Figure AI vs Unitree

AI Sophistication vs. Mass-Market Accessibility

The most consequential difference between Figure AI and Unitree is their approach to intelligence. Figure's Helix architecture — pairing a vision-language model with a vision-language-action model — represents perhaps the most advanced AI system deployed on a humanoid robot today. By separating scene understanding from motor control, Figure can update its reasoning capabilities independently, running inference at 200 Hz for real-time reactivity. The Figure 03's tactile fingertip sensors, detecting forces as small as three grams, enable manipulation precision that few competitors can match.

Unitree's AI stack is more pragmatic. Its locomotion is impressive — trained via reinforcement learning in simulation and transferred to real hardware — enabling dynamic walking, stair climbing, and even acrobatic demonstrations like martial arts routines and wall climbing. But Unitree's robots lack the kind of general-purpose reasoning that would let them autonomously interpret complex, unstructured tasks. Unitree's Dex5 hand shows manipulation promise, but the overall autonomy gap remains significant. The question is whether that gap matters at Unitree's price point.

The Economics of Humanoid Robotics

Figure AI's per-unit economics remain opaque, but with partner-only deployments and a $39 billion valuation built on potential rather than revenue, Figure is operating deep in the venture-funded territory of frontier technology development. The company's bet is that its AI capabilities will eventually justify premium pricing across massive markets — automotive, logistics, and ultimately household labor.

Unitree has already proven a viable business. With $257 million in 2025 revenue, a 335% growth rate, and adjusted net income of $90 million, Unitree is profitable and scaling. Its upcoming R1 humanoid at $4,900 could make humanoid robots as accessible as high-end consumer electronics. This mirrors the broader Chinese strategy visible in EVs and drones: use manufacturing scale and aggressive pricing to capture volume, then iterate on capability. The $610 million IPO filing on Shanghai's STAR Market suggests confidence in sustained growth.

Deployment Maturity and Real-World Validation

Figure AI has the stronger deployment story in high-value industrial settings. Its 11-month deployment at BMW's Spartanburg plant — where Figure 02 robots contributed to the production of 30,000 cars — represents one of the most substantive real-world humanoid deployments to date. The transition to Figure 03, with pilot testing at BMW Leipzig beginning summer 2026, shows a credible pathway from prototype to production workhorse in automotive manufacturing.

Unitree's deployments are more numerous but less demanding. The G1 excels as a research and development platform, and its open SDK and new App Store (with 1,200+ developers) are building the kind of ecosystem that drives long-term adoption. But Unitree has not yet demonstrated sustained deployment in a high-stakes industrial production environment. The H2, shipping in April 2026, may change this calculus as Unitree targets industrial and domestic applications with a full-size platform.

Hardware Philosophy

Figure 03 was redesigned from the ground up for both mass manufacturing and home environments. At 5'8" and 61 kg, it's human-scale — designed to operate in spaces built for people. The soft materials, multi-density foam coverings, wireless inductive charging, and washable textiles signal a company already designing for household deployment, not just factory floors. A 2.3 kWh swappable battery delivers 5 hours of runtime.

Unitree's G1 is deliberately compact — 4'3" and 35 kg — which limits its capability envelope but enables the aggressive price point. It's a robot you can pick up and move. The new R1 doubles down on this compact, affordable philosophy, while the H2 addresses customers who need a full-size platform. Unitree's product lineup strategy — offering multiple form factors at different price points — resembles a consumer electronics company more than a traditional robotics firm.

Ecosystem and Developer Community

Unitree has a significant lead in ecosystem development. The Go2 quadruped, starting at $1,600, has become one of the most widely used platforms in reinforcement learning research and computer vision development. This installed base feeds directly into Unitree's humanoid ecosystem. The launch of the Robot App Store — with applications spanning logistics, manufacturing, and service robotics — is a strategic move to become the platform on which third-party robotics applications are built.

Figure AI's ecosystem is more vertically integrated. Its proprietary teleoperation fleet generates training data at scale, and the Helix architecture is designed as a foundation model inference platform. Figure is building a closed ecosystem optimized for capability, while Unitree is building an open ecosystem optimized for adoption. Both strategies have historical precedents — Apple vs. Android, if you like — though the humanoid robotics market is far too early to know which will prevail.

Geopolitical and Strategic Context

The Figure AI vs. Unitree comparison cannot be separated from the broader U.S.–China technology competition. Figure AI's investor base — Microsoft, NVIDIA, OpenAI, Intel, Jeff Bezos — reads like a who's who of American tech. Unitree's backers include Tencent, Alibaba, Ant Group, and China Mobile, with Chinese government subsidies supporting the broader humanoid robotics industry. Trade restrictions, tariffs, and export controls could significantly affect both companies' addressable markets.

China's "Big 5" humanoid firms — Unitree, AgiBot, UBTECH, Leju Robotics, and Fourier Intelligence — collectively shipped more humanoid units in 2025 than the rest of the world combined, accounting for nearly 90% of global shipments. This volume advantage creates a data flywheel: more deployments generate more real-world data, which improves AI models, which makes robots more capable, which drives more deployments. Whether Western companies can match this velocity with superior AI will be the defining question of the industry over the next several years.