In-Depth Analysis of Investment Prospects in the Embodied Intelligence Track: Commercialization of Generative AI in Robotics and Investment Opportunities

Based on the data I have collected, I now provide you with a systematic and comprehensive

.

---

---

Embodied Intelligence refers to a technical form that integrates generative AI into physical entities such as robots, endowing them with capabilities like perception, learning, and interaction with the surrounding environment. Unlike traditional AI which “thinks in computers”, embodied intelligence emphasizes deep integration between AI and the physical world [1]. Humanoid robots are currently the most representative carrier of embodied intelligence.

According to the latest report released by Frost & Sullivan, the global embodied intelligence robot market size is projected to surge from $11.71 billion in 2024 to $101.07 billion in 2030, with a compound annual growth rate (CAGR) of 43.2% from 2024 to 2030 [2]. This growth rate far outpaces most technology sub-tracks.

2025 has seen unprecedented fervor in financing for the embodied intelligence field, with the following characteristics:

• Tashi Zhihang (它石智航):
  Secured over $120 million in two angel rounds within months of its establishment [3]
• Leju Robotics (乐聚机器人):
  Raised RMB 1.5 billion in a Pre-IPO round [3]
• Yinhe General (银河通用):
  Secured a single $300 million financing round, setting an industry record, with a valuation exceeding RMB 20 billion [3]
• Figure AI (US):
  Valuation skyrocketed from $2.6 billion to $39 billion in its Series C round [3]
• Physical Intelligence (US):
  Valuation jumped from $2.4 billion to $5.6 billion [3]
• Skild AI:
  Is in talks with SoftBank and NVIDIA for over $1 billion in financing, with an expected valuation of $14 billion [3]

Data from IT Juzi shows that as of December 21, 2025, 8 major internet giants including Baidu Ventures, Lenovo Capital, Ant Group, iFlytek Ventures, Meituan Longzhu, JD Cloud, and Alibaba have made a total of 62 investments in the embodied intelligence track, with investment amounts ranging between RMB 1.45 billion and RMB 3.4 billion [1].

As of November 2025, nearly 30 companies in the robotics industry chain have submitted listing applications to the Hong Kong Stock Exchange, covering the entire industrial chain from core components, whole-machine manufacturing to system integration [3].

Enterprise	Listing Status	Capital Background
Geek+ (极智嘉)	Listed on HKEX	Sequoia China, Hillhouse Capital, Tencent Investment
CloudMinds (云迹科技)	Listed on HKEX	Leading comprehensive VC/PE firms
Unitree Robotics (宇树科技)	In counseling period	Matrix Partners China, CICC Capital
Leju Robotics (乐聚机器人)	In progress	Local industrial funds
Ziyuan Robotics (智元机器人)	Acquiring a listing platform	Huawei-affiliated capital

---

The integration of generative AI and robotics is achieving a qualitative leap:

Breakthroughs in

enable robots to understand complex instructions and plan multi-step task sequences. For example, Dyna Robotics’ robots can fold 700 towels in 24 hours with a 99.4% success rate [4].

• Natural Language Interaction:
  Robots can conduct natural conversations with humans, receive task instructions, and provide real-time feedback on execution progress [2]
• Cross-Modal Perception and Reasoning:
  Process multi-modal inputs such as text and images, understand complex instructions, and independently plan tasks [2]
• Leap in Cognitive Intelligence:
  Significant enhancements in environmental understanding, autonomous decision-making, and collaborative interaction capabilities [2]

• Precision assembly and flexible logistics in automotive production lines [2]
• Warehouse sorting, material handling in manufacturing [4]
• Deployment cases in hotel service scenarios [5]

• Commodity sorting, meal delivery, and cleaning maintenance in retail, catering, and building management [2]
• Popularization in scientific research and education markets (teaching tools, competition platforms, children’s education) [6]

• Healthcare: Rehabilitation care, remote diagnosis assistance, surgical assistance [2]
• Extension to agricultural scenarios [2]

Wang Hao, CTO of Independent Variable Robotics, believes that current embodied intelligence is at the “GPT-2 level” and is expected to reach the GPT-3 level in 1-2 years [4]. This means the technical path is clear, but the true “GPT moment” is still some way off.

• Insufficient Data “Fuel”:
  Embodied large models lack ready-made data, with datasets being the main barrier [1]
• Limited Generalization Capability:
  Significant gaps remain compared to skilled workers in workstations requiring high flexibility and judgment [3]
• Cost-Efficiency Issues:
  Humanoid robots are still costly and inefficient, requiring strategic investment [3]

---

is the world’s first generative universal embodied brain company, having completed an additional angel round of financing exceeding RMB 10 million, with the investor being Xuhui Capital [7]. The company is stationed in Xuhui District’s “Model Speed Space”, a core carrier of the artificial intelligence industry.

Moushen Intelligence is positioned as a

developer, which differentiates it from most enterprises in the embodied intelligence track that focus on robot ontology hardware R&D. Its technical route focuses on:

• R&D of the robot “brain” layer, rather than the “cerebellum” (motion control) [1]
• Deep integration of multi-modal large models and robot control
• Building of versatility and generalization capabilities

According to public information, the funds from this financing round will be mainly used for:

• Computing Power Procurement:
  Supporting large model training and inference
• Expansion of Engineering Team:
  Accelerating the commercialization of technology
• Construction of Robot Experiment Platform:
  Building data closed-loop infrastructure

• Track Scarcity: The world’s first company positioned as a “Generative General Embodied Brain” developer
• Policy Support: Stationed in Shanghai Xuhui District’s “Model Speed Space”, enjoying regional industrial policy dividends
• Technical Differentiation: Avoiding the red ocean of hardware competition and entering the high-barrier “brain” layer

• Technology not yet fully validated, business model to be tested by the market
• Facing competitive pressure from international giants (such as Physical Intelligence, Skild AI)
• Low maturity of embodied intelligence “brain” technology, long R&D cycle [1]

---

Time Node	Market Size	CAGR
2024	$11.71 billion	-
2030	$101.07 billion	43.2%
2035 (Hybrid Robots)	Nearly $200 billion	25.2% (2030-2035)

Data Source: Frost & Sullivan Analysis [2]

Analysts at Morgan Stanley predict in their latest research report:

• 2050:
  The global humanoid robot market size will reach
  $5 trillion
• Number of Robots:
  May exceed 1 billion, 90% of which will be used in industrial and commercial fields

1. Maturity of Large Model Technology:
   The overall capabilities of large models are becoming stable, already serving as a reliable basic capability layer for embodied intelligence systems [4]
2. Decline in Computing Power Costs:
   The unit cost of equivalent computing power is declining over the long term, allowing startups to afford training costs [4]
3. Maturity of Hardware Supply Chain:
   Technologies of key components such as motors and reducers continue to mature, with costs decreasing steadily [4]

• Standardization by MIIT:
  On December 26, 2025, the Standardization Technical Committee for Humanoid Robots and Embodied Intelligence under the Ministry of Industry and Information Technology was formally established [1]
• Local Policy Support:
  Shanghai has set up a RMB 1 billion special fund for embodied intelligence, while Beijing Yizhuang has gathered over 300 ecological enterprises [3]

• Rising labor costs and population aging
• Demand for intelligent transformation of manufacturing
• Demand for cost reduction and efficiency improvement in the service industry

---

• Joint Modules and Reducers:
  Unitree Robotics has cut prices to $5,900, driving a steep decline in the cost curve [4]
• 6-Axis Force Sensors:
  Core components of robot “dexterous hands”
• Tactile Sensors:
  Key link in human-robot interaction

• VLA/VLA Model Developers:
  Integrated vision-language-action solutions
• Multi-Modal Large Model Platforms:
  Competition centered on versatility and generalization capabilities
• Data Collection and Simulation Platforms:
  Solving the data “fuel” bottleneck

• Industrial Scenario Whole Machines:
  Automotive production lines, warehouse logistics
• Commercial Service Robots:
  Hotels, retail, catering
• Specialized Applications:
  Healthcare, agriculture, construction

• Computing Power Infrastructure:
  Platform-level players such as NVIDIA and Google [4]
• Embodied R&D Paradigms:
  Such as the “ACE Embodied R&D Paradigm” by Daxiao Robotics [8]

• Overinflated Valuations:
  Some enterprises have not yet achieved revenue or released specific products, yet their valuations can still climb to billions of yuan [3]
• Order Authenticity:
  Multi-hundred-million-yuan orders are mostly framework agreements, with actual delivery amounts accounting for only 10%-20% of the total agreement value [3]
• Related-Party Order Phenomenon:
  Similar to the “related-party order” issue that once emerged in the new energy vehicle industry [1]

• Easy to Go from “0 to 1”, Hard to Scale from “1 to N”:
  Large-scale deployment faces multiple constraints [3]
• Data Bottleneck:
  High cost of real-scenario data collection, gap between simulated data and real data [6]
• Reliability Issues:
  The operational efficiency of robots is generally still lower than that of skilled workers [3]

• Long Commercial Validation Cycle:
  Large-scale commercial application may still take 2-3 years [4]
• Intensified Competition:
  There are already over 150 humanoid robot companies in China, more than half of which are startups or “cross-industry” entrants [9]
• Narrow Exit Channels:
  Most embodied intelligence enterprises have only been established for 1-3 years, making traditional exit methods such as IPO and mergers and acquisitions difficult [3]

Phase	Focus Areas	Evaluation Indicators
Early Stage	Technical routes, team background	Paper publications, team resumes
Middle Stage	Scenario deployment capability	Order amount, customer quality
Mature Stage	Mass production delivery capability	Shipment volume, ROI, profit margin

Current investment logic has diverged: Companies focusing on the “cerebellum” (motion control) are now required to demonstrate mass production capabilities and orders, while companies focusing on the “brain” can still tolerate early-stage losses and focus on technological breakthroughs [3]

---

[1]. According to a report by China Insights Consultancy, as downstream application scenarios are gradually unlocked, especially in retail services, commercial exhibitions, and industrial manufacturing, conditions are forming for products to shift from single-point trials to batch replication.

• ROI data from real scenarios
• Matching degree between shipment volume and delivery capability
• Conversion rate of framework agreements to actual deliveries

• Whether the founder has a dual background in AI + robotics
• Whether the core technical team comes from top laboratories or leading enterprises
• Whether the team has a “long-distance running” mindset [9]

• Whether it has data collection and processing capabilities
• Generalization capability of VLA/VLA models
• Gap with international leading technical routes

• Whether it avoids direct competition with giants (e.g., Unitree Robotics in hardware, Moushen Intelligence in the brain)
• Whether downstream scenarios are clear and scalable
• Whether it has benchmark customers or orders

Strategy	Recommended Allocation	Risk Preference
Core Allocation	Leading validated enterprises (e.g., Unitree Robotics, Yinhe General)	Low
Growth Allocation	Embodied “brain” technology companies (e.g., Moushen Intelligence)	Medium
Frontier Allocation	Core component suppliers	Medium-High
Wait-and-See	Concept-only companies with no actual products	High

1. Severe Short-Term Volatility:
   Two extremes coexist, such as K-Scale Labs failing to secure financing and closing down, while Figure AI is flush with funding [4]
2. Valuation Correction Risk:
   Current valuations more reflect expectations rather than performance, carrying bubble risks
3. Uncertainty in Technical Routes:
   The industry has not yet formed unified technical standards, carrying risks of route betting

---

1. The track is on the eve of a boom:
   2025-2026 is a critical period for embodied intelligence to shift from “pioneer showcase” to “rational advancement”. The technical path is clear, and conditions for commercialization are taking shape.
2. Generative AI is the core driver:
   Large model technology enables robots to have a complete capability chain of “understanding - planning - execution”, which is the fundamental reason for the boom of embodied intelligence.
3. Investors need to distinguish between “brain” and “cerebellum” tracks:
   Companies focusing on the “brain” (e.g., Moushen Intelligence) can still tolerate early-stage losses and focus on technological breakthroughs; companies focusing on the “cerebellum” need to demonstrate mass production capabilities and orders.
4. Beware of bubbles and grasp certainty:
   Current overinflated valuations and “artificial hype” around orders coexist. Investment should return to the essence of business, focusing on real-scenario ROI and delivery capabilities.

As the world’s first generative universal embodied brain company, Moushen Intelligence has the following investment value:

• Track Scarcity:
  Enters the “brain” layer of embodied intelligence, avoiding the red ocean of hardware competition
• Policy Dividends:
  Stationed in Shanghai’s “Model Speed Space”, enjoying regional industrial support
• Technical Differentiation:
  Forms a complementary position with peers focusing on the “cerebellum”

Technology has not yet been fully validated, facing competition from international giants, and requires preparation for a “long-distance race”. It is recommended to position this as a in the embodied intelligence track, with the allocation ratio controlled within 5-10% of the investment portfolio.

---

[1] Science and Technology Innovation Board Daily - 2025 for Humanoid Robots: Value Verification, Capital Frenzy, and a Crossroads (https://www.eet-china.com/mp/a463650.html)
[2] Frost & Sullivan - 2025 Global and China Embodied Intelligence Robot White Paper (https://www.frostchina.com/content/insight/detail/696616c4945e6056fa4e07a1)
[3] 36Kr - 2025 Year-End Special: A Year of Frenzy in Embodied Intelligence: Robots Still Can’t Do Real Work (https://m.36kr.com/p/3610836423574531)
[4] 36Kr - 2025 Annual Review of the Embodied Intelligence Industry: From Pioneer Showcase to Rational Advancement (https://eu.36kr.com/zh/p/3615952634507783)
[5] Sina Finance - Dialogue| After the “Listing Rush”, Is Embodied Intelligence Still Worth Investing In? (https://finance.sina.com.cn/roll/2025-12-23/doc-inhcuunk0700069.shtml)
[6] Huxiu - 2026 Embodied Intelligence Outlook and Warning (https://www.huxiu.com/article/4825538.html)
[7] Woshipm - Financing & Acquisition| Moushen Intelligence Completes Angel Round of Financing Worth Tens of Millions of Yuan (https://www.woshipm.com/news)
[8] Jiemian News - Daxiao Robotics Releases ACE Embodied R&D Paradigm and Enlightenment World Model 3.0 (https://www.jiemian.com/lists/48_46.html)
[9] People’s Daily - Four Questions on Artificial Intelligence: Will Embodied Intelligence Reach a Turning Point? (http://finance.people.com.cn/n1/2026/0103/c1004-40637792.html)

---

January 14, 2026
Jinling AI Financial Research Team