At BMW’s sprawling Spartanburg, South Carolina, plant, where roughly 11,000 workers assemble more than 1,500 vehicles every day, a new kind of colleague has started clocking in. Figure, the robotics startup founded in 2022 by serial entrepreneur Brett Adcock, has entered a formal commercial agreement with BMW Manufacturing to deploy its general-purpose humanoid robots on active production lines. It is the first publicly confirmed deal to place walking, multi-task humanoid machines inside one of the world’s highest-volume automotive factories.
Adcock has gone further in public statements reported by Bloomberg and other outlets, claiming that Figure’s own manufacturing line can now produce one humanoid robot roughly every 30 minutes, that the company has built around 350 units to date, and that some of those machines are running 24/7 shifts with no human supervisor on the floor. If those numbers hold up under independent scrutiny, Figure would be operating the fastest humanoid production line on the planet and running one of the most aggressive autonomous-labor deployments in manufacturing history.
Those are big “ifs.” Here is what can be confirmed, what cannot, and why the distinction matters for workers, investors, and anyone watching the accelerating race to put humanoid robots to work.
What the BMW deal actually says
The partnership announcement, distributed through PR Newswire in January 2024, confirms that Figure and BMW Manufacturing have agreed to explore placing general-purpose humanoid robots on BMW’s automotive production lines. The release names both parties, describes the robots as general purpose, and outlines joint work on defining use cases, evaluating safety procedures, and integrating the machines into existing logistics and quality-control workflows.
The phrase “general purpose” is doing heavy lifting. Traditional industrial robots are bolted to fixed stations and programmed for a single repetitive motion: welding one seam, tightening one bolt. Figure’s humanoid, designated the Figure 02, is designed to walk between stations, manipulate different objects with dexterous hands, and adapt to changing conditions using onboard AI that includes a custom vision-language model for task planning. Deploying that kind of machine inside a real production environment represents a commercial bet that the technology has graduated from controlled demos to handling at least some of the variability of live assembly work.
The press release language, however, points to a phased rollout. References to “defining use cases” and “evaluating safety procedures” suggest early deployments will be limited in scope, closely monitored, and targeted at specific tasks where robots can slot in without disrupting the broader line. That is standard practice when automakers trial new automation. It is not the same as handing an entire shift to unsupervised machines.
The production and autonomy claims
The most eye-catching numbers, one robot every 30 minutes, 350 units built, and fully unsupervised 24/7 operation, do not appear in the BMW partnership announcement or in any audited filing reviewed for this report. They trace primarily to statements Adcock has made in interviews and on social media through early and mid-2025, amplified by outlets including TechCrunch and Bloomberg.
These claims are plausible on paper. Figure closed a $675 million Series B round in February 2024 at a reported $2.6 billion valuation, with backing from Microsoft, OpenAI’s startup fund, NVIDIA, Amazon founder Jeff Bezos, and other heavyweight investors. That war chest could fund rapid production scaling. And Adcock has a track record of moving fast: he previously founded Archer Aviation, which went public via SPAC in 2021.
But plausible and verified are not the same thing. No independent manufacturing audit, no third-party production log, and no regulatory filing has confirmed the unit count or the cycle time. The distinction between supervised and unsupervised operation is especially important. A robot working alongside a human operator who can intervene when sensors fail or software misbehaves presents a fundamentally different risk profile than one running through an overnight shift with zero human presence. If the latter is truly happening at scale, it would set a new benchmark for autonomous industrial labor. Without primary documentation, such as safety certifications, operational logs, or statements from plant management, the claim remains a company-sourced assertion.
Equally absent from the public record are performance metrics. No reviewed source contains data on robot error rates, task completion speeds, maintenance intervals, or downtime frequency. Numbers like mean time between failures, average cycle time per task, and comparative defect rates versus human workers would all help clarify whether the technology is delivering results that justify scaling or whether early deployments are hitting the mechanical and software hiccups common in first-generation systems.
The workforce question no one has answered publicly
BMW’s Spartanburg complex is the company’s largest plant by production volume worldwide and a major employer in upstate South Carolina. Introducing humanoid robots into that environment raises immediate questions about labor impact, and so far, neither Figure nor BMW has addressed them in detail.
In unionized European plants, organized labor groups typically negotiate terms around automation that govern shift structures, retraining programs, and job-security guarantees. Spartanburg’s workforce is not unionized under the United Auto Workers, which changes the negotiating dynamics but does not eliminate the concerns. Workers and labor advocates will want to know which tasks are being targeted first, whether the robots are intended to augment existing roles or eventually replace them, and what retraining or transition support is on the table.
BMW has publicly framed automation investments at Spartanburg as responses to labor shortages and ergonomic challenges rather than headcount reduction. Whether that framing holds as humanoid capabilities expand is an open question. The company’s quarterly financial disclosures and any future statements from plant leadership will be the most reliable indicators of how the deployment is actually affecting jobs.
Where Figure fits in the humanoid race
Figure is not building humanoid robots in a vacuum. Tesla has been developing its Optimus robot and has shown it performing warehouse tasks at its own facilities. Agility Robotics began deploying its Digit robot at an Amazon fulfillment center in late 2023. Apptronik, backed by funding from Google parent Alphabet, is targeting logistics and manufacturing with its Apollo humanoid. Chinese firms including Unitree and Fourier Intelligence are also scaling production.
What separates the BMW deal is the partner’s profile. BMW operates one of the most quality-obsessed production systems in global manufacturing. A willingness to trial humanoid robots on its lines, even in a limited pilot, signals that at least one tier-one automaker believes the technology has crossed a credibility threshold. If the pilot produces measurable gains in throughput or cost, other automakers will face pressure to follow. If it stalls or scales back, it will temper expectations across the sector.
How to track what happens next
For workers in automotive manufacturing and adjacent industries, the practical step is to start monitoring official disclosures rather than social-media hype. BMW’s quarterly and annual reports will eventually reflect any material changes in per-vehicle labor costs or capital expenditure tied to robotics. Occupational safety agencies may issue inspection reports or incident logs from the Spartanburg site that shed light on how autonomous operations are functioning day to day. Figure itself will likely need to share more granular data as it pursues additional partnerships or a potential public offering.
For the broader industry, the signal to watch is replication. If other major automakers announce similar humanoid-robotics agreements in the coming months, it will suggest the technology is crossing from experimental curiosity to mainstream consideration. If deployments remain confined to narrow pilots at a single plant, the technical or economic barriers are still significant.
As of June 2025, the most accurate description of Figure’s work with BMW is that it marks an important but early step in testing general-purpose humanoid robots on real automotive production lines. The partnership is confirmed. The bolder claims about production speed, fleet size, and full autonomy are not yet independently verified. The human consequences, for Spartanburg’s 11,000 workers and for manufacturing labor more broadly, will depend on how quickly verifiable evidence catches up with the ambition. Readers and decision-makers who hold that distinction clearly will be better positioned to separate signal from noise as the next wave of announcements arrives.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
