Summary of Key Points
This article reveals four new trends in current technology investment by analyzing the list of young investors at the WAIC (World Artificial Intelligence Conference): capital is shifting from pursuing fancy technical terms to focusing on the practical application of technologies. Young investors are concentrating on areas such as AI's integration into the physical world, the ongoing benefits of large models, data bottlenecks leading to the development of new underlying technologies, and the need for patient capital in the challenging fields of hard technology. They place more emphasis on the combination of technology and industry rather than short-term hype around conceptual ideas.
Detailed Explanation
1. AI Moving from a “Toy on the Screen” to a “Tool in Reality”
In the past, our perception of AI was limited to chatbots (like ChatGPT) and the generation of images/videos. However, now AI is moving beyond the screen to become practical tools in real-world applications—such as robots for cargo handling, delivery services, autonomous driving vehicles, and even maintenance robots in space.
For example, embodied intelligence (robots that can perceive their environment and perform actions like humans) is very popular, but investors point out that the “bubble” lies not in the technology itself, but in people’s expectations. A robot that can simply hold a glass of water during a demonstration may be useful if it can perform the task 100 times without error and at a reduced cost in a real factory setting. Many robots are still being sold to universities for research purposes, but they are far from becoming profitable commercial products.
Another trend is “edge intelligence”—AI systems that are not only hosted in the cloud but also built directly into devices like smartphones, glasses, and robots. This approach eliminates the need for constant internet connectivity, improves response times, and allows for revenue generation through ongoing services (for example, regular maintenance of the robots). The number of autonomous delivery vehicles is expected to increase from 100 to 2,000 by 2025, operating in over 170 cities, with continuous testing of efficiency and cost-effectiveness—a true sign of practical application.
2. Large Models Are Not a “Fad”; Building a Self-Improving Cycle
Many believe the hype around large models has subsided, but their benefits are still not fully realized. Similar to how applications like Taobao and WeChat emerged years after the introduction of TCP/IP protocols, many companies have yet to effectively incorporate large models into their operations.
The so-called “intelligent flywheel” is a crucial concept: users use the product → generate data → the data improves the model → the improved model attracts more users → which in turn generates even more data, creating a self-sustaining cycle. Companies like Zhipu AI have seen their revenue increase fivefold in three years and have become leaders in the large-model space; Kuaishou’s Keling AI has started generating revenue after just 45 days of being integrated into its creative processes, earning 1 billion yuan by 2025. These companies’ success relies not on the model itself but on the “user-data-model” cycle, turning the model into a system that improves with use.
3. Insufficient Data? Scientific Baseline Models Fill the Gap
Large models require large amounts of high-quality data, but the internet’s text and image resources are becoming scarce, and manual annotation is expensive. A solution could be “self-learning” methods like AlphaGo Zero, which improved by playing against itself without human guidance. Another approach is to develop scientific baseline models—universal systems that can be applied across various fields such as life sciences and materials science. Although scientific data is specialized and difficult to obtain, this could be the key to the next wave of AI growth.
4. Hard Technology Requires “Patient Investment”; Speed Is Not Everything
Fields like commercial aerospace and quantum computing have long verification cycles. The future path for quantum computing is still uncertain (with both photonic and neutral atom approaches having their advantages and disadvantages), and commercial aerospace faces challenges such as launching, approval processes, and supply chain management. However, this does not mean these technologies are worthless. In 2025, private companies launched rockets 23 times and raised 18.6 billion yuan in funding, indicating rapid industry development.
Investors’ logic is to invest before the technology is fully mature (or by then, the opportunity may have passed), but they also avoid treating distant possibilities as immediate returns. They focus on whether the team has the necessary skills (e.g., core research and development capabilities) and whether there is a real market demand (e.g., the need for cleaning up space debris with more satellites). For example, the emerging field of space embodied robots is driven by the increasing number of satellites and the resulting need for in-orbit maintenance.
Conclusion
Today’s technology investment is no longer about who has the most appealing story; it’s about who can turn technology into profitable businesses. Young investors understand technology better and are willing to wait for the right opportunities, which may represent the new logic of the future tech industry.