Figure 02 Robot Starts Real-World Work at BMW

Figure AI has officially begun deploying its Figure 02 humanoid robot at BMW's manufacturing facility in Spartanburg, South Carolina, marking one of the most significant real-world tests of a general-purpose humanoid robot in industrial history. The deployment represents a critical leap from controlled lab demonstrations to the unpredictable, high-stakes environment of a working automotive factory.

This partnership between a leading robotics startup and one of the world's premier automakers signals that the humanoid robotics industry is moving beyond proof-of-concept into genuine commercial viability — a transition many experts predicted was still years away.

Key Facts at a Glance

• Figure 02 is the company's second-generation humanoid robot, featuring improved dexterity, AI reasoning, and battery life over its predecessor
• BMW's Spartanburg plant is the automaker's largest facility globally, producing roughly 1,500 vehicles per day
• The robots are initially handling parts inspection, bin picking, and component placement tasks alongside human workers
• Figure AI has raised over $750 million in funding, with a valuation exceeding $2.6 billion as of early 2024
• The partnership was first announced in January 2024, with phased deployment beginning in late 2024 and expanding into 2025
• Unlike single-purpose industrial arms, Figure 02 is designed to perform multiple tasks without hardware reconfiguration

Figure 02 Brings Major Hardware and AI Upgrades

The Figure 02 represents a substantial upgrade over the company's first-generation prototype. The robot stands approximately 5 feet 6 inches tall, weighs around 130 pounds, and features 16 degrees of freedom in its hands alone — giving it near-human dexterity for manipulation tasks.

One of the most significant improvements is the onboard compute system. Figure 02 runs custom AI models that allow it to perceive its environment, reason about tasks, and adapt to unexpected situations in real time. The robot's vision system uses multiple cameras and depth sensors to create a 3D understanding of its workspace.

Battery life has also been extended significantly. While the original Figure 01 could operate for roughly 2 hours on a single charge, Figure 02 pushes that to approximately 5 hours of continuous operation. This improvement alone makes the robot far more practical for factory shift work, where downtime for recharging directly impacts productivity.

The robot's AI stack integrates a large language model — developed in partnership with OpenAI — that allows it to understand natural language instructions and translate them into physical actions. This means factory supervisors can potentially reassign the robot to new tasks through verbal commands rather than complex reprogramming.

BMW's Spartanburg Plant Serves as the Ultimate Proving Ground

BMW's decision to open its Spartanburg facility for this test is not trivial. The plant is the company's largest worldwide, spanning over 7 million square feet and employing more than 11,000 workers. It produces popular models including the X3, X4, X5, X6, and X7 SUVs.

The factory environment presents challenges that no laboratory can replicate. Temperature fluctuations, noise, vibration from heavy machinery, unpredictable human movement, and the sheer pace of a production line all create conditions that stress-test every aspect of a robot's capabilities.

Initially, the Figure 02 units are stationed at specific workstations handling tasks that are ergonomically difficult or repetitive for human workers. These include:

• Sheet metal inspection — checking body panels for defects before assembly
• Bin picking — selecting and retrieving specific parts from unsorted containers
• Component insertion — placing clips, fasteners, and small components into assemblies
• Material transport — moving parts between nearby workstations
• Quality verification — using onboard sensors to confirm correct part placement

BMW has emphasized that the robots are not replacing workers but rather filling roles that are difficult to staff or that pose repetitive strain risks. This framing aligns with broader industry messaging around collaborative robotics.

How Figure Compares to Competitors in the Humanoid Race

Figure AI is far from alone in pursuing the humanoid robotics market. Tesla's Optimus (also known as the Tesla Bot) has been demonstrated performing simple warehouse tasks, though its real-world deployment timeline remains unclear. Boston Dynamics continues to develop its Atlas platform, which recently transitioned from a hydraulic to a fully electric design. Chinese competitors like Unitree and Agility Robotics with its Digit robot are also advancing rapidly.

What distinguishes Figure's approach is speed of commercialization. While many competitors remain in the demonstration phase, Figure has moved to actual factory deployment with a paying customer — a milestone that lends significant credibility.

Compared to traditional industrial robots like those from FANUC, ABB, or KUKA, humanoid robots offer a fundamentally different value proposition. Traditional robots excel at one specific task with extreme precision and speed. Humanoid robots sacrifice some of that specialization for versatility — the ability to navigate human-designed spaces, use human tools, and switch between tasks.

The cost equation remains a critical question. Traditional industrial robot arms can cost between $25,000 and $400,000 depending on capability. Figure has not publicly disclosed Figure 02's price point, but industry analysts estimate general-purpose humanoid robots will need to hit the $30,000 to $50,000 range to achieve mass adoption — a target that Figure CEO Brett Adcock has publicly acknowledged.

The AI Brain Behind the Machine

Perhaps the most underappreciated aspect of Figure 02 is its software architecture. The robot operates on a layered AI system that combines multiple approaches:

Foundation models handle high-level reasoning and task planning. When the robot receives an instruction like 'pick up the bracket and place it in the fixture,' the language model breaks this down into a sequence of sub-tasks.

Vision-language models connect what the robot sees with its understanding of objects, spatial relationships, and task context. This allows Figure 02 to identify parts it has never seen before and reason about how to grasp and manipulate them.

Reinforcement learning policies govern low-level motor control — the precise joint angles, grip pressures, and movement trajectories needed to execute physical actions smoothly. These policies are trained in simulation and then fine-tuned on real hardware.

This multi-layered approach represents the state of the art in embodied AI, a field that seeks to give physical robots the kind of general intelligence that large language models have brought to text and code generation. The BMW deployment provides invaluable real-world training data that will accelerate improvement across all three layers.

What This Means for Manufacturing and the Workforce

The implications of successful humanoid robot deployment extend far beyond a single factory. The global manufacturing sector faces a well-documented labor shortage. In the United States alone, the National Association of Manufacturers projects that 2.1 million manufacturing jobs could go unfilled by 2030 due to retiring workers and a lack of new entrants.

Humanoid robots that can be rapidly trained and redeployed across tasks could help address this gap without requiring the massive infrastructure changes that traditional automation demands. Because humanoids are designed to work in human spaces, factories would not need to be redesigned around the robots.

For businesses evaluating this technology, several practical considerations emerge:

• ROI timeline — Early deployments are expensive, but costs will decrease as production scales
• Integration complexity — Humanoid robots require robust wireless infrastructure, safety systems, and AI management platforms
• Regulatory landscape — Workplace safety standards for humanoid robots are still being developed by agencies like OSHA and international bodies
• Workforce transition — Companies will need retraining programs to help workers collaborate with and supervise robotic coworkers
• Data security — Robots with cameras and AI processing raise questions about workplace surveillance and data handling

Looking Ahead: From Pilot to Production Scale

Figure AI's near-term roadmap focuses on expanding the number of Figure 02 units at the BMW Spartanburg plant throughout 2025, with the goal of demonstrating reliable multi-shift operation across multiple task types. Success at BMW could trigger partnerships with other automakers and manufacturers.

The broader humanoid robotics market is projected to reach $38 billion by 2035, according to Goldman Sachs estimates. That forecast assumes that key technical hurdles — battery life, dexterity, AI reliability, and cost — continue to improve at current rates.

For Figure AI, the BMW deployment is more than a product test. It is a proof point for an entire industry thesis: that general-purpose humanoid robots can deliver economic value in real-world settings today, not in some distant future. If the Spartanburg deployment succeeds, it will accelerate investment, competition, and adoption across the sector.

The age of humanoid robots in the workplace is no longer a question of 'if.' With Figure 02 clocking in at BMW, the question has become 'how fast.'