![Figure AI Review: Robots, Helix AI & Everything You Need to Know [2026]](https://cdn.prod.website-files.com/67ea6793adf74f7d3087e4e4/6996680d95006625b2dbbf5d_zl1Ak4A.jpeg)
Figure AI is one of the most ambitious humanoid robotics companies in the world. With backing from OpenAI, Microsoft, Jeff Bezos, NVIDIA, and Intel—plus a $39 billion valuation—they're redefining what humanoid autonomy looks like. Here's everything you need to know.
Figure AI has emerged as one of the most ambitious and well-funded humanoid robotics companies in the world. With backing from OpenAI, Microsoft, Jeff Bezos, NVIDIA, and Intel, plus a valuation that hit $39 billion in late 2025, Figure isn't just building robots—they're redefining what humanoid autonomy looks like.
In this comprehensive review, I'll break down everything you need to know about Figure AI: their robot lineup, the revolutionary Helix AI system, real-world deployment results at BMW, and whether Figure lives up to the hype.
Quick Summary: Figure AI at a Glance
| Aspect | Details |
|---|---|
| Founded | May 2022 by Brett Adcock |
| Headquarters | San Jose, California |
| Valuation | $39 billion (September 2025) |
| Total Funding | ~$1.7 billion+ |
| Key Investors | OpenAI, Microsoft, NVIDIA, Jeff Bezos, Amazon, Intel Capital |
| Robot Lineup | Figure 01 (retired), Figure 02 (commercial), Figure 03 (latest) |
| AI System | Helix / Helix 02 Vision-Language-Action (VLA) |
| Key Deployment | BMW Spartanburg plant (30,000+ vehicles produced) |
| Target Markets | Manufacturing, logistics, home assistance |
| Production Goal | 100,000 robots over next 4 years |
Who Is Figure AI? Company Background
Figure AI was founded in May 2022 by Brett Adcock, a serial entrepreneur who previously founded Archer Aviation (eVTOL aircraft, NASDAQ: ACHR) and Vettery (acquired by Adecco for $100M+). Adcock assembled Figure's founding team from alumni of Boston Dynamics, Tesla, Google DeepMind, and Apple—a who's who of robotics and AI talent.
The company's mission is deceptively simple: give AI a physical body. While chatbots and large language models have transformed digital interactions, Figure believes the real transformation happens when AI can manipulate the physical world—folding laundry, loading dishwashers, assembling products on factory floors.
Funding and Valuation Timeline
Figure's funding trajectory reflects extraordinary investor confidence in humanoid robotics:
| Round | Date | Amount | Valuation | Key Investors |
|---|---|---|---|---|
| Seed | 2022 | $70M | ~$300M | Parkway Venture Capital |
| Series A | May 2023 | $100M | ~$1B | Intel, Parkway |
| Series B | Feb 2024 | $675M | $2.6B | OpenAI, Microsoft, NVIDIA, Jeff Bezos, Amazon, Intel |
| Series C | Sept 2025 | $1B+ | $39B | Parkway, Brookfield, NVIDIA |
The February 2024 Series B was a watershed moment. Having OpenAI, Microsoft, and NVIDIA all invest in a robotics startup signaled that the biggest names in AI see humanoid robots as the next frontier. The partnership with OpenAI was particularly significant—it positioned Figure to leverage cutting-edge language models for robot reasoning.
Figure's Robot Lineup: From 01 to 03
Figure 01: The Prototype That Started It All
Figure 01 was the company's proof-of-concept humanoid, unveiled on March 2, 2023 and deployed in limited testing through 2023-2024. Standing at 5'6" (168 cm) and weighing 132 lbs (60 kg), Figure 01 demonstrated that Figure could build a functional bipedal humanoid.
Key Figure 01 specs:
- Height: 5'6" (168 cm)
- Weight: 132 lbs (60 kg)
- Payload: 44 lbs (20 kg)
- Battery Life: ~5 hours
- Degrees of Freedom: 40+
Figure 01's main purpose was learning—both for the company and for the AI models that would eventually become Helix. It was used for the initial BMW partnership testing and helped Figure understand what manufacturing environments actually demand from a humanoid.
Figure 02: The Commercial Workhorse
Figure 02 represented Figure's first commercially viable humanoid. Announced in 2024 and deployed at BMW's Spartanburg plant, Figure 02 proved that humanoids could work real shifts in real factories.
| Specification | Figure 02 |
|---|---|
| Height | 5'6" (168 cm) |
| Weight | 155 lbs (70 kg) |
| Payload Capacity | 44-55 lbs (20-25 kg) |
| Battery | 2.25 kWh lithium-ion |
| Runtime | 5+ hours |
| Compute | 3x more powerful than Figure 01 |
| Cameras | 6 onboard cameras |
| Hand DoF | 16 degrees of freedom per hand |
| Walking Speed | 1.2 m/s |
The Figure 02's key innovations were:
- Torso-integrated battery: Lowered center of gravity for better balance
- Improved actuators: Faster and more precise movements
- Enhanced perception: Six cameras for 360° environmental awareness
- Commercial durability: Designed for 10-hour shift reliability
After 11 months at BMW (1,250+ hours of runtime, 90,000+ parts loaded), Figure retired Figure 02 to make way for Figure 03. The lessons learned—particularly around forearm reliability and wrist electronics—directly shaped the next generation.
Figure 03: Built for Homes and Scale
Figure 03, introduced in late 2025, represents Figure's most ambitious robot yet. It's not just an industrial workhorse—it's designed to eventually enter homes.
| Specification | Figure 03 |
|---|---|
| Height | ~5'6" (estimated, similar to F.02) |
| Weight | 9% lighter than Figure 02 |
| Camera System | 2x frame rate, 25% latency, 60% wider FOV |
| Palm Cameras | Embedded in each hand for in-hand visual feedback |
| Tactile Sensors | Fingertip sensors detecting forces as small as 3 grams |
| Hand Design | Softer, more compliant fingertips |
| Battery | UN38.3 certified, multi-layer safety protection |
| Charging | 2 kW wireless inductive charging via foot pads |
| Data Offload | 10 Gbps mmWave wireless |
| Covering | Soft textiles (washable, replaceable) |
What makes Figure 03 special:
- Home-safe design: Multi-density foam, soft textile covering, reduced mass
- Tactile intelligence: Can feel a paperclip's weight (3 grams) with fingertip sensors
- Palm cameras: Visual feedback when main cameras are occluded (reaching into cabinets)
- Wireless everything: Inductive charging, wireless data offload—no cables needed
- Mass manufacturing ready: Designed for BotQ factory production at 12,000 units/year initially
Helix: The AI Brain Behind Figure's Robots
Hardware matters, but Helix is what makes Figure's robots genuinely intelligent. Helix is a Vision-Language-Action (VLA) model—a neural network that directly converts visual input and language commands into robot actions.
What Is a Vision-Language-Action Model?
Traditional robot programming works like this:
- Engineer writes code for each specific task
- Robot follows predetermined movements
- Any new task requires new code
Helix works differently:
- Robot sees environment through cameras
- Human gives natural language command ("Pick up the ketchup")
- Helix translates vision + language into motor actions—in real-time
This is transformative. Instead of programming thousands of individual behaviors, Figure can simply tell the robot what to do.
Helix Architecture: System 1 and System 2
Helix uses a "dual-system" architecture inspired by cognitive psychology:
System 2 (S2): The Slow Thinker
- 7-billion-parameter Vision-Language Model
- Operates at 7-9 Hz
- Handles scene understanding, language comprehension
- Pretrained on internet-scale data
- Produces semantic "latent goals" for S1
System 1 (S1): The Fast Reactor
- 80-million-parameter visuomotor transformer
- Operates at 200 Hz
- Translates S2's goals into precise motor commands
- Controls wrists, torso, head, individual fingers
- Handles real-time adjustments
This separation is elegant: S2 can "think" about what to do while S1 handles the split-second motor control needed for smooth movements. It's similar to how humans consciously decide to pick up a cup (slow, deliberate) while the actual reaching-and-grasping happens automatically (fast, reactive).
Helix Capabilities
With Helix, Figure robots can:
- Pick up virtually any object: Thousands of novel items via simple commands like "Pick up the desert item" (and Helix knows a cactus qualifies)
- Operate appliances: Drawers, refrigerators, dishwashers
- Multi-robot collaboration: Two robots working together on shared tasks
- Zero-shot generalization: Handle objects never seen during training
The "pick up anything" capability is particularly impressive. Helix learned from only ~500 hours of demonstration data—a fraction of what other VLA systems require—yet generalizes to thousands of novel objects.
Helix 02: Full-Body Autonomy (January 2026)
Helix 02, unveiled on January 27, 2026, extended the original Helix from upper-body control to full-body control. This is a massive leap.
The New System 0 Layer
Helix 02 adds a third layer to the architecture:
| System | Role | Speed | Parameters |
|---|---|---|---|
| System 2 | Scene understanding, language | 7-9 Hz | 7B |
| System 1 | Full-body joint targets | 200 Hz | 80M |
| System 0 | Whole-body balance & coordination | 1,000 Hz | 10M |
System 0 is trained on 1,000+ hours of human motion data and handles the physics of staying upright while moving and manipulating. It replaced 109,504 lines of hand-engineered C++ code with a single neural network.
The Dishwasher Demonstration
To showcase Helix 02, Figure released a 4-minute continuous task video: a humanoid robot autonomously unloading a dishwasher, walking across a kitchen, placing items in cabinets, reloading the dishwasher, and starting it.
Key stats from this demonstration:
- Duration: 4 minutes continuous
- Actions: 61 separate loco-manipulation actions
- Resets: Zero
- Human intervention: Zero
- Teleoperation: None
Figure called this "the longest horizon, most complex task completed autonomously by a humanoid robot to date." Whether or not that's strictly true, it's undeniably impressive—especially the seamless integration of walking, reaching, balancing, and fine manipulation.
New Dexterity Tasks
Helix 02's palm cameras and tactile sensors enable tasks that were impossible with vision alone:
- Extracting individual pills from a medicine organizer
- Dispensing precise syringe volumes (5 ml)
- Unscrewing bottle caps with controlled force
- Picking small metal pieces from cluttered bins
Real-World Results: The BMW Deployment
Talk is cheap. The real test of any industrial robot is whether it can survive a factory floor. Figure's 11-month deployment at BMW's Spartanburg plant provides hard data.
Deployment Overview
- Location: BMW Manufacturing, Spartanburg, South Carolina
- Duration: 11 months (deployed within 6 months of Figure 02 release)
- Task: Sheet-metal loading for welding fixtures
- Shift: 10 hours/day, Monday-Friday
By the Numbers
| Metric | Result |
|---|---|
| Parts Loaded | 90,000+ |
| Runtime Hours | 1,250+ |
| Vehicles Contributed To | 30,000+ BMW X3s |
| Estimated Robot Steps | 1.2+ million |
| Distance Walked | 200+ miles |
Key Performance Indicators
The task had strict requirements:
- Cycle time: 84 seconds total, 37 seconds for loading
- Placement accuracy: >99% success per shift (5 mm tolerance)
- Interventions: Zero per shift (goal)
The challenge: placing three sheet-metal parts within 5 mm tolerance in just 2 seconds—while moving fast enough to keep up with the line.
What Figure Learned
The BMW deployment wasn't just about proving capability—it generated invaluable data for Figure 03:
- Forearm reliability: The forearm was Figure 02's top failure point. For Figure 03, they eliminated the distribution board and dynamic cabling entirely.
- Thermal management: Tight packaging in the forearm created heat issues. Figure 03 uses redesigned wrist electronics.
- Field calibration: Consistent cross-robot performance required new calibration tools.
BMW hasn't announced plans to deploy Figure 03 yet, but the partnership validated Figure's approach to humanoid manufacturing.
Pricing and Availability
Current Status
Figure robots are not available for consumer purchase. As of February 2026:
- Figure 02: Retired from production (fleet returned to HQ)
- Figure 03: Early commercial deployments; not consumer-available until late 2026 at earliest
- BotQ Production: Ramping to 12,000 units/year capacity
Estimated Pricing
While Figure hasn't published official pricing, industry estimates suggest:
| Model | Estimated Price | Target Market |
|---|---|---|
| Figure 02 | $30,000-$50,000 | Commercial/industrial |
| Figure 03 | $50,000-$100,000 (speculative) | Commercial, eventually home |
For context, Tesla's Optimus is targeting ~$25,000-$30,000, while Unitree's G1 starts at $16,000. Figure is positioning higher on capability rather than competing purely on price.
Commercial Availability
Figure is currently focused on:
- Select commercial partners (manufacturing, logistics)
- Scaling BotQ production
- Building the supply chain for 100,000 robots over 4 years
CEO Brett Adcock has stated the goal is to have Figure 03 "in select homes" by late 2026, but this will likely be limited pilot programs rather than broad consumer availability.
Figure AI vs. Competitors
How does Figure stack up against other humanoid players?
| Company | Robot | Height | Weight | Price Est. | Key Strength |
|---|---|---|---|---|---|
| Figure | Figure 03 | 5'6" | ~140 lbs | $50-100K | Helix AI, full-body autonomy |
| Tesla | Optimus Gen 2 | 5'8" | 127 lbs | $25-30K | Scale, cost, Tesla ecosystem |
| 1X Technologies | NEO | 5'5" | 66 lbs | ~$30K | Lightweight, home-focused |
| Agility Robotics | Digit | 5'9" | 141 lbs | ~$200K+ | Logistics-optimized |
| Unitree | G1 | 4'3" | 77 lbs | $16K | Affordable, research-friendly |
| Boston Dynamics | Atlas | 4'11" | 196 lbs | Not for sale | Most athletic movements |
Figure's advantages:
- Most sophisticated VLA system (Helix 02)
- Proven factory deployment (BMW)
- Strong funding and tech partnerships
- Home-ready design (Figure 03)
Figure's challenges:
- Not yet at Tesla's scale ambitions
- Higher price point than budget competitors
- Home deployment still 1-2 years away
Pros and Cons
Pros
- ✅ Most advanced AI system: Helix 02's full-body autonomy is industry-leading
- ✅ Proven industrial deployment: 11 months at BMW with measurable results
- ✅ World-class investors: OpenAI, Microsoft, NVIDIA backing validates approach
- ✅ Home-safe design: Figure 03's soft materials and safety features
- ✅ Wireless charging & data: No cables = true autonomy
- ✅ Vertical integration: BotQ factory enables quality control and cost reduction
- ✅ Dexterous hands: 16 DoF hands with tactile sensing
Cons
- ❌ Not available to consumers: Commercial partnerships only for now
- ❌ High price point: More expensive than Tesla Optimus, Unitree G1
- ❌ Limited production: 12,000/year capacity vs. Tesla's mass-manufacturing ambitions
- ❌ Early-stage home capabilities: Dishwasher demos ≠ reliable home assistant
- ❌ Figure 02 retired: Previous generation already obsolete
- ❌ Battery life unclear: Figure 03 specs not fully disclosed
Frequently Asked Questions
How much does a Figure AI robot cost?
Figure hasn't released official pricing. Industry estimates suggest Figure 02 was $30,000-$50,000 for commercial deployments, and Figure 03 may be $50,000-$100,000. These robots are not currently available for consumer purchase.
Can I buy a Figure robot for my home?
Not yet. Figure 03 is designed with home environments in mind (soft materials, wireless charging, safety features), but consumer availability isn't expected until late 2026 at earliest—and even then, it will likely be limited pilot programs.
What is Helix AI?
Helix is Figure's proprietary Vision-Language-Action (VLA) model. It allows Figure robots to understand natural language commands ("Pick up the red cup") and translate them directly into motor actions. Helix 02, released January 2026, extended this to full-body control including walking and balance.
What happened to Figure 02?
Figure retired the Figure 02 fleet in November 2025 after the BMW deployment concluded and Figure 03 entered production. The lessons learned from Figure 02's 1,250+ hours of factory runtime directly informed Figure 03's design improvements.
How does Figure compare to Tesla Optimus?
Tesla Optimus is targeting lower price (~$25,000-$30,000) and higher volume (millions of units eventually). Figure is pursuing higher capability with Helix AI and has proven factory deployment. Tesla has more manufacturing scale; Figure has more sophisticated AI integration.
Is Figure AI publicly traded?
No. Figure AI is a private company. It has raised over $1.7 billion in venture funding at a $39 billion valuation (September 2025). There's no announced timeline for an IPO.
What can Figure robots actually do?
Based on demonstrated capabilities:
- Pick and place objects (including novel items)
- Load/unload dishwashers and appliances
- Navigate home and factory environments
- Operate drawers, refrigerators, cabinets
- Multi-robot collaboration on shared tasks
- Fine manipulation (pills, syringes, bottle caps)
Who are Figure AI's main investors?
Key investors include OpenAI, Microsoft, NVIDIA, Jeff Bezos (personal investment), Amazon, Intel Capital, Parkway Venture Capital, and Brookfield Asset Management.
The Bottom Line
Figure AI is building exactly what the humanoid robot industry needs: capable hardware paired with genuinely intelligent AI. The Helix system—especially Helix 02's full-body autonomy—represents the most sophisticated integration of language understanding and physical manipulation I've seen in a commercial humanoid.
The BMW deployment proves Figure isn't just making demo videos. 90,000+ parts loaded, 30,000+ vehicles contributed to, zero-reset shifts—that's real work.
But let's be clear: Figure 03 isn't ready for your living room yet. The dishwasher demos are impressive but carefully controlled. Real homes are chaotic, unpredictable, and full of edge cases that will test any AI system.
Who should care about Figure AI right now?
- Manufacturing companies exploring humanoid automation
- Logistics operations with repetitive physical tasks
- Investors tracking the humanoid robotics sector
- Robotics researchers and engineers
Who should wait?
- Consumers looking for home robots (check back in 2027)
- Anyone expecting sub-$30,000 pricing
- Those who need robots immediately
Figure has the funding, the talent, the AI, and the manufacturing roadmap. The question isn't whether humanoid robots will work in factories and homes—it's how fast Figure can scale. With 100,000 robots targeted over four years and a $39 billion valuation backing them, Figure AI is one of the most serious bets in robotics.
For individual robot reviews, see our Figure 01 Review, Figure 02 Review, and Figure 03 Review.
