Figure AI has opened its Sunnyvale "Robot Campus" to journalists, offering a detailed look at the infrastructure behind its Figure 03 humanoid robot. CEO Brett Adcock presented the "Never Fall" resilience protocol, 2kW wireless inductive charging built into the robot feet, and a consumer leasing model priced at $400–$600 per month. The tour revealed a company that has moved decisively from research prototype to production machine oriented around 24/7 reliability.
Key Takeaways
- The "Never Fall" protocol, powered by the Vulcan AI controller, allows Figure 03 to limp to a maintenance bay after complete mechanical leg failure
- All robots run on the Helix 02 VLA model installed on-board; no internet connection required for task execution
- 2kW wireless inductive charging built into the robot feet; hot-swap logistics ensure sub-30-second unit replacement
- Brett Adcock targets a lease model of $400–$600/month for home robots — modeled on the automotive industry
- Figure 4 is in design; Adcock describes it as the "iPhone 1 moment" for consumer robotics
A Campus Built to Stress-Test Robots
The Sunnyvale campus is more than a headquarters. Brett Adcock describes it as a stress-testing environment with the stated goal of having more humanoid robots walking the hallways than humans. By keeping robots operating continuously in the same disordered, unpredictable spaces that characterize real homes and factories, Figure AI validates its hardware and software under conditions that no controlled laboratory can fully replicate.
The company currently employs around 500 people and is scaling its internal fleet to "a few hundred" Figure 03 units. The campus includes a system integration lab with dedicated burn-in stations, where every robot must complete repetitive deep squat sequences and mobility tests before deployment.
The Never Fall Protocol and Vulcan
In commercial humanoid robotics, reliability matters more than raw capability. A robot that collapses during an overnight shift, unable to stand up, is not a deployable product.
Figure answer is the Vulcan project — a reinforcement-learned controller that allows Figure 03 to adapt to mechanical failure in real time. In a live demonstration, a robot with a locked left knee autonomously redistributed its center of mass, compensating for the dead joint and hobbling toward a service station without human intervention. The system can velocity lock a failing limb, enabling the machine to self-triage and reach a maintenance bay unassisted.
This capability is designed for what Adcock calls lights-out operations — 2 AM shifts where no human technician is present and the robot must assess its own condition, navigate to maintenance, and return to work.
Helix 02: On-Board and Offline
Every Figure 03 runs on Helix 02, a Vision-Language-Action neural network installed directly on GPU modules mounted in the robot torso. The model manages vast numbers of potential body states and executes all task logic locally, without requiring an internet connection. The machines carry 5G and Wi-Fi hardware, but these are used only for data synchronization and model updates — not for real-time control.
Adcock cites two reasons for this design choice. First, privacy: home customers do not want every robot motion streamed to a cloud server. Second, operational stability: in unseen environments — a first-time visit to a new apartment layout, an unfamiliar kitchen — the robot must function on local inference alone.
Wireless Charging and Hot-Swap Logistics
Figure has replaced plug-in charging with 2kW wireless inductive charging integrated into the robot feet. The standard operational cycle is four to five hours of work followed by one hour of charging. When a unit drops to approximately 10–15% battery, a fresh robot undocks and replaces it within under 30 seconds.
This hot-swap system is designed to eliminate the operational interruptions that plague shift-based deployments. Whether the task is industrial package sorting or domestic cleaning, the work does not stop when a battery runs low. The replacement cycle is designed to be transparent to the customer — no monitoring, no scheduling required.
The Economics of a Home Robot
Adcock has been publicly specific about the consumer pricing strategy. The model he describes is borrowed from the automotive industry: a monthly lease in the range of $400 to $600, giving households access to a humanoid without a large upfront purchase.
To reach this price point, Figure is focusing on production cost reduction and non-technician serviceability. The robots wear fabric suits with rear zippers, allowing unskilled users to replace damaged aesthetic components without opening the chassis. The high-top sneakers serve dual purposes: inductive charging dock interface and a new passive toe design that assists the robot when crouching or rising from the floor.
Figure 4, currently in design, is framed by Adcock as the product that will define the consumer humanoid category — the iPhone 1 moment for the industry. Like the original iPhone, it does not need to be perfect; it needs to be good enough, reliable enough, and affordable enough to open a market that does not yet exist at scale.
Why This Matters
Figure is the first humanoid company to present a coherent operational model for home deployment: not just hardware and software, but the full ecosystem — a reliability protocol (Never Fall), wireless power infrastructure, consumer leasing, hot-swap logistics, and modular servicing. This addresses the question the industry has been unable to answer for years: how does a humanoid robot become a product rather than a demonstration?
The contrast with competitors is instructive. Tesla Optimus, Boston Dynamics, and Agility Robotics remain focused on industrial and warehouse applications. Figure is the only company that has publicly named a consumer lease price and described the operational infrastructure required to support it at scale. If Adcock can maintain cost discipline and the reliability claims hold in field conditions, Figure may define the home robot category before anyone else has the chance to compete in it.
What's Next?
- Figure 4 is in design phase; no launch date announced — Adcock's "iPhone 1 moment" framing sets a high expectation for both price and market impact
- The leasing model requires financial and service partners; no agreements have been announced
- Transfer learning from the global fleet is the key technical test: does Helix 02 improve faster as the fleet scales?
Sources
- Humanoids Daily — Inside the Campus: Figure's Plan for 24/7 Autonomy and Leasable Home Robots - https://www.humanoidsdaily.com/news/inside-the-campus-figure-s-plan-for-24-7-autonomy-and-leasable-home-robots
- Humanoids Daily — From Pixels to Torque: Figure Unveils Helix 02 - https://www.humanoidsdaily.com/news/from-pixels-to-torque-figure-unveils-helix-02-and-the-era-of-whole-body-autonomy
- Humanoids Daily — Vulcan: Figure Unveils AI Balance Policy That Prevents Death from Hardware Failure - https://www.humanoidsdaily.com/news/vulcan-figure-unveils-ai-balance-policy-that-prevents-death-from-hardware-failure
