics operations. The pitch to those customers is that one Figure robot can replace one human worker for a 24/7 schedule at lower total cost than the human’s wages plus benefits.

The reason this is plausible in 2026 and was not in 2016 is that the robot’s brain is now a foundation model — software trained on enormous amounts of video, language, and action data the way large language models like ChatGPT are trained on text. The robot can be shown a new task and learn it from demonstration, rather than needing engineers to hand-code every motion. Figure calls its in-house model Helix.

Who founded Figure AI and what’s their background?

Brett Adcock founded Figure in 2022. He’s a serial founder who built and sold Vettery (a recruiting platform) to Adecco for around $100 million, then co-founded Archer Aviation (the eVTOL electric-aircraft company that’s now publicly traded). Adcock’s track record matters because building a humanoid robotics company is roughly the same engineering challenge as building an electric aircraft — actuators, batteries, control software, supply chain — applied to a different form factor. The investors who backed Archer also backed Figure on partly that thesis.

The company’s early team was assembled aggressively from Boston Dynamics, Tesla, Apple, and IHMC. Adcock has been public about preferring to hire people who’ve already built shipping robotics products rather than career academics. The bet is that Figure’s competitive advantage is execution speed, not technical novelty.

What robots has Figure built?

Robot	Year	Status	Notable specs
Figure 01	2023–2024	R&D / early demo	First proof-of-concept; demonstrated walking, manipulation, OpenAI integration
Figure 02	2024–2025	BMW Spartanburg deployment	Built 30,000+ cars at 99% accuracy; the workhorse of Figure’s commercial proof
Figure 03	Unveiled Oct 2025	Production-grade, ramping in BotQ	5’8″, 61 kg, 20 kg payload, wireless feet-charging at 2kW, ~5hr battery, 9% lighter than 02, tactile fingertips sensing 3g of force, 60% wider camera FOV

Figure 03 is the model that matters going forward. It was officially unveiled October 9, 2025 after weeks of teaser videos from Adcock. The headline design changes versus Figure 02 are practical: it’s smaller and lighter (humans can move it more easily), the meshed fabric covering plus foam padding reduces injury risk for human coworkers, and the wireless charging integrated into the feet means the robot can refuel by standing on a charging pad rather than being plugged in. The tactile fingertips are the upgrade that matters most for dexterous manipulation — sensitivity down to 3 grams of force lets the robot handle delicate items without crushing them.

What does Figure’s BMW deployment actually look like?

BMW’s Spartanburg, South Carolina plant is the largest BMW plant in the world by output. Figure 02 robots have been deployed there since 2024 on body-in-white tasks — specifically sheet-metal placement and parts handoffs in the unbuilt-car-frame stage of vehicle assembly. As of early 2026, BMW and Figure publicly reported that the robots had contributed to 30,000+ vehicles built with 99% accuracy on the tasks assigned to them, and that BMW saw approximately 40% reduction in per-unit cost on those specific operations.

The caveat: BMW and Figure are reporting their own numbers. There is no independent third-party audit of those figures. The numbers are credible because Figure’s investors would have audited them as part of the Series C process, and because BMW’s reputation is on the line — but anyone evaluating Figure should treat the 99%/40% numbers as company-reported rather than industry-verified.

Importantly: BMW’s European pilot at Plant Leipzig, announced February 2026, will start with a different humanoid — Hexagon Robotics’ AEON, a Zurich-based wheeled humanoid — rather than Figure 03. BMW and Figure said publicly they are evaluating Figure 03 for additional use cases. Read it however you like, but the implication is that Figure has not yet won the next BMW deployment outright. That matters for the “Figure dominates auto” narrative.

What is Helix, Figure’s foundation model?

Helix is Figure’s in-house Vision-Language-Action (VLA) model — the software that processes what the robot sees (vision), what it’s told to do (language), and decides what physical actions to take (action). It’s analogous to Google DeepMind’s RT-2, Physical Intelligence’s Pi Zero, or Tesla’s Optimus brain. Helix is closely held; Figure has shared demonstration videos and high-level architecture posts but has not published the model weights or full architectural details.

The strategic bet here is that Figure’s vertically integrated stack (Figure designs the hardware, the model, the deployment software, and the operating-environment integrations all in-house) produces a tighter feedback loop than the alternative where you license a foundation model from a separate company. The opposing bet — taken by Skild AI, by Physical Intelligence, and arguably by Hyundai/Boston Dynamics through the Google DeepMind partnership — is that foundation models will become commoditized and the platform/hardware split will win.

What is BotQ and why does it matter?

BotQ is Figure’s purpose-built robot manufacturing facility. It came online in 2025 with an initial design capacity of 12,000 humanoids per year, and Figure has stated a plan to scale BotQ (or BotQ-like facilities) to 100,000 robots per year over the following four years. The company’s supply-chain target — to source 3 million actuators in that window — is the actuator-volume equivalent of building enough motors for 200,000+ humanoids if you assume 15 actuators per robot.

This is the part of the Figure story that’s easy to miss. A startup that’s never built more than a few hundred robots committing to a multi-million-actuator supply chain is signaling that they intend to be a manufacturing company, not a research lab. Whether they hit those numbers is the open question; the size of the commitment itself is what’s noteworthy.

Who has invested in Figure?

Figure’s investor list reads like a roster of the AI infrastructure stack: Microsoft, NVIDIA, Intel Capital, OpenAI Startup Fund, Jeff Bezos (personal), Parkway Venture Capital, Brookfield Asset Management, Macquarie Capital, Salesforce, T-Mobile Ventures, and Qualcomm Ventures are among the named backers. Across multiple rounds the company has raised approximately $1.9 billion. The Series C in September 2025 was the round that took the post-money valuation to $39 billion.

The investor mix is unusual and informative. NVIDIA invests where it sees long-term chip demand. Microsoft and OpenAI invest where AI workloads will land. Bezos invests where physical-world workflows can be reinvented (Amazon being the obvious adjacency). Salesforce and T-Mobile are operational customers in waiting. The combination signals that strategic investors believe Figure could be one of the platforms onto which AI-driven physical labor gets deployed, not just a vendor of one product.

How does Figure compare to its main competitors?

Company	Robot	Capital raised	Production-scale deployments	Foundation-model approach
Figure AI	Figure 02/03	~$1.9B	BMW Spartanburg (30K cars)	Vertically integrated (Helix)
Boston Dynamics	Atlas (electric)	Owned by Hyundai	In production March 2026 (Hyundai RMAC, Google DeepMind)	Co-developed with Google DeepMind
Tesla	Optimus	Inside Tesla Inc.	Internal Tesla factories; V3 ramps mid-2026	Closely-held (Dojo / FSD-adjacent)
Agility Robotics	Digit	Mid-100s of millions	GXO (100K+ totes), Amazon, Toyota Mfg Canada	In-house
Apptronik	Apollo	~$1B, $5B valuation	Mercedes-Benz pilot, GXO R&D	Co-developed with Google + Mercedes
1X Technologies	NEO	Mid-100s of millions	Pre-orders for home; commercial pilots	OpenAI-aligned
Unitree	G1 / H1	Private (Chinese)	Research labs globally; industrial pilots	In-house

Figure is the most capitalized and has the most-cited commercial deployment. Boston Dynamics has the longer technical pedigree and now the first formal enterprise-grade production status. Tesla has by far the largest potential manufacturing scale but the smallest track record of customer-facing deployments. Agility has the most actually-deployed humanoid hours in the field, just at smaller scale per site. None of these companies will be the only winner; the realistic 2030 picture has at least three of them at meaningful scale.

What can go wrong for Figure?

• The Helix bet doesn’t pan out. If a third-party foundation model (Pi Zero, Skild Brain, a DeepMind release) becomes the obviously-best brain across humanoid hardware, Figure’s vertical-integration premium evaporates.
• The BMW economics don’t generalize. If Figure 03 deploys at Mercedes, Toyota, Ford, or another major auto OEM and the per-unit-cost reduction is far below 40%, the early Figure story looked better than the steady state.
• BotQ doesn’t ramp. Building robots at scale is the hard problem. Tesla has demonstrated production-ramp risk repeatedly with cars; humanoid production is harder than cars.
• Chinese competition on price. Unitree at $16K, UBTech, Fourier, AgiBot all push the price floor in ways that hurt premium U.S. humanoid pricing power.
• Foundation-model commoditization. If “good enough” robotics foundation models become widely available, the differentiation moves to hardware, supply chain, and deployment know-how — areas where Figure is good but not uniquely so.
• Regulation in human-shared spaces. Industrial deployments are easy to certify. Hospital, retail, and home deployments are not, and the unsolved safety-standards problem could delay Figure’s expansion beyond manufacturing.
• Adcock’s track record. Adcock is impressive but the failure mode for charismatic-founder companies is “moves too fast, ships before ready, breaks customer trust.” This is a real risk that observers track.

Frequently asked questions

Can I buy a Figure robot?

Not as a consumer. Figure sells to large industrial customers under enterprise contracts. Public list pricing isn’t published. Figure 03 was designed with the eventual home market in mind, but home delivery is years out.

Is Figure publicly traded?

No. Figure is a private company. There has been speculation about an IPO given the $39 billion valuation and the maturity of the funding cadence, but no formal announcement as of mid-2026.

How does Figure 03 compare to Tesla Optimus?

Figure 03 is in production now at BotQ. Tesla Optimus V3 has been announced for summer 2026 launch but is not yet shipping at the same scale, and Tesla has not yet placed Optimus at a non-Tesla customer at production volume. On hardware, Optimus is targeting much higher hand dexterity (22 degrees of freedom per hand, 50 actuators across both); Figure 03 is targeting tactile sensitivity in the fingertips. Different bets. Both real engineering investments.

Does Figure compete with Amazon’s warehouse robots?

Indirectly. Amazon’s own warehouse robotics are mostly non-humanoid — Kiva-derived autonomous mobile robots, sortation systems, robotic arms. Amazon has invested in Agility Robotics (whose Digit humanoid works at Amazon sites), not Figure. So far, Figure’s commercial footprint is automotive and industrial assembly, not Amazon-style retail logistics. That could change.

What’s the difference between Figure and Boston Dynamics?

Boston Dynamics has 30+ years of robotics research history and is owned by Hyundai Motor Group. Their humanoid Atlas (in its new electric form, in production since March 2026) ships in a much smaller volume than Figure plans for but with a deeper technical pedigree. Figure is a 4-year-old startup taking a manufacturing-first approach to scale. Both deserve to win. Both probably will, in different markets.

Should I trust Figure’s 99% accuracy and 40% cost reduction numbers from BMW?

Provisionally yes, with caveats. The numbers are company-reported (by both Figure and BMW), they were available to Figure’s Series C investors during due diligence, and BMW would not put its reputation behind numbers it didn’t believe. They are not independently audited as of public reporting. Treat them as credible benchmarks for the specific tasks Figure 02 was deployed on, not as universal claims about humanoid economics.

Sources

• Figure AI — Figure 02/03 product information and announcements
• BMW Group — BMW Physical AI deployment in production
• Brett Adcock — Founder profile and writings
• The Robot Report — Industry coverage of Figure AI deployments
• Humanoids Daily — Sector news and Figure 03 unveiling coverage

You may also like

• AI in Robotics in 2026: The Complete Guide
• What is a Humanoid Robot? (Glossary)
• What is Embodied AI? (Glossary)
• What is Optimus (Tesla Robot)?
• AI in Drones in 2026
• What Is a Large Language Model?