Figure 02: The Robot Built for Real Work

Figure 02 launched in August 2024 as the most commercially advanced humanoid robot then available — a 60 kg, 1.70 m bipedal platform with 16 degrees of freedom in each hand, capable of carrying 25 kg, and already deployed at a BMW Group manufacturing facility in Spartanburg, South Carolina. While competitors were still running lab demos, Figure was generating real production data.

The Figure 02 represents a clear philosophical bet: ship early, learn from real environments, iterate fast. Founded in 2022 by Brett Adcock — who previously founded Archer Aviation and Vettery — Figure AI has moved from zero to factory deployment in under three years, backed by $675 million in Series B funding from a consortium that includes Jeff Bezos, Microsoft, Nvidia, Intel, Amazon, and OpenAI. By September 2025, a third round brought the total valuation to $39 billion.

Illustration

Figure 02 performing assembly tasks at the BMW Spartanburg plant. Source: Figure AI

Speed as Strategy: The Figure AI Story

Figure AI was founded in 2022 by Brett Adcock with an explicit mandate: build a commercially deployable humanoid robot faster than anyone thought possible. Adcock's prior companies — Vettery (acquired by Adecco) and Archer Aviation (NYSE: ACHR) — gave him a playbook for hard-tech startups: hire fast, move faster, and don't let perfect be the enemy of deployed.

Figure 01, released in early 2024, was the company's first public platform. It was deliberately optimized for speed-to-deployment over technical elegance — a robot capable enough to generate real-world data, not impressive enough to be a research marvel. That philosophy paid off: the BMW partnership announced in January 2024 gave Figure access to a real factory floor before the Gen 1 hardware was even fully proven.

Figure 02, announced in August 2024, represented the lessons learned from that first deployment wave. Every major subsystem was redesigned: integrated cabling replaced exposed wiring, a battery was moved into the torso for better weight distribution, compute was tripled with NVIDIA RTX GPU-based modules, and the hands were completely rebuilt with 16 DOF per hand and 25 kg carrying capacity. The result is a robot that's ready for structured industrial environments — not a demo, but a platform generating production hours.

Under the Hood: Figure 02's Hardware Architecture

Figure 02's hardware represents a clean-sheet redesign from Figure 01, driven by six months of real deployment data from the BMW facility. The engineering team identified three core failure modes in Figure 01: cable routing that snagged in tight spaces, insufficient compute for real-time VLA inference, and hands that lacked the fine-motor resolution for precision tasks. All three were addressed in Figure 02.

Hands: The Core Differentiator

The Figure 02 hands are arguably the most sophisticated on any commercially deployed humanoid. Each hand has 16 degrees of freedom — more than the Optimus (11 DOF in Gen 2, 22 in Gen 3 hands) and far more than Boston Dynamics Atlas (which prioritizes whole-body strength over hand dexterity). The hands can carry up to 25 kg and manipulate objects with enough precision to handle automotive assembly components without damage.

The follow-on Helix platform (announced February 2025) pushed this further with 35 total DOF, human-like wrist articulation, embedded palm cameras, and custom tactile sensors capable of detecting forces as small as three grams. Those sensors feed directly into the Helix VLA model, enabling closed-loop manipulation control that doesn't require pre-programmed grasp strategies for every object type.

Vision and AI Compute

Figure 02 is equipped with six RGB cameras providing overlapping fields of view — covering the workspace around both hands, forward locomotion, and overhead clearance. The cameras feed a custom onboard vision-language model running on NVIDIA RTX GPU-based modules, delivering 3× the inference throughput of Figure 01. The compute architecture is split between a high-level planner (System 2, running at 7–9 Hz) for task reasoning and a low-level controller (System 1, running at 200 Hz) for joint-level motion execution — a design shared with the newer Helix platform.

Illustration

Figure 02's five-fingered hands feature 16 degrees of freedom each — enabling precise manipulation of automotive components at the BMW Spartanburg plant.

Specs Comparison: Figure 02 vs. The Field

Figure AI: The Full Picture

Figure AI's investor roster is a signal of confidence as much as a source of capital. The February 2024 Series B round was led by a consortium that reads like a who's-who of technology and industrial capital: Jeff Bezos personally participated, alongside Microsoft, Nvidia, Intel, Amazon Industrial Innovation Fund, and OpenAI. The $675 million raised in a single round — at a $2.6 billion valuation — was among the largest single rounds in robotics history at the time. By September 2025, a third financing round valued the company at $39 billion, with investors including Qualcomm, T-Mobile, Salesforce, and Brookfield Asset Management.

The funding is being deployed in two directions simultaneously: accelerating Figure 02 hardware iteration and building the software stack. The March 2025 announcement of BotQ — a dedicated humanoid manufacturing facility in California targeting 12,000 robots per year — signals that Figure is now in production-scaling mode, not just R&D mode. BotQ is designed with an unusual recursive element: Figure's own robots will help build future robots, creating a compounding feedback loop for both manufacturing throughput and robot learning.

The October 2025 introduction of Figure 03 — a complete hardware and software redesign aimed at general-purpose home environments — further signals that Figure's commercial strategy spans both industrial B2B and eventual consumer B2C markets. Figure 03 adds wireless inductive charging, soft exterior materials for home safety, washable textiles, and UN38.3-certified battery safety layering. The separation of industrial (Figure 02, deployed at BMW) and consumer (Figure 03, built for homes) use cases positions Figure to pursue multiple addressable markets simultaneously.

What Figure 02 Means for Manufacturing

The BMW Spartanburg deployment isn't just a proof of concept — it's a commercial template. BMW isn't running a charity pilot; they're deploying robots where they generate measurable ROI. The specific tasks Figure 02 performs at Spartanburg — sheet metal handling, component transport, inspection assistance — are precisely the roles that are physically demanding, highly repetitive, and increasingly difficult to staff in the current labor market. Average turnover rates in automotive assembly exceed 30% annually; robot turnover is zero.

The implications for automotive manufacturing are significant. BMW's Spartanburg plant produces approximately 1,500 vehicles per day and employs over 11,000 people. Even a modest deployment of 50–100 humanoid robots at a facility that size represents a meaningful productivity uplift on specific tasks. More importantly, the data those robots generate — about task performance, failure modes, manipulation strategies — feeds back into Figure's AI training pipeline, accelerating capability development for future deployments.

The broader labor market signal is nuanced. Humanoid robots in 2026 are not replacing human workers wholesale — they're filling specific task slots in structured environments where repeatability and endurance matter more than adaptability. The more meaningful displacement risk in manufacturing is further out: as VLA models like Helix become capable of handling novel objects and unstructured environments, the task set robots can perform expands. Industry analysts at Goldman Sachs projected in 2024 that the humanoid robot market could reach $38 billion annually by 2035, which implies a significant acceleration of robotic penetration into manufacturing, logistics, and services.

The 12-Month Outlook: Figure in 2026–2027

Figure's near-term roadmap centers on three concurrent tracks. First, scaling the BMW deployment: the January 2024 partnership agreement outlined plans for up to 1,000 Figure robots at BMW facilities — the current deployment is in double digits. Reaching hundreds of deployed units will require both BotQ manufacturing capacity and the software reliability that comes from iterating on real deployment data.

Second, Figure 03 is targeting consumer markets in 2026–2027. The home robot use case is fundamentally different from automotive manufacturing — unstructured environments, unpredictable human cohabitation, extreme safety requirements. Figure 03's soft exterior materials, wireless charging, and multi-density foam coverings reflect genuine engineering investment in home-safe operation. BotQ's 12,000-unit annual capacity would need to scale substantially to meet any meaningful consumer demand, but the architecture is being established now.

Third, Helix VLA is Figure's platform play. The vision-language-action model that currently runs on Figure 02 is the same model that will run on Figure 03 — and on any future hardware. If Helix becomes the industry's leading generalist manipulation model, Figure doesn't just sell robots; they control the AI layer that makes humanoid robots commercially viable. That's a platform business, not a hardware business, and it's valued accordingly.

Frequently Asked Questions

Conclusion: Deployment Is the Moat

Figure 02 isn't the most technically impressive humanoid robot on the market. Boston Dynamics Atlas has more raw physical capability; Tesla Optimus has the advantage of scale manufacturing from a company that builds 1.8 million vehicles annually. But Figure 02 has something both competitors lack as of early 2026: a real commercial deployment generating real production data. Every shift Figure 02 works at BMW is a data collection session that makes the next version smarter.

At a $39 billion valuation with BotQ ramping toward 12,000 units per year and three robot generations in three years, Figure is executing with a pace that the field hasn't seen before. The question for 2026 and 2027 isn't whether Figure 02 works — it's whether the BMW deployment can be replicated across other OEMs, other industries, and eventually, homes. That's the $39 billion question.