Brett Adcock, founder and CEO of humanoid robotics company Figure AI, confirmed this year that he dissolved Figure’s technology partnership with OpenAI — a collaboration that had provided some of the AI models powering Figure’s humanoid robots — stating directly that his own internal AI research team was “running circles” around OpenAI’s robotics-specific efforts. The decision reflects Figure’s broader strategy of building its own foundation models specifically for robotics rather than relying on general-purpose AI labs whose primary focus remains chatbots and software agents rather than physical, embodied robots operating in the real world.
Figure AI’s own valuation has grown sharply alongside this strategy shift: the company reached a $39 billion post-money valuation following a Series C round that closed with more than $1 billion in committed capital in September 2025 — a roughly 15-fold increase from Figure’s $2.6 billion valuation in early 2024, making it the most valuable pure-play humanoid robotics company by a significant margin. The company hit record production levels at its BotQ manufacturing facility in March 2026 and has said it plans to ship 100,000 humanoid robots over the next four years, with BotQ scaling toward an annual production capacity of 12,000 units.
Real Commercial Deployments, Not Just Demos
Figure has pointed to concrete commercial deployments as evidence its robots are performing genuine, sustained work rather than staged demonstrations: an 11-month deployment at BMW’s Spartanburg, South Carolina plant during which Figure’s robots loaded more than 90,000 sheet-metal parts supporting production of over 30,000 X3 vehicles, and a fully autonomous 30-hour warehouse shift the company has cited as a landmark test of its robots’ ability to operate without constant human supervision. Adcock has separately said he expects Figure’s robots to eventually remove nearly all Chinese-manufactured components from its hardware by summer 2026, part of a broader push toward domestically sourced supply chains for critical robotics hardware.
A Second, Very Different Venture on the Side
Beyond Figure AI, Adcock has applied what he’s described as the same high-intensity operating approach to Cover, a school-safety security startup using terahertz imaging radar technology licensed from NASA’s Jet Propulsion Laboratory, with beta units planned for deployment to real schools by the end of 2026 — a notably different application of advanced sensing technology from Figure’s own consumer- and industrial-robotics focus.
A Founder Known for an Unusually Intense Operating Style
Adcock has built a public reputation across his ventures — Figure AI and, before it, Vettery and Hello Alfred — around an unusually high-intensity, all-consuming approach to running a company, one he’s described applying identically to the security startup Cover as to Figure’s own robotics work. That operating style is directly relevant to understanding why Adcock was willing to walk away from a partnership with an AI lab as prominent as OpenAI once he judged his own team’s internal capability had surpassed it — a decision that assumes a level of confidence in an internal team’s speed and quality that a more cautious, consensus-driven founder might have been reluctant to act on this early. It’s worth noting that Adcock’s own account of the OpenAI split — his team simply outpacing OpenAI’s robotics work — has not been independently confirmed or disputed by OpenAI itself in public statements, and should be read as Adcock’s own characterization of the decision rather than a jointly agreed account from both companies. Figure’s own next real test will be whether its internally built models can sustain the same pace of improvement at scale that justified walking away from an outside partnership in the first place — a bet the company’s ambitious 100,000-robot shipping target over the next four years will make increasingly difficult to hide from public view either way — real, deployed robots operating continuously in real facilities are a far harder claim to overstate than a research paper or a stage demo. That real-world accountability is arguably a healthier long-term discipline for the whole humanoid robotics category than the more speculative, demo-driven hype cycles that have characterized parts of the industry so far.
What This Means for Philippine Founders
Adcock’s decision to end a partnership with a much larger, more prominent AI lab once his own team’s internal capability caught up and surpassed it is a genuine lesson for any Philippine startup currently relying on an outside AI vendor’s models or technology: that dependency is worth continuously reassessing as a startup’s own technical capability matures, rather than treating an early-stage partnership as a permanent, unquestioned arrangement. Figure’s real, sustained industrial deployments (the BMW plant work in particular) are also a useful benchmark for any Philippine manufacturing, logistics, or industrial-automation company evaluating humanoid robotics vendors: a company that can point to nearly a year of continuous real-world factory deployment, not just a controlled demo, is operating at a genuinely different maturity level than most robotics startups can currently claim.
Share this article