Li Auto理想汽车

About

Li Auto was founded in 2015 by Li Xiang and is headquartered in Beijing, dually listed on NASDAQ (LI) and HKEX (2015.HK). Li Xiang explicitly declared 'we are not an EV company or a smart-device company, but an embodied intelligence company,' defining the car as 'the biggest robot.' In January 2026 the company completed a major organizational restructuring, establishing a humanoid robot team (codename Nexus), a software-body team and a foundation-model team; at GTC 2026 in March it released the MindVLA-o1 foundation model with a native multimodal MoE architecture; in February it announced the 'industry's first embodied-intelligence car' — the new-generation Li Auto L9 Livis (priced at RMB 559,800, with 2 in-house Mach M100 chips and 2,560 TOPS of compute); a two-wheeled robot is planned for release within 2026, and the LIVIS AI glasses went on sale in December 2025. It is the world's second automaker after Tesla to develop its own foundation model; 2026 R&D spending is expected at RMB 12B (about 50% AI-related) with a cash reserve of about RMB 98.9B. But the core auto business is under pressure: a Q1 2026 net loss of RMB 2.276B with Q2 losses expected above RMB 2B; by mid-2026 at least 8 Li Auto executives had left to start embodied AI companies, forming a 'Li Auto school' ecosystem.

Technical Approach

Full-stack in-house foundation models (MindGPT/MindVLA-o1) + chips (Mach M100) + operating system (Xinghuan OS) + embodied bodies (two-wheeled/bipedal robots) — no more than 3 companies worldwide can cover all of these simultaneously; transfers autonomous-driving perception, decision-making and planning capabilities to robots, with over 50% overlap between the LiDAR/camera/motor/battery hardware supply chain and robotics; the Mach M100 chip is 'one chip, many uses' — a common compute platform for cars and robots.

Commercialization（7）

• ◆Announced a major reorganization at the January 2026 all-hands meeting, establishing the humanoid robot team (codename Nexus), software-body team and foundation-model team; Zhan Kun took over the foundation-model business
• ◆The new-generation L9 was announced in February 2026 as an embodied-intelligence flagship SUV; MindVLA-o1 was released at GTC 2026 on March 17
• ◆Held the Livis Day launch event on June 15, 2026, where Li Xiang proposed 'the first half is the car, the second half is embodied intelligence' and announced Q4 assisted-driving alignment with Tesla FSD V14
• ◆Within 2026, plans to publicly release the two-wheeled robot (for factory scenarios) and deliver the AI system built around the Mach M100 chip
• ◆Technology reuse: smart-driving perception (3D BEV, OCC, 3D ViT) and decision-planning capabilities transfer to robots, with over 50% supply-chain overlap and massive road-scenario data transferable to robot environment-understanding models
• ◆Financials: Q1 2026 net loss of RMB 2.276B, with Q2 losses expected above RMB 2B (revenue about RMB 24.1B, down 20% YoY); 2026 R&D spending expected at RMB 12B, about 50% AI-related; cash reserves of about RMB 98.9B
• ◆'Li Auto school' ecosystem: by mid-2026 at least 8 executives had left to build embodied AI startups, and Li Auto invests directly (e.g., in 斜跃智能) to build an embodied AI ecosystem

Position & Strengths（5）

• ◆The most thorough strategic pivot among carmakers: Li Xiang explicitly declared 'not an EV or smart-device company, but an embodied intelligence company,' defining the car as 'the biggest robot'
• ◆Full-stack in-house: foundation models (MindGPT/MindVLA-o1) + chip (Mach M100) + OS (Xinghuan OS) + embodied bodies (two-wheeled/bipedal robots) — no more than 3 companies worldwide can do all of these
• ◆Mass-production data moat: the VLA driver model has 80% monthly usage, 12.254 million cumulative command uses, and 250 million km of assisted driving over Spring Festival
• ◆RMB 100B-level cash reserves: about RMB 98.9B of cash supports heavy R&D investment
• ◆'Li Auto school' ecosystem network: invests in startups by former employees such as 斜跃智能, building out embodied AI