On August 19, 2026, Unitree Robotics rang the opening bell on Shanghai’s STAR Market (ticker: 688836) — and within minutes, its shares had rocketed more than 629%, pushing the company’s market cap past RMB 444.9 billion (roughly USD 62 billion). It is now officially China’s first publicly listed humanoid robot company.
For anyone who has walked through a Shenzhen tech showroom on one of our tours, the name Unitree needs little introduction — its robot dogs and, more recently, humanoid platforms have become fixtures of the exhibition circuit we bring our clients through. What’s newsworthy today is less the product and more the market’s verdict on it: a nearly seven-fold opening pop is an extraordinary signal for embodied intelligence as an investable, commercializable category.
Unitree’s investor roster reads like a who’s-who of Chinese venture capital — Meituan, Hillhouse (Sequoia China), Matrix Partners China, and Photosynthesis Capital (光合创投) among them. Photosynthesis first backed the company in mid-2024 and has continued adding to its position in subsequent rounds. Notably, this marks the fund’s second landmark exit in quick succession, following earlier bets on MetaX (沐曦) and Zhipu AI (智谱) — and it was also an early investor in Innolight (中际旭创), the optical-module giant that recently completed a dual A+H listing.
Photosynthesis partner Zhu Jia described the listing as “a new chapter in China’s embodied-intelligence era,” crediting Unitree’s fully self-developed hardware stack — from quadruped robots to full-scale humanoid platforms — for building what he called a genuine moat through engineering execution and supply-chain control.
Why hardware, and why now
According to Zhu, the fund’s conviction traces back to 2020, when Unitree was still primarily known for four-legged robots. The real inflection point came in 2024, when the company’s first humanoid — capable of backflips and rapid bipedal walking — shipped, followed within six months by a more mature second-generation model. That shift, he noted, opened up an application space “an order of magnitude” larger than anything quadruped robots could address.
Around the same time, Photosynthesis observed something telling in its due diligence across China’s leading embodied-AI labs: regardless of which foundation model a company was racing to build, nearly all of them were standardizing on Unitree hardware for their physical robot platforms. The logic is straightforward — while embodied models are still iterating rapidly, the underlying robot body needs to be the reliable constant: durable enough for round-the-clock training, without derailing R&D timelines through frequent part failures, and priced sensibly enough to deploy at scale.
Zhu drew a direct parallel to NVIDIA’s role in the large language model boom: just as OpenAI and Anthropic’s GPU purchases fueled NVIDIA’s first wave of explosive growth, he expects the same dynamic to repeat for Unitree as the “physical infrastructure” of embodied intelligence — a hardware layer that early on captures foundational value, and over time becomes indispensable across a widening range of applications.
A deeper bet on embodied intelligence
Beyond Unitree itself, Photosynthesis has been building out a broader position across the embodied-intelligence value chain — from Zhiyuan Robotics (自变量机器人) on the model side, to component makers in joints and dexterous hands including Quanzhibo, Tianji, and Sharpa Robotics (whose tactile-AI dexterous hand was recently featured in NVIDIA and Unitree’s jointly designed H2Plus humanoid reference platform), down to a global precision-bearing manufacturer supplying an often-overlooked but hard-to-replace component.
The fund’s definition of “embodied intelligence” also extends well beyond humanoids on the ground: it has led funding rounds for aerial-robotics startup Wefly Intelligence (微分智飞), whose drones now run VLA (vision-language-action) models for fully autonomous obstacle avoidance, and deep-sea robotics company Deep Sea Smart Human (深海智人), which is applying embodied AI to subsea cable-laying, oil and gas development, and mineral exploration.
Zhu’s read on the broader trend: embodied AI is shifting from pure data-scale accumulation toward models that learn physical-world rules the way a toddler does — through observation and trial and error — and this shift is expected to diversify robot form factors dramatically. He points to a telling industry signal: several major embodied-robotics makers have announced production-scale-up plans this year, targeting roughly 10x shipment growth.
[Reference — this post is written based on an original article by IPO早知道 (WeChat ID: ipozaozhidao)]
For inquiries related to AI/robotics industry visits, corporate training programs, and on-site interpretation in Shenzhen, please contact us.