Why Figure matters
Figure AI is the best-capitalized pure-play humanoid-robot company in the world, and the clearest test yet of a specific bet: that the winner in general-purpose robotics will be whoever vertically integrates the AI, the hardware, and the factory that builds the robots — all under one roof. Most rivals specialize in one layer. Figure is trying to own all three at once, and its September 2025 Series C valued that ambition at $39 billion, an extraordinary mark for a company with only nascent commercial revenue.
The thesis is that humanoids are the right form factor for the largest automation market that exists — human work in spaces built for humans. Rather than redesign warehouses and homes around robots, Figure builds a machine shaped like a person, trained on a single neural network, that can in principle be pointed at sorting packages today and tidying a kitchen tomorrow. Whether that generality is real or marketing is the central question hanging over the entire sector, and Figure has staked more capital on the affirmative answer than anyone else.
From stealth to scale in four years
Founder and CEO Brett Adcock is a serial entrepreneur — he previously built Vettery (sold to Adecco) and the eVTOL company Archer Aviation — and he started Figure in 2022 by self-funding a $100M seed round. The pace since has been unusually fast: Figure exited stealth in early 2023, had a robot walking dynamically within months, and by February 2024 had raised a $675M Series B at a $2.6B valuation from an investor list that read like a who's-who of AI and tech — Microsoft, the OpenAI Startup Fund, NVIDIA, Intel Capital, Amazon's industrial fund, and Jeff Bezos.
Each robot generation marked a strategic shift. Figure 01 proved the company could build a walking humanoid at all. Figure 02, unveiled in 2024, was a workforce-oriented redesign that validated the company's first real factory deployment. Figure 03, introduced in 2025, was re-architected around two goals the earlier robots were never built for: running Figure's own Helix AI, and being manufactured cheaply at very high volume.
The OpenAI break and the Helix bet
The defining decision in Figure's history came in February 2025, when Adcock ended the company's collaboration agreement with OpenAI — an unusual move given OpenAI was both a partner and an investor. His stated reasoning was that solving embodied AI at scale requires vertical integration: a robotics company that leases its 'brain' from a foundation-model vendor is structurally dependent on a potential competitor's roadmap. Weeks later, Figure unveiled Helix, its in-house vision-language-action model.
Helix is built as a dual-system architecture. A slower 'System 2' — an internet-pretrained vision-language model — handles scene understanding, language, and goal sequencing, while a fast 'System 1' visuomotor policy translates that reasoning into high-frequency joint commands. The January 2026 Helix 02 release extended control to the entire body, unifying walking, balance, and manipulation in one network rather than separate stacks. This is the riskiest part of Figure's strategy — and, if it works, the deepest moat.
What the robot can actually do
Figure's most credible proof point to date is autonomy under load. In May 2026 the company live-streamed four Figure 03 humanoids running a fully autonomous 8-hour package-sorting shift on Helix 02 — detecting barcodes, picking packages, and orienting them onto a conveyor at roughly human throughput — and later pushed extended runs past 40 hours and 22,000 packages without reported failure. Sustained, unsupervised, multi-hour operation is the threshold that separates a demo from a deployable worker, and few competitors have shown it.
Figure 03 also gained genuine mobility. A 2026 upgrade to the low-level 'System 0' controller added perception-conditioned whole-body control, letting the robot climb stairs and cross ramps and uneven terrain using only onboard stereo cameras — trained end-to-end in simulation across thousands of randomized terrains and transferred zero-shot to the real machine. The robot stands around 5'6"–5'8", runs roughly 4–5 hours per charge, and is built with softer surfaces, palm cameras, tactile fingertips, and wireless charging aimed at both commercial floors and, eventually, the home.
Manufacturing as the real product
Figure's least-discussed but possibly most important asset is BotQ, its own humanoid factory. A robot company that cannot build robots cheaply and reliably will never reach scale, and Figure has treated manufacturing as a first-class engineering problem. In early 2026 it reported a 24x throughput improvement in under 120 days — from one Figure 03 per day to one per hour — with 350+ units produced, 80%+ end-of-line first-pass yield, and battery packs shipping at 99.3% first-pass yield.
The stated trajectory is aggressive: roughly 55 robots per week and climbing toward an initial 12,000-units-per-year line, with a longer-term goal of 100,000 robots over four years. Adcock has also said Figure is decoupling its supply chain from China. Hitting these numbers is what would let Figure undercut rivals on price and flood the market — and it is exactly the kind of manufacturing ramp that has humbled many hardware companies before.
Business model and early commercial traction
Figure's commercial model is robot-as-a-service: customers pay for robot labor rather than buying machines outright, with third parties referencing pricing in the neighborhood of ~$1,000 per robot per month. Its first paid deployment was an 11-month Figure 02 program at BMW's Spartanburg plant, where the fleet ran daily shifts and contributed to tens of thousands of vehicles before being retired and its lessons rolled into Figure 03.
The more important signal came in May 2026, when Catalyst Brands — parent of JCPenney, Aéropostale, and Brooks Brothers — signed a commercial agreement to deploy Figure 03 fleets at its Reno, Nevada distribution center, one of the first public commercial humanoid agreements in U.S. retail logistics. A separate strategic partnership with Brookfield opens a large footprint of residential, office, and logistics real estate as future deployment and training environments. The open question is whether these convert from pilots into standing, paying production at scale.
Funding, financials, and what to watch
Figure has raised on the order of $1.9B across seed, Series A, the $675M Series B, and a Series C that exceeded $1B in September 2025 at a $39B post-money valuation, backed by Parkway Venture Capital, Brookfield, NVIDIA, Intel Capital, Microsoft, Qualcomm, Salesforce, and others. That war chest is the company's biggest advantage in a capital-intensive race; revenue, by contrast, remains early and undisclosed.
The bear case is straightforward: a $39B valuation on minimal revenue prices in near-flawless execution, and humanoid demos have a long history of outrunning real-world reliability. The bull case is that Figure is further along on the three things that actually matter — in-house AI, a working high-volume factory, and signed commercial customers — than anyone else. The things to watch over the next year are concrete: does Helix 02 stay in continuous paying production at Catalyst Brands rather than one-off livestreams; does BotQ hold its yield as volumes climb; do home pilots materialize inside the Brookfield footprint; and does Figure's funding edge translate into a durable lead before better-resourced entrants like Tesla and Boston Dynamics close the gap.