Li Auto Announces Embodied Intelligence Strategy, Maintains Buy Rating

Overview

This report analyzes Li Auto’s Livis Day event held on June 15, 2026, focusing on comprehensive upgrades across software, autonomous driving, and embodied intelligence. Deutsche Bank believes that through proprietary models, chips, and operating systems, Li Auto is redefining vehicles as AI agents entering the physical world, while maintaining a Buy rating on its stock.

Core views

Li Auto has established an embodied intelligence vision centered on safety, capability, and efficiency. On the model front, the company introduced two new models—Mach Mind-Pro and Mach Mind-Edge—where the former focuses on voice, memory, and agent capabilities, and the latter serves as the industry’s first mass-produced edge-device-native embodied intelligence agent, specializing in visual processing and real-time decision-making. Concurrently, alongside the newly launched L9, the Mach VLA model was unveiled, achieving a 40% reduction in end-to-end latency through full-stack reconstruction. On the hardware and system level, Li Auto released the M100 chip, built on a 5nm automotive-grade process, delivering 1,280 TOPS of computing power per chip—several times more powerful than the Thor U chip when running self-developed intelligent driving models. Coupled with Halo OS, designed specifically for embodied intelligence, and a fully controlled chassis, the system response speed surpasses human reaction by 40%. The company has also constructed an end-to-end defense framework, ensuring that chips, compilers, OS, and algorithms operate on a unified trusted foundation. In terms of product iteration planning, Li Auto has outlined a clear OTA roadmap: 30% improvement in intelligent driving efficiency by July; by September, the system will leverage imitation learning to directly handle complex road conditions (such as pothole avoidance and full-scenario automatic parking), reducing task completion time by 30%; and by December, further refinements will address false-touch prevention, route memorization, and end-to-end latency. The company aims to bring its intelligent driving model performance up to the level of Tesla’s FSD V14 by Q4.

Analysis framework

The institution primarily evaluates Li Auto’s competitive barriers in the field of embodied intelligence by dissecting its technical architecture (models, chips, OS) and iteration cadence (OTA roadmap). The analytical logic emphasizes horizontal comparisons of technical parameters (e.g., against Tesla’s FSD or Nvidia’s Thor chip) as well as the tangible improvements in user experience and safety resulting from technology implementation, thereby assessing long-term product competitiveness and valuation support.

Methodology notes

• Competitive and Strategic FrameworkMoat / competitive advantage

  Technical synergy achieved through full-stack self-development

  The report underscores how Li Auto’s vertically integrated control over chips, OS, and algorithms enables deep software-hardware collaboration (e.g., latency optimization, response speed enhancement), forming a core competitive advantage distinct from automakers reliant on external suppliers.

• Industry/Industrial Analysis Framework

  Embodied Intelligence (AI Agent) Implementation Paradigm in the Automotive Industry

  The report introduces the concept of embodied intelligence, analyzing how automobiles are evolving from mere transportation tools into physical-world AI agents capable of understanding intent and autonomously executing tasks—a new dimension for evaluating automakers’ future technological attributes.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Li Auto (2015.HK)

  Beneficiary, leader in embodied intelligence technologies

  Strengths

  Full-stack self-development capabilities (chips, OS, models), clear OTA iteration path, rapidly approaching industry benchmarks in intelligent driving

  Comparison

  Compared to automakers dependent on external suppliers, Li Auto boasts stronger software-hardware synergy and cost-control potential; its intelligent driving goals directly target Tesla’s FSD

  Risks

  Technology implementation falling short of expectations, intensifying market competition

Key data

• Target Price96.00 HKD12-month target price set by Deutsche Bank
• Current Stock Price56.85 HKDClosing price as of June 15, 2026
• End-to-End Latency Optimization40%Percentage reduction in latency achieved through full-stack reconstruction
• M100 Chip Computing Power1,280 TOPSComputing power per single chip; dual-chip configuration delivers 2,560 TOPS
• Intelligent Driving Efficiency Improvement30%Expected efficiency gain following the July OTA upgrade
• Reinforcement Learning Data Growth15xMultiple-fold increase in training data for the Mach VLA model

Impact & implications

The report posits that by implementing embodied intelligence technologies, Li Auto not only enhances product safety and efficiency but also positions itself as a continuously evolving AI agent, thereby strengthening user loyalty and boosting brand premium. If its intelligent driving capabilities meet expectations and match Tesla’s FSD V14 by Q4, it will further solidify its leading position in the high-end NEV market, supporting upward pressure on the stock price.

What to watch

• Actual outcomes and user feedback from the July, September, and December OTA upgrades
• Q4 comparative testing results between Li Auto’s intelligent driving model and Tesla’s FSD V14
• Progress and stability of M100 chip and Halo OS integration into production vehicles