Summary of Key Points
Li Kaifu's AI company, ZeroOneAllThings (formerly one of the "Six Big Models"), has recently abandoned the development of ultra-large-scale foundational models and shifted to providing enterprise AI solutions. The company plans to list in Hong Kong by 2027. This transformation signifies a change in its valuation logic from focusing on technical potential to emphasizing its ability to deliver tangible business results. It needs to prove that it is more than just an AI outsourcing firm and must overcome challenges such as scaling, converting orders into actual revenue, and managing risks in overseas markets.
1. Why abandon the "high-end" big models and focus on enterprise AI?
In simple terms, the big model space is highly competitive, making it difficult for startups to compete. Big models are extremely costly to train, requiring billions of dollars in computing power and manpower. Companies like Alibaba, Tencent, and Baidu have significant resources at their disposal, giving ZeroOneAllThings a tough time competing on a technical level. Enterprise AI, on the other hand, is more immediately profitable. By developing customized AI solutions for businesses (e.g., intelligent customer service for banks or production optimization for factories), ZeroOneAllThings can quickly generate orders and revenue without the continuous investment required by big model projects.
2. The shift in valuation logic: from "storytelling" to "real money"
Previously, ZeroOneAllThings was valued based on the strength of its models. Now, as an AI service company, its value is determined by its ability to generate actual revenue. There are three main valuation frameworks for AI companies in the capital market:
- Foundation Model Companies (e.g., Zhipu, MiniMax): These companies have not yet turned a profit but are valued based on their technical potential, often at 10–50 times their expected future annual revenue (or even up to 100 times in a hot market).
- AI Software Companies (e.g., ServiceNow): Their value is based on stable annual subscription fees, customer renewal rates, and gross margins. Mature companies can be valued at 5–20 times their future revenue, with additional consideration for profit margins.
- AI Outsourcing Companies (e.g., general IT consulting): These have the lowest valuations, typically at 10–25 times current profits or 0.8–5 times annual revenue.
ZeroOneAllThings aims to align itself with the second category and avoid the third. To achieve this, it needs to develop a scalable software platform that can be reused for multiple clients, rather than developing custom solutions for each one.
3. Emulating Palantir requires overcoming scalability challenges
ZeroOneAllThings aspires to follow the example of Palantir, a US-based company that serves governments and businesses with AI solutions worth over $300 billion. Palantir's success lies in creating a scalable software platform that can quickly adapt to different client needs, reducing development costs. However, ZeroOneAllThings is still in the transition phase, using open-source models (e.g., DeepSeek, TongyiQianwen) and custom data to create customized solutions. If it has to reinvent its approach for each client, it will remain an outsourcing firm with lower valuations.
4. Orders do not equal revenue: capital markets look for real capabilities
Although ZeroOneAllThings has received orders worth 1.5 billion yuan, this does not equate to actual revenue. Revenue generation involves multiple steps: contract signing, development, delivery, client acceptance, revenue confirmation, and payment collection. Any issue (e.g., customer dissatisfaction or delays) can result in missed payments. Capital markets focus on the quality of revenue (whether it comes from stable subscriptions), gross margins, and cash flow.
ZeroOneAllThings aims to achieve quarterly profitability by 2027 and must show that its orders can generate sustainable revenue without requiring additional recruitment of engineers for each new project.
5. The overseas market: a potential advantage with risks
Half of ZeroOneAllThings' business comes from overseas markets (Asia, Europe, South America), which can boost its valuation due to higher subscription fees and more stable revenue streams. However, there are risks:
- Data compliance: Different countries have varying data privacy regulations (e.g., GDPR in the EU), which can lead to fines if not followed properly.
- Localization: Local business practices vary significantly between regions (e.g., differences between European and Chinese companies).
- Geopolitics: Political tensions (e.g., Sino-US relations) can affect business operations.
The overseas market is both an opportunity and a challenge. It can increase valuation if managed effectively but may also hinder growth if not handled carefully.
Conclusion
ZeroOneAllThings' transformation focuses on shifting from a technology-driven to a business-driven approach. To achieve a high valuation upon listing, the company must address three key issues: avoiding the role of an AI outsourcing firm, achieving scalability, and converting orders into stable revenue. While the overseas market offers opportunities, it also presents significant challenges. This transition reflects a broader trend in the AI industry, where companies that can provide tangible business value are more likely to be recognized by investors.