# Huawei Invests in Physics-Based AI: Extreme Vision
Huawei has made another strategic move in the **world model** arena — investing in a physics-based AI startup.
## Extreme Vision Secures Massive A1 Funding
**Extreme Vision** has completed a new A1 funding round worth **hundreds of millions of RMB**, jointly invested by **Huawei Hubble** and **Huakong Fund**.
This is the **third consecutive funding** round for the company in just two months.
---
## Company Overview
- **Founded:** 2023
- **Core Focus:** China’s first *“pure-blood” physics AI* company, dedicated to **world model development**.
- **Product Coverage:** Full-stack solutions (hardware + software) for:
- Automotive world models
- Embodied foundation models
- World model platforms
- **Core Application Areas:** Autonomous driving & embodied intelligence — sectors Huawei is actively investing in.
**Team Background:** Members from elite institutions:
- Tsinghua University
- Chinese Academy of Sciences (CAS)
- University of Science and Technology of China
---
## Key Milestones
### 1. Public Recognition (2024)
At **Qiji Chuangtan 2024**, in collaboration with **Tsinghua University’s Department of Automation**, Extreme Vision launched **YiSu**:
- **China’s first** native video generation model supporting ultra-long **16-second sequences**.
- Emphasized *physical-world understanding* in video generation.
- Demonstrated potential in autonomous driving & embodied intelligence.
---
### 2. Leadership
**Huang Guan — Founder & CEO**
- PhD in AI, Tsinghua University, Department of Automation
- Previous roles at **Microsoft**, **Samsung**, **Horizon Robotics**
- Serial entrepreneur in AI & autonomous driving
> In a 2024 speech, Huang stated:
> “All paths toward general intelligence converge on *end-to-end large models*, making every problem a high-quality data challenge. World models are the most critical source of such data for embodied intelligence.”
---
### 3. DriveDreamer Development
#### DriveDreamer (2023)
- **World’s first** *real-world-driven autonomous driving world model*.
- Attracted significant attention in industry and academia.
#### DriveDreamer 4D (2024)
Collaboration partners:
- CAS Institute of Automation
- Li Auto
- Peking University
- Technical University of Munich
**Significance:**
- **First** application of world models to enhance **4D driving scene reconstruction**.
- Provides **data generation** & **closed-loop simulation** for autonomous driving and embodied intelligence.
- Multiple fixed-cooperation agreements with leading OEMs; serving dozens of clients.
---
## Core Scientific Team
- **Zhu Zheng — Co-founder & Chief Scientist**
- PhD (CAS Institute of Automation, 2019)
- Postdoctoral research at Tsinghua University
- Google Scholar: 17,000+ citations, h-index: 50
- Key contributions:
- *SiamRPN*, *DaSiamRPN*: Influential deep learning object tracking algorithms
- *BEVDet*: Adopted by automotive companies for 3D perception
- **Sun Shaoyan — Co-founder**
- Former Director at Alibaba Cloud
- Former GM, Horizon Robotics closed-loop data product line
- **Mao Jiming — Partner & VP of Engineering**
- Former Architect at Baidu & Yingche
- Led Baidu Apollo’s simulation technology team
---
## Shared Vision with Huawei
Extreme Vision’s mission: **Build a data engine for autonomous driving & embodied intelligence using world models.**
Huawei aligns closely with this vision:
- Prefers **WA (World Action)** models over **VLA (Vision-Language-Action)**.
- Advocates **direct mapping** from perception (vision) to decision-making, bypassing language.
**Huawei Intelligent Automotive CEO Jin Yuzhi:**
> “We will not follow the VLA path. While it may seem like a shortcut, it is not the way to truly achieve autonomous driving.
> We value WA, removing the Language step, directly controlling the vehicle via perception data.”
---
## Related AI Content Platforms: AiToEarn
For AI-powered content creators & researchers:
- [AiToEarn官网](https://aitoearn.ai/) integrates **content generation**, **publishing**, and **monetization**.
- Multi-platform publishing:
- Douyin, Kwai, WeChat, Bilibili, Xiaohongshu, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, X (Twitter)
- Analytics & AI model rankings: [AI模型排名](https://rank.aitoearn.ai)
---
## Huawei’s WEWA Concept
**WEWA** = *Cloud-based World Engine* + *Vehicle-based World Behavior Model*
Research collaborations extend world model applicability beyond smart vehicles.
---
## Academic Advancements
**Example Project:** **WorldGrow**
- Partners: Shanghai Jiao Tong University, Huazhong University of Science and Technology
- Capability: Generate a **272 m² indoor scene** in **30 minutes** on a single GPU
- Highly realistic and coherent textures
---
## Strategic Implications
Huawei Hubble’s backing of **Extreme Vision**:
- **First** investment in a company with “world model” as its core business.
- Solidifies Huawei’s direction in **spatial intelligence** and **embodied intelligence**.
- Complements prior investment in **Qianxun Intelligence**.
---
### The Bigger Picture
World models aim to **unify perception and action for machines**.
Similarly, platforms like AiToEarn unify **creation and distribution** for human–AI collaboration — merging technical innovation with monetization.
---
**— End —**
