Figure AI

Figure AI is an American robotics company founded in 2022 by Brett Adcock that designs and builds general-purpose humanoid robots — robots with a human-shaped body (two arms, two legs, a head) that are powered by advanced artificial intelligence. The company's flagship products are Figure 01 and Figure 02, humanoid robots designed to work alongside human workers in factories, warehouses, and other industrial environments. Figure robots can walk, carry objects, manipulate components with their hands, understand natural spoken instructions, navigate complex environments with obstacle avoidance, and learn new tasks through AI training. The company's ultimate vision is to build robots capable of doing virtually any physical task a human can do — creating what Brett Adcock calls "a new category of robot: the general-purpose humanoid." Figure AI's robots are being commercially deployed at BMW manufacturing facilities and the company has partnerships with Microsoft, OpenAI, and NVIDIA.

Figure AI was founded by Brett Adcock in 2022. Brett is a serial entrepreneur from Illinois who studied Finance at the University of Illinois at Urbana-Champaign. Before Figure AI, he co-founded Vettery — an AI-powered job recruitment platform that was acquired by the Adecco Group in 2018 — and Archer Aviation — an electric air taxi company that he took public on the New York Stock Exchange in 2021 via SPAC merger. Figure AI is his third and most ambitious venture. Brett serves as both the Founder and CEO of Figure AI, playing a hands-on role in the company's technology development and business strategy. His track record of building, scaling, and successfully exiting previous technology companies in technically demanding fields (AI recruiting, electric aviation) has been a key factor in attracting elite investors and engineering talent to Figure AI.

Figure AI has raised over $675 million in funding, primarily through a landmark $675 million Series B round closed in February 2024 — one of the largest single funding rounds in the history of robotics. This round attracted an extraordinary roster of investors: Microsoft, OpenAI, NVIDIA, Jeff Bezos (personally), Intel, LG Innotek, Samsung, ARK Invest, Parkway Venture Capital, Align Ventures, and others. The round valued Figure AI at approximately $2.6 billion — making it one of the most valuable private robotics companies in the world less than two years after founding. The funding is being used to scale robot development and manufacturing, expand the engineering team, advance AI model development through the Helix system, grow commercial deployment programmes with enterprise partners, and establish the operational infrastructure needed to move from prototype scale to production manufacturing scale.

Figure 01 is the company's first commercial humanoid robot — a human-sized, AI-powered robot designed to work in physical environments built for people. It stands 5 feet 6 inches (168 cm) tall, weighs approximately 132 pounds (60 kg), and can carry payloads of up to 44 pounds (20 kg). Figure 01 uses a combination of cameras and depth sensors to see and navigate its environment in three dimensions. It is powered by the Helix AI system, which processes visual and sensor inputs to make decisions about how to move and interact with objects. The robot has articulated hands that can grasp and manipulate objects, and can walk on two legs with the dynamic balance needed to navigate real factory and warehouse environments. Figure 01 was the robot deployed at BMW's manufacturing facility in 2024 — the first commercial deployment of a Figure humanoid robot in a real industrial setting performing genuine production tasks.

Figure 02 is the significantly upgraded second-generation Figure robot, announced in mid-2024. Compared to Figure 01, Figure 02 has several major improvements. Its hands have 16 degrees of freedom — dramatically more dexterous than Figure 01's hands — allowing it to perform much more precise and complex manipulation tasks, including fine assembly work that requires near-human hand capability. Figure 02 integrates OpenAI's language models directly at the chip level, giving it the ability to understand natural spoken instructions and respond verbally in real time. It also has improved on-board AI computing hardware capable of running more sophisticated AI models, a better camera and sensor suite with higher resolution and better depth sensing, and is approximately 50% more energy efficient than Figure 01. Figure 02 represents a major step toward the level of capability needed for the broader range of tasks Figure AI ultimately wants its robots to perform.

In March 2024, Figure AI announced a landmark commercial partnership with BMW — one of the world's most respected and technically demanding car manufacturers. Under this partnership, Figure 01 humanoid robots are being deployed in BMW's manufacturing facilities in the United States to perform real production tasks alongside human workers. This deployment is historically significant because it represents one of the first times a general-purpose humanoid robot has been commercially deployed in an elite industrial manufacturing environment for genuine production work — not just a research demonstration. For Figure AI, the BMW partnership provides validation that its robots can perform useful work in one of the world's most demanding and quality-conscious industrial environments. For BMW, the partnership positions the company at the leading edge of manufacturing technology and gives it early access to humanoid robotic capabilities that could transform its production operations in the coming years.

Figure AI and OpenAI have a deep technical partnership that goes beyond a typical investor relationship. OpenAI has invested in Figure AI and is also providing its AI technology to power the language understanding capabilities of Figure robots. Specifically, OpenAI's language models are integrated into Figure 02 to enable the robot to understand natural spoken instructions in real time, reason about tasks, engage in dialogue with human coworkers, and explain its actions and reasoning verbally. This means a human supervisor can tell a Figure robot "that shelf needs restocking from the pallet in aisle 7" and the robot understands the instruction completely — no technical programming language needed. The partnership was demonstrated publicly when Figure 01 was shown having a full natural conversation with a researcher while simultaneously performing physical tasks, showcasing the seamless integration of language AI and robot physical intelligence that the OpenAI collaboration enables.

Figure AI has made safety the highest design priority for its humanoid robots — understanding that commercial enterprise adoption is impossible without demonstrable, certified safety for human co-workers. Figure robots incorporate multiple layers of safety systems. Mechanically, the robots have force-limited joints that cannot exert dangerous levels of force on people or objects, soft external materials on potentially contact-prone areas, and stability systems designed to prevent dangerous falls. From an AI perspective, the robots continuously monitor the positions and movements of people nearby and automatically slow down or stop when humans enter their safety zone. The robots are programmed with fail-safe behaviours that trigger safe stopping sequences if any sensor, computing component, or communication system fails or behaves unexpectedly. All commercial deployments undergo extensive safety testing and certification before robots work near human employees. That said, humanoid robots working alongside humans represent a genuinely new category of workplace equipment, and the regulatory and certification frameworks for ensuring their safety are still being developed in many jurisdictions.

Helix is Figure AI's proprietary AI model — the "brain" that powers its robots' intelligent, autonomous behaviour. Unlike traditional robot control systems that execute pre-programmed sequences of movements, Helix is a neural network that perceives the environment through the robot's sensors, reasons about what actions to take, and generates appropriate movement commands in real time. Helix is trained using multiple AI techniques: imitation learning (where the robot watches humans perform tasks and learns to replicate them), reinforcement learning (where the robot practices tasks in simulation and improves through reward signals), and real-world experience (where the robot learns from actually performing tasks in deployment environments). A key feature of Helix is collective learning — as robots accumulate experience in different environments and tasks, that learning can be shared across the entire Figure robot fleet, so every robot benefits from the experiences of all other robots. This creates a compounding intelligence effect where the AI becomes more capable as more robots are deployed and more tasks are performed.

Figure AI and Tesla Optimus are the two most talked-about humanoid robot companies, but they come from very different directions. Figure AI is a pure-play humanoid robotics company — everything it does is focused on building and deploying the best possible humanoid robot. Tesla, by contrast, is primarily a car and energy company that is developing Optimus as an internal tool for its factories and potentially as a commercial product. Both companies have demonstrated impressive robots and achieved some level of commercial deployment in their own facilities. Figure AI's advantage is its OpenAI language AI integration, its focused team of robotics specialists, and its commercial deployment at BMW — a third-party, highly demanding customer. Tesla's advantages include enormous manufacturing scale (if Optimus works, Tesla can produce it at car-factory scale), massive AI computing infrastructure through the Dojo supercomputer, and Elon Musk's influence and resources. Both companies are racing toward the same goal, and it is possible both can succeed in what may be a very large market.

Figure AI is initially targeting manufacturing and warehousing as its primary commercial markets. Manufacturing — specifically automotive manufacturing, as demonstrated by the BMW partnership — offers large, relatively structured environments where the use cases are well-defined, the economic value of the tasks is clear, and the customers (large manufacturers) have the budget and technical sophistication to be early adopters of new technology. Warehousing and logistics is the second primary target, driven by the enormous and growing labour shortage in e-commerce order fulfilment. After establishing strong positions in manufacturing and warehousing, Figure AI plans to expand into healthcare (physical support tasks in hospitals), hospitality, retail, construction, and other industries where physical labour is in shortage. The longest-term vision — home assistance and elder care — represents the largest ultimate market but requires the highest levels of robot capability and safety before it can be practically addressed.

Figure AI robots learn new tasks through several complementary methods that work together. The first is simulation training — before a robot ever attempts a new task in the real world, it practices thousands or millions of variations of that task in a virtual simulation environment. This allows the AI to learn the fundamental mechanics of a task quickly and safely before transferring to physical hardware. The second method is human demonstration — a human operator can show the robot how to perform a task by physically demonstrating it, and the robot's AI learns from observing the human's movements. The third method is real-world reinforcement — the robot attempts tasks in the actual environment and improves based on feedback about what worked and what did not. Finally, collective learning allows experience from all deployed robots to be aggregated and shared across the fleet — so if a robot in a German factory learns a better technique, that knowledge can be transferred to robots operating in American warehouses. Together these methods allow Figure robots to acquire new capabilities much faster than traditional robots, which require extensive manual programming for every new task.

Figure AI has not publicly disclosed the pricing for its robots, and the commercial model is still being developed as the company scales from prototype to commercial production. The likely business model is Robots-as-a-Service (RaaS) — where enterprise customers pay a monthly or annual fee per robot rather than purchasing robots outright. This subscription model is common in commercial robotics because it makes the economic decision for enterprise customers much easier: rather than a large capital expense for buying robots, they pay an ongoing operational expense similar to staffing costs, but with more predictable performance and availability. The per-robot subscription fee would need to be less than the cost of the human labour the robot replaces to make economic sense for customers. As production scales and manufacturing costs decrease — which has happened with every technology product from smartphones to solar panels — the subscription pricing can decrease to reach smaller customers, while margins improve as unit costs fall.

Figure 01, the first-generation robot, stands 5 feet 6 inches (168 cm) tall — roughly average human height — and weighs approximately 132 pounds (60 kg). It can carry payloads of up to 44 pounds (20 kg). It walks on two legs and can navigate varied terrain. Figure 02, the second-generation robot, maintains a similar overall form factor but has significantly improved hand dexterity with 16 degrees of freedom across both hands, improved sensor systems for better environmental perception, and more powerful on-board computing hardware for running more sophisticated AI in real time. Both robots are designed to fit within the physical envelope of a human — able to walk through standard doorways, work at human-height surfaces, sit in vehicles designed for people, and use tools and equipment designed for human hands. This human-scale design is central to Figure AI's strategy of deploying robots into existing human-designed environments without requiring expensive modifications.

This is one of the most important and debated questions around humanoid robotics. The honest answer is nuanced. In the short to medium term, Figure AI and similar companies are targeting jobs that are genuinely difficult to fill with human workers — physically demanding, repetitive, sometimes dangerous roles in manufacturing and warehousing that have persistent labour shortages. In this context, robots may fill roles that would otherwise be unfilled rather than displacing employed workers. In the longer term, as robots become more capable and more affordable, they will inevitably be able to do a wider range of physical tasks that humans currently perform. This could displace some workers, particularly in repetitive physical jobs. However, technology has historically created more jobs than it destroys — past waves of automation led to new categories of work that could not be anticipated beforehand. The transition will require societies to invest in worker retraining and reskilling, and policymakers will need to address the economic distribution questions that arise when robots significantly improve productivity. Figure AI, and its investors including Jeff Bezos and Microsoft, will need to engage constructively with these social questions as their technology scales.

Jeff Bezos invested personally in Figure AI — as distinct from his venture fund Bezos Expeditions which also makes investments — which signals this is a matter of personal conviction, not just portfolio management. Bezos has publicly expressed enthusiasm about robotics throughout his career, and Amazon under his leadership became one of the world's largest robotics operators in its warehouses. Bezos understands the economics of physical labour at enormous scale: Amazon employs hundreds of thousands of warehouse workers globally, and even small improvements in labour productivity or reductions in labour cost represent billions of dollars in economic impact. A humanoid robot that can perform warehouse tasks — particularly the complex, variable tasks that specialised robots cannot handle — could be transformative for companies like Amazon. Bezos's personal investment also reflects his general philosophy of backing transformative technologies early, when the outcome is uncertain but the upside is potentially enormous. He was an early investor in Google before it was a household name, for similar reasons of technological conviction at an early stage.

The future of humanoid robots is one of the most debated and exciting topics in technology. If companies like Figure AI succeed in their mission, humanoid robots could become as transformative as the personal computer or smartphone — ubiquitous physical AI assistants that change how physical work is done across every industry. In the nearer term (5-10 years), success means commercially deployed humanoid robots working at scale in manufacturing and warehousing, demonstrably improving productivity and safety for enterprise customers. In the medium term (10-20 years), improvements in AI and manufacturing costs could enable deployment in healthcare, hospitality, retail, and agriculture. In the longer term, home assistant robots — capable of cooking, cleaning, and supporting elderly people living independently — could reach consumer markets. The total addressable market is potentially enormous: Goldman Sachs has estimated the humanoid robot market could reach hundreds of billions of dollars by 2035. The key uncertainties are how quickly AI capabilities will improve, how manufacturing costs will fall as production scales, how regulatory frameworks will develop, and how quickly societies and workers will adapt to humanoid robots as a standard part of work environments.

Figure AI is headquartered in Sunnyvale, California — located in the heart of Silicon Valley. Sunnyvale is an ideal location for a robotics company: it is close to a dense concentration of robotics engineering talent, AI researchers, hardware engineers, and venture capital. It is also near many of Figure AI's technology partners including NVIDIA (Santa Clara), Intel (Santa Clara), and within commuting distance of the San Francisco offices of Microsoft and OpenAI. As of 2024, Figure AI has been growing its team rapidly following the $675M Series B funding round. The company has focused on attracting elite robotics engineers, AI researchers, and hardware specialists — many from leading robotics companies and research institutions. While Figure AI has not publicly disclosed its headcount, it has grown significantly from a small founding team to a substantial engineering organisation capable of developing and manufacturing advanced humanoid robots at an impressive pace.

Boston Dynamics and Figure AI are both building advanced robots, but they differ in age, approach, and focus in important ways. Boston Dynamics was founded in 1992 — making it over three decades old — and has built extraordinary robots that are famous for their physical movement quality: Spot (the quadruped robot dog) and Atlas (the humanoid). Boston Dynamics' robots demonstrate some of the most impressive dynamic motion in the industry. However, Boston Dynamics has historically been a research and engineering organisation first, with commercial applications coming more recently through Spot. Figure AI, founded in 2022, is focused from day one on commercial deployment and AI integration — the goal is not just impressive robot movement but robots that are useful enough to deploy in real factories and smart enough to understand natural instructions. Figure AI has integrated OpenAI's language AI at a level Boston Dynamics has not reached. Figure AI also has a fundamentally different ownership and funding structure: it is a venture-backed startup with aggressive commercial targets, while Boston Dynamics has gone through several ownership changes (MIT, Google, SoftBank, now Hyundai) that have reflected different commercial strategies. Both companies are impressive and serve different needs in the developing robotics market.

Figure AI robots are designed to work in any environment that was built for humans — which includes the vast majority of industrial and commercial environments. Because human workers are the design assumption for virtually every factory, warehouse, hospital, hotel, and retail environment in the world, a robot with human size and proportions can walk through the same doors, reach the same shelves, use the same equipment, and navigate the same aisles without any modification to the environment itself. This is Figure AI's key deployment advantage over specialised robots: a dedicated conveyor-belt robot requires infrastructure designed specifically for it; a Figure humanoid robot simply walks in and starts working. That said, there are environments where current Figure robots would face challenges: outdoor environments with rough terrain require more sophisticated locomotion than current indoor-optimised walking systems provide; extreme temperature environments (very hot or very cold) present hardware challenges; and highly unstructured environments with unpredictable obstacles require more sophisticated AI than is currently available. As Figure robots' AI and hardware improve through successive generations, the range of environments they can successfully operate in will expand.