Learning the use case with AI is the only path here
In the unfolding contest between nations and corporations to define the future of physical labor, Figure AI has secured a second major American partner and set a target of 100,000 humanoid robots deployed within four years — a threshold that would transform robotics from a promise into an infrastructure. The announcement arrives as China accelerates its own strategic ambitions in the field, making the question of who scales first not merely a business matter but a civilizational one. At stake is whether the next generation of machines that work alongside humans will be shaped by one set of values, incentives, and engineering cultures — or another.
- Figure AI's CEO has publicly committed to shipping 100,000 humanoid robots in four years, backed by a mystery partner described only as 'one of the biggest U.S. companies,' igniting immediate speculation across the industry.
- The pressure is not just commercial — China has declared humanoid robotics a national strategic priority, with manufacturers like Agibot already claiming production of 1,000 units by late 2024, compressing the window for American firms to establish dominance.
- Figure's robots are evolving at a pace that suggests real momentum: Figure 02 moves seven times faster than its predecessor, and the company is already testing a third generation in the lab just over two years after incorporation.
- Rather than programming robots with fixed instructions, Figure is deploying end-to-end neural networks that allow machines to learn tasks from experience — a bet that adaptability, not rigidity, is the only viable path to scale.
- The strategy of concentrating on a few high-volume customers rather than many small ones means that if execution holds, each deployment becomes a massive AI training engine, turning scale itself into a compounding technological advantage.
Figure AI has announced a deal with a second major commercial customer — identity undisclosed, though CEO Brett Adcock described it as one of the largest companies in the United States — that puts the firm on a path to shipping 100,000 humanoid robots over the next four years. Combined with its existing partnership with BMW, the announcement marks a shift from robotics as aspiration to robotics as operational commitment.
What sets Figure apart is the speed of its iteration. The company launched its first humanoid robot just 31 months after incorporation. Its successor, Figure 02, moves at nearly 2.7 miles per hour — seven times faster than the original — and a third generation is already running in the laboratory. The progression reflects genuine engineering momentum rather than incremental refinement.
Central to Figure's approach is artificial intelligence as the primary teaching mechanism. Rather than writing explicit instructions for every task, the company uses end-to-end neural networks that allow robots to learn from experience and adapt to environmental variation. Adcock has described watching these systems work as feeling like 'pure magic' — a phrase that captures both the novelty and the genuine capability now emerging from the lab.
The competitive landscape is intensifying. Tesla, Agility Robotics, and a growing field of Chinese manufacturers are all pursuing similar goals, with China having declared humanoid robotics a national strategic priority and targeting deep industrial integration by 2027. For Figure, the race is not just about building better robots — it is about building enough of them, fast enough, that the data generated in the field accelerates every subsequent generation. Scale, in this contest, is not just a business metric. It is the engine of the technology itself.
Figure AI has signed a second major commercial customer, and the company's leadership is now talking openly about shipping 100,000 humanoid robots over the next four years. The identity of this new partner remains confidential, though CEO Brett Adcock has described it as "one of the biggest U.S. companies," which has sparked immediate speculation about whether it might be a large retailer or a technology firm with substantial labor demands. The announcement arrived via LinkedIn, where Adcock emphasized the company's preference for deep partnerships with high-capacity customers rather than spreading resources across many smaller clients. "Between both customers, we believe there is a path to 100,000 robots over the next four years," he wrote.
Figure AI's first publicly identified customer is BMW, which signed on late last year. The automotive giant's involvement signals that the company sees real commercial value in deploying these machines at scale. The new deal, combined with BMW's commitment, positions Figure as a serious contender in an industry that has long seemed perpetually on the horizon but never quite arrived in everyday operations. What makes this moment different is the volume target: 100,000 units in four years is not a lab projection or a theoretical roadmap. It is a concrete business commitment tied to actual customers with actual needs.
The company's speed of execution sets it apart from competitors. Figure AI launched its first humanoid robot, Figure 01, just 31 months after the company was incorporated. Figure 02 followed, and the company is already running Figure 03 in the laboratory. The progression matters because it shows iteration happening at a pace that suggests genuine engineering momentum rather than incremental tweaks. Figure 01, which debuted in January 2024, moved at roughly 17 percent of average human walking speed. Figure 02 has increased that pace sevenfold, now traveling at 3.9 feet per second, or about 2.68 miles per hour. Still slower than a typical human stride, but the trajectory is clear.
What distinguishes Figure's approach is its reliance on artificial intelligence to teach robots how to perform tasks rather than programming them with explicit instructions. Adcock noted that the company recently began running an end-to-end neural network on the new customer's specific use case. "Learning the use case with AI is the only path here as heuristics would be impossible to write," he explained. This matters because it means the robots can adapt to variations in their environment and learn from experience rather than following rigid, pre-written routines. Each time these AI systems work through a task, Adcock said, it "feels like pure magic"—a telling phrase that captures both the genuine capability and the sense that something genuinely novel is happening.
The commercial opportunity is real, but so is the competition. Tesla's Optimus project, Agility Robotics, Unitree, Apptrovik, Sanctuary AI, and Agibot are all pursuing similar goals. The competitive pressure is sharpest from China, where the government has made humanoid robotics a strategic priority. In December 2024, Agibot announced it could produce 1,000 humanoid robots by year's end. MagicLab introduced its third-generation humanoid, MagicBot, designed to compete with established players like Boston Dynamics' Atlas. These moves align with China's 2023 strategy to embed humanoid robots deeply into its industrial base by 2027, aiming for widespread adoption across multiple sectors.
For Figure AI, the path forward depends on execution. Shipping 100,000 robots to real-world environments over four years would create an enormous feedback loop for AI training and design refinement. Each robot in the field would generate data, reveal problems, and suggest improvements. The scale itself becomes a competitive advantage: more robots in more locations means more learning, faster iteration, and lower costs through volume manufacturing. Adcock's emphasis on high-capacity customers rather than scattered small deployments reflects this logic. A single major customer with thousands of robots in operation is worth far more than dozens of smaller clients with a handful each. The question now is whether Figure can deliver on the timeline and whether the new customer—whoever it is—can absorb and effectively deploy machines that are still being refined in real time.
Notable Quotes
Between both customers, we believe there is a path to 100,000 robots over the next four years.— Brett Adcock, CEO of Figure AI
Learning the use case with AI is the only path here as heuristics would be impossible to write.— Brett Adcock, on Figure's approach to robot task adaptation
The Hearth Conversation Another angle on the story
Why does it matter that Figure AI won't name this second customer?
Because the secrecy itself tells you something. If it were a mid-sized logistics company, they'd probably announce it for the publicity. The fact that it's staying quiet suggests it's either a household name that doesn't want to be associated with job displacement concerns, or it's so large that the deal is already locked in and doesn't need marketing.
The 100,000 robots in four years—is that actually feasible, or is it a number designed to impress investors?
It's ambitious but not impossible if you have the manufacturing capacity and the customer demand. The real constraint isn't engineering anymore; it's whether these robots can actually do useful work reliably enough that companies want to deploy them at scale. Figure's betting that AI learning solves that problem faster than traditional programming would.
Why is the speed increase from Figure 01 to Figure 02 so significant?
Because speed correlates with capability. A robot that moves seven times faster can cover more ground, respond quicker, and handle more tasks per shift. It's not just about velocity—it's a proxy for how much the engineering has improved in a very short time.
The neural network approach—why is that different from what competitors are doing?
Most robotics companies still rely heavily on pre-programmed instructions. Figure is saying: teach the robot to learn the task itself. That's a fundamentally different bet. It's riskier because you can't fully predict what the robot will do, but it's more flexible because the robot adapts to variations humans didn't anticipate.
How worried should Figure be about the Chinese competition?
Very. China has government backing, lower labor costs, and a massive domestic market to test in. But Figure has something China's companies don't yet have: major Western customers like BMW who will push them to meet real commercial standards. That's worth a lot.