Humanoid Robot Race Heats Up as China’s Unitree and US Rival Figure Take Divergent Paths
The global humanoid robot industry is reaching an inflection point, with capital shifting from backing speculative technology to funding companies with clear paths to commercialization. In this high-stakes race, China's industry leader Unitree and U.S.-based Figure AI represent two starkly different strategies, raising critical questions for investors about which model will ultimately dominate the nascent trillion-dollar market.
The contrast has been thrown into sharp relief by recent developments. Figure AI last month unveiled its third-generation robot, Figure 03, and secured a staggering US$ 39 billion valuation following a US$ 1 billion funding round, making it the world's most valuable humanoid robot startup. The company has aggressively marketed its product for both home and commercial use, claiming a production line ready to manufacture 100,000 units annually.
Meanwhile, in China, market frontrunner Unitree is facing intensified scrutiny. While it continues to dominate in public procurement contracts from universities and state-owned enterprises, the company is being questioned over its conservative AI development strategy and a perceived lack of breakthrough commercial applications. The pressure mounts as investors in 2025, dubbed the "first year of commercialization," demand to see tangible orders and revenue streams.
This divergence encapsulates a core debate shaping investment in the sector: whether to support pragmatic, hardware-focused firms with steady, albeit less glamorous, revenue, or to place massive bets on high-valuation, AI-centric visionaries promising a transformative future. The outcome will not only determine the leaders of the next technological wave but also signal the viability of humanoid robots in the global economy.
Race for Orders Signals Market Shift
The mantra for humanoid robot companies in 2025 is clear: secure orders or risk being left behind. As one investor noted, "If you don't have mass production capabilities and the ability to generate your own revenue, we're not looking." This sentiment is driving a frenzy of commercial activity across China. UBTECH Robotics has reported nearly 500 million yuan in contracts for its Walker series robots, with some deliveries scheduled for 2025. Agibot has also garnered orders worth hundreds of millions of yuan for its new G2 model.
The trend extends beyond the top players. Leju Robotics announced an RMB 82.95 million yuan (approximately US$11.5 million) order in July, while EngineAI Robotics has a strategic agreement for at least 2,000 units over three years. Similarly, Noetix Robotics has surpassed 2,500 total orders for its N2 unit, and Galbot expects to deploy around a thousand robots this year. For robotics firms seeking funding in 2025, having orders in the hundreds or thousands is becoming a prerequisite.
China’s Leader Faces Questions on AI, Vision
Despite its leading position, exemplified by winning 25 public procurement bids so far in 2025, Unitree finds itself on the defensive. Critics question why the industry leader isn't pioneering advancements in robotic "brains" and "dexterous hands." The company’s R&D spending—totaling approximately 350 million yuan over the past three years, with less than 20% allocated to algorithms—pales in comparison to rivals like Agibot, which spends 400-500 million yuan on large-model development annually.
Adding to the skepticism, Unitree CEO Wang Xingxing has publicly criticized the mainstream Vision-Language-Action (VLA) architecture for AI models, arguing it is inefficient and not a viable path forward. His company has instead developed its own UnifoLM-WMA-0 world model, a move seen by some as out of step with prevailing industry trends. While Wang defends the company's valuation as justified from a long-term perspective, its strategy is viewed by some as overly "conservative" and a potential risk if a major AI breakthrough occurs, a sentiment underscored by reports earlier this year of investors seeking to sell Unitree shares.
Figure AI’s $39 Billion Bet on a Consumer Future
In stark contrast to Unitree's pragmatism, Figure AI embodies ambitious, future-forward marketing. The newly launched Figure 03 was presented in slick videos featuring fabric coverings and multiple "outfits," a clear nod to its consumer-friendly aspirations. The company emphasizes that its focus on solving complex tasks in the home will produce a versatile robot capable of handling any job. However, this high-flying vision is accompanied by significant skepticism.
Figure has yet to provide a live, in-person demonstration of its product, with all information coming from heavily produced demo videos. The company has faced accusations of imitating a rival’s design and was the subject of a Fortune magazine report in April that questioned the substance of its partnership with BMW, alleging exaggeration. CEO Brett Adcock has furiously denied these claims but also acknowledged that the company’s goal of a fully autonomous home robot is not yet a reality, setting a target of 2026. This "hype first, build later" approach is not unique to Figure, with even Tesla’s Optimus project reportedly facing delays.
Industry Navigates ‘Pilot Purgatory’
Across the industry, the gap between demonstration and real-world value remains wide. Many high-profile deployments are still limited to pilot programs or promotional events, such as UBTECH's robots performing inspections at a NIO's factory or Tesla's Optimus appearing at a Disney movie premiere. These applications are often dismissed by critics as "subsidized demand" from partners or investors rather than genuine, scalable commercial use. Unitree's CEO has defended the value of its extensive work with universities, noting it has contributed to over 3,000 academic papers.
Historically, such early-stage support has been crucial. The rise of Japan’s robotics industry and the U.S. semiconductor sector were both fueled by long-term government and corporate investment that de-risked early development. However, significant technical and business model hurdles persist today, from hardware limitations like overheating motors to the difficulty of creating standardized solutions that work across different industrial clients. As the industry matures, a consensus is forming that the robot's "brain," or its AI model, constitutes 80% of its value. With China’s supply chain already capable of producing commoditized hardware, the ultimate winner will likely be the one that cracks the code on intelligence, regardless of its current strategy.