Why did JPMorgan significantly raise Unitree Robotics' target price to 995 yuan?

TL;DR

· The core of JPMorgan's shift is moving forward the time window for accelerated commercialization of embodied intelligence, thereby granting higher valuations to existing hardware platforms and developer ecosystems.

· Enhanced planning and reasoning capabilities of large models are further shifting the competitive focus of the robotics industry toward real-world execution, data collection, and feedback iteration.

· Unitree's installed base holds potential platform value, but whether cumulative production scale can translate into effective data and sustained revenue remains to be verified.

· Profit forecast adjustments are mainly concentrated in the medium term: the 2026 humanoid robot sales forecast remains unchanged, while the 2028 forecast is raised from approximately 41,600 units to 111,600 units.

· The RMB 995 price target corresponds to a 30x price-to-sales ratio on 2028 forecast revenue, setting high demands for accelerated commercialization and realization of platform value.

JPMorgan's stance on Unitree Robotics has seen a notable reversal.

In a report published on October 25, the bank upgraded Unitree Robotics' rating by two notches from "Underweight" to "Overweight," and raised its December 2027 price target from RMB 300 to RMB 995. The rating and price target adjustments have also been confirmed by public reports.

The core variable driving this adjustment is the change in the timeline for embodied intelligence breakthroughs and the resulting revaluation of hardware platform value.

Previously, JPMorgan remained cautious after conducting research at the China International Industry Fair: although domestic robot manufacturers had established early hardware advantages, demand mainly came from R&D, developer platforms, and small-batch pilots, and had not yet fully translated into repeatable commercial orders. Sufficient evidence was still lacking on whether enterprise customers would be willing to make sustained purchases and whether robots could generate measurable return on investment.

The Hangzhou World Physical AI Conference and Unitree Partner Ecosystem Conference changed the bank's assessment of the industry's pace of development. According to the report, Unitree founder Wang Xingxing shortened the expected timeline for physical AI breakthroughs from "up to ten years" to "two to three years," or even "one to two years."

JPMorgan accordingly believes that if the window for technological breakthroughs and application adoption arrives earlier, Unitree's existing robot installed base, developer ecosystem, and mass production capabilities will carry higher strategic value.

Breakthrough window moves forward, hardware platform value rises

Under a longer technology maturity cycle, early shipments from robot manufacturers mainly prove that products have certain market demand. But as the breakthrough window shortens, already-deployed robots may also become important infrastructure for model training, scenario testing, and application development.

According to management information cited in the report, Unitree's G1 has cumulatively rolled off the production line by approximately 10,000 to 20,000 units since its release in May 2024. This figure refers to cumulative production output, not equivalent to the scale of terminal commercial deployment.

What JPMorgan values is that this batch of hardware has the opportunity to let more developers become familiar with the Unitree platform and support data generation in real environments, control strategy validation, failure case collection, and product iteration. If robot capabilities show significant improvement within the next one to two years, hardware already deployed in universities, laboratories, and application development teams may help Unitree more quickly capture new demand.

Ecosystem building is also unfolding around this direction. The report states that this conference attracted more than 3,000 participants, with showcased scenarios covering urban governance, emergency firefighting, education and research, culture and entertainment, commercial services, and industrial applications. Products include quadruped robots, humanoid robots, and the Dex5-S five-fingered dexterous hand.

Unitree also proposed three initiatives: enhancing large-scale mass production capability, deploying physical AI small models adapted to hardware through in-house development and external cooperation, and building large-scale computing clusters. Its robot self-evolution system attempts to improve algorithm development efficiency through a cycle of "AI autonomous R&D—simulation training—real-world testing—feedback iteration."

Whether these investments can constitute a platform advantage depends on whether developer participation can continuously bring reusable software, real-task data, and commercial applications. Hardware scale provides a starting point, while data feedback and ecosystem monetization remain steps that need to be completed.

Advances in large models push competition into the real world

Another rationale behind JPMorgan's rating upgrade this time is the boost in demand for robot hardware driven by improved large model capabilities.

With progress in task decomposition, memory, context management, and cross-tool execution capabilities, AI's capabilities in high-level planning and supervision have strengthened. The problems robot companies need to solve are further concentrated on how to translate these capabilities into reliable, safe, and economically viable physical execution.

Digital systems can generate action plans, but robots still need to complete perception, movement, grasping, and contact control in real environments. These tasks depend on sensors, actuators, batteries, heat dissipation, calibration, safety design, durability, and on-site maintenance capabilities.

Therefore, enhanced large model capabilities may increase the value of mature robot bodies and deployment systems. Model companies, universities, and system integrators need real hardware to collect data, validate control strategies, and test model performance in complex environments.

Simulation and synthetic data can accelerate development, but they struggle to fully capture object variability, contact variations, human interference, and equipment wear and tear. Failure cases and feedback from real-world deployments still determine whether a system can continuously improve.

Under this framework, Unitree's potential value extends to serving as the interface for AI models entering the physical world: more hardware deployments bring more testing opportunities, test results support improvements to models and control systems, and enhanced capabilities in turn drive application expansion.

However, this loop still requires commercial evidence to support it. Whether installed base scale can generate high-quality data, whether data can effectively flow back, and whether Unitree can generate revenue from the ecosystem cannot be inferred from shipment volumes alone.

Valuation Anchor Shifts to 2028, Order Fulfillment Remains Key

JPMorgan's reassessment of the industry cadence is directly reflected in its sales forecasts.

The bank maintained its 2026 humanoid robot sales forecast for Unitree at approximately 14,200 units, raised its 2027 forecast from approximately 26,200 units to 31,700 units, and significantly upgraded its 2028 forecast from approximately 41,600 units to 111,600 units. The 2030 forecast was raised from approximately 74,100 units to 214,300 units.

According to its estimates, Unitree's humanoid robot sales in 2028 and 2030 would correspond to global market shares of approximately 21% and 12%, respectively.

This means the target price adjustment primarily reflects a re-rating of the medium-term growth trajectory rather than an equivalent change in current-year sales. JPMorgan expects Unitree's revenue for 2026 through 2028 to be approximately RMB 2.819 billion, RMB 4.888 billion, and RMB 13.383 billion, respectively; adjusted net profit is estimated at approximately RMB 616 million, RMB 1.125 billion, and RMB 3.202 billion, respectively. All figures are institutional forecasts.

The RMB 995 target price is based on 30x price-to-sales on 2028 forecast revenue, i.e., using future revenue scale as the valuation anchor in advance. This pricing framework requires the company to achieve significant volume ramp-up over the coming years while maintaining strong profitability.

Capital market supply is also a factor JPMorgan considered. The bank observed that IPO and spin-off listing progress for robotics companies has been slower than previously expected, with regulatory reviews placing greater emphasis on recurring order revenue and business model stability. This may prolong the scarcity premium for already-listed leaders, though it represents the bank's judgment on the financing and listing environment.

Scarcity of listed targets can support valuations, but companies still need to cross the commercialization threshold. Key risks listed by JPMorgan include: slower-than-expected enterprise adoption, with pilot projects failing to convert into repeat orders; intensifying competition compressing pricing and profit margins; and regulatory and geopolitical changes affecting overseas expansion.

This rating shift reflects JPMorgan's reassessment of the pace of embodied intelligence development. In the next phase, what will determine whether this judgment holds will be sustained customer procurement, operational reliability in real-world scenarios, and whether revenue growth can translate into stable profits.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps