Unitree's listing on the STAR Market signals the humanoid robotics industry's official entry into an era of public market pricing, marking not just a single company's debut but a systematic recalibration of the entire sector's valuation framework. J.P. Morgan views this IPO as a catalyst that will transition the sector from a "scarcity premium-driven proxy trade" toward price discovery based on pure-play listings, with valuation anchors being reset and capital flowing toward companies demonstrating quantifiable commercial execution and scalable delivery capabilities.

Retail demand for the Unitree IPO reached an extraordinary 5,000-8,000 times oversubscription, with the offer price set at RMB 150.80 per share, corresponding to a market value of approximately RMB 61 billion. On its first trading day, the stock surged over 460%, pushing the company's valuation to RMB 341.8 billion. Meanwhile, AgiBot is preparing for a Hong Kong listing with a target valuation of roughly USD 20 billion (approximately RMB 145 billion), implying a forward P/S ratio of around 35x based on FY26 estimates. Both valuations stand significantly above those of existing listed robotics OEMs, which will compress the scarcity premium enjoyed by incumbents like UBTech and Estun Automation, while driving more granular market pricing.

At the strategic level, J.P. Morgan anticipates a 1-6 month digestion period for the sector: the increase in new supply and the resetting of valuation benchmarks will shift capital allocation away from the "only listed pure-play" logic toward rewarding companies with verifiable shipment records, multi-customer replication capabilities, and a clear path to gross margin improvement.

The IPO Wave Ends the Scarcity Premium, Ushering in an Event-Driven Pricing Phase

Unitree's listing represents a sector-wide valuation reset rather than an isolated corporate event. Looking at the IPO pipeline, following Unitree, AgiBot is expected to complete its Hong Kong listing within 6-12 months. In the United States, Agility Robotics has completed a merger with Churchill Capital Corp XI via a SPAC deal at a USD 2.5 billion valuation, while Figure AI is expanding its production capacity to 12,000 units annually with a target IPO in 2027-2028. As pure-play entities enter public markets, the scarcity premium on existing proxy names will be systematically compressed.

J.P. Morgan points out that this compression is structurally inevitable: pressure from new supply arises not only from investors needing to reduce existing positions to free up capital for new subscriptions, but also from market behavior, as IPO candidates are often viewed as "purer new-cycle plays," temporarily diverting incremental demand away from incumbents. The depth of the digestion cycle will depend on the macro and sentiment backdrop: if risk appetite holds, the sector can quickly absorb new supply and reprice; if macro conditions turn volatile, markets will demand a higher margin of safety and penalize names lacking near-term catalysts.

Primary market funding activity confirms this divergence. July recorded 34 deals totaling approximately RMB 12 billion, a notable decline from June's 47 deals and RMB 32 billion peak. Capital is shifting from "thematic financing" to "capability-driven financing," with markets distinguishing between hardware execution, model/data capabilities, and the quality of real-world deployment.

Physical AI Differs From the LLM Market: Adoption Curves Resemble Smartphones or Autonomous Driving

A core premise for understanding the humanoid robotics sector is recognizing its structural differences from large language models (LLMs). LLM platforms like ChatGPT, Anthropic, Kimi, Zhipu, and MiniMax can scale distribution at near-zero marginal cost, whereas humanoid robots require manufacturing, transportation, deployment, maintenance, and supervision, while bearing safety and legal liabilities in case of failure. This distinction redefines what constitutes a "milestone": multi-customer commercial reliability, quantifiable fleet learning loops, and ultimately safe in-home deployment are the real inflection points, not a singular "ChatGPT moment."

The market structure therefore follows a "winner-takes-more-by-layer" rather than "winner-takes-all" dynamic: demand is highly fragmented across structured factory settings, semi-structured logistics environments, and unstructured household scenarios, while varying regional safety regulations further support a multi-OEM landscape. However, significant concentration effects persist in bottleneck segments, namely component suppliers and horizontal platforms that become the "default choice." This is precisely why J.P. Morgan's recent risk/reward assessment favors scalable component makers: these companies can benefit across multiple OEM product roadmaps without bearing the adoption risk of any single platform.

China's Robot Training Network Accelerates Physical AI Commercialization

China's top-down policy framework is systematically driving the industry from laboratory validation toward scaled production and real-world deployment. At the policy level, the "15th Five-Year Plan" designates intelligent robots as one of six emerging pillar industries, with embodied intelligence listed among six future industries. The 2026 national "Six Networks" infrastructure plan commits over RMB 7 trillion in investment, providing computing, connectivity, energy, and logistics infrastructure for large-scale robot and AI deployment. In February 2026, the Ministry of Industry and Information Technology issued the "2026 Humanoid Robot and Embodied Intelligence Standards System Construction Guidelines," China's first top-level standard covering the entire industrial chain and full lifecycle; in June, the MIIT and SASAC jointly launched the "2026 Special Action for Real-World Scenario Training of Humanoid Robots and Embodied Intelligence."

At the infrastructure level, China has established a nationwide robot training network: Shanghai's "Qilin" training ground currently hosts over 100 on-site robots, targeting 1,000 by 2027; Beijing and Hangzhou have deployed over 120 and 130 robots respectively, covering dozens of real-world training scenarios; regional hubs are emerging in Henan, Guangdong, and Shandong. This system constitutes a "data factory" for physical AI, designed to accelerate data accumulation, model iteration, and scaled commercial adoption.

On core component localization, the government has set a target of 80% domestic sourcing for critical components by the end of 2026. Planetary roller screws currently have a localization rate of approximately 40-50%, while six-axis force sensors remain below 30%, both designated as key focus areas. Major cities including Beijing, Shanghai, Shenzhen, Suzhou, Chengdu, Nanjing, Shaoxing, Wuhan, Hefei, Hebei, and Ganzhou have established government-backed robotics industry funds, each ranging from RMB 5-10 billion, specifically supporting OEMs, key components, and innovative applications. This has extended order visibility for leaders such as Leader Harmonious Drive Systems (harmonic reducers), Orbbec (3D vision), and Hengli Hydraulic (ball screws) through 2027.

For investors, the current priority is distinguishing "companies that can tell a story" from "companies that can prove scalable delivery," with shipment credibility, production cadence, gross margin improvement trajectories, and multi-customer replication capability emerging as the decisive variables for relative returns in the next cycle.