At the AI summit of the 25th China Venture Capital & Private Equity Annual Forum, hosted by Zero2IPO and PEdaily recently, NIO Capital Partner Yao Li joined a roundtable titled “The Offense and Defense of AI Investing,” where he shared his insights on the evolving AI industry and the distinct characteristics shaping the landscapes in China and the United States.

 

Li emphasized that AI is a long-cycle track. For investment targets, he focuses on three dimensions: technology leverage, model boundaries, and the data flywheel. Reflecting on these dimensions will also help startups better define their products and businesses.

 

Key Insights from Yao Li:

 

 

2025: A Year of "Return" for the AI Market

 

Li noted that 2025 is a year of high market enthusiasm, with “too many projects to review.” From his observation of the industry—particularly the AI sector—he summarized the trend with the keyword “return,” which carries several layers of meaning:

 

First, he pointed to a return to engineering and technology fundamentals. He noted that while discussions around large language models in recent years heavily emphasized scale—such as token processing capacity—the focus in 2025 is shifting back toward foundational research.

 

Second, he highlighted a return to the application scenario. Many AI companies, including those at the application layer, are now re-anchoring themselves in specific verticals and use cases. They are concentrating on how to leverage AI technology to address concrete problems within those scenarios and refine their products accordingly.

 

Third, Li observed a return to business fundamentals. Beginning in 2025, a growing number of enterprises, including startups, are rigorously evaluating their unit economics. From an investment standpoint, he emphasized the importance of the metric "Dollar per Intelligence"—which assesses the value derived relative to the cost of computational power consumed.

 

Fourth, he described 2025 as marking a return of investment value for autonomous driving. NIO Capital began its AI investment approximately seven to eight years ago, during a period of significant excitement around autonomous driving. This year, portfolio company Pony.ai achieved dual listings in the US and Hong Kong markets, while another key investment, Momenta, maintains a leading industry position. Overall, Li assessed 2025 as a dynamic year.

 

 

Cost-effectiveness: China's Advantage Becomes Apparent

 

Li outlined that the differences in AI development between China and the US are twofold. Firstly, he noted China's strength in hardware, supported by a richer ecosystem spanning consumer electronics, smart devices, and robotics. He referenced the intense competition seen in recent years—from the “hundred-model war” in large language models to what he described as a current “hundred-robot war” in the robotics industry.

 

Secondly, he emphasized the dimension of cost-effectiveness. While the US companies often pursue a “brute force” approach—scaling computing power and model size—Li pointed out that Chinese enterprises have strategically focused on enhancing the cost-efficiency of its models, building on existing foundations. He observed that from foundational models to applications and hardware, cost-effectiveness remains a relatively prominent advantage for Chinese companies. This is evidenced, he noted, by the fact that many US humanoid robotics companies source hardware from Shenzhen due to China's competitive edge in speed, efficiency, and performance.

 

 

Illustrating this with the Robotaxi sector, Li compared the US industry leader Waymo, which operates a fleet of over 2,000 vehicles, with China' s Pony.ai—a NIO Capital portfolio company that hit its target of 1,000 vehicles ahead of schedule this year and is poised for multi-fold growth next year, bringing it to a comparable scale. More importantly, he highlighted that Pony.ai achieves positive unit economics even with lower fares, whereas many US companies, in his view, still face a longer path to profitability.

 

 

Long-cycle Track: Focusing on Three Dimensions

 

Li stressed that AI is a long-cycle track measured in decades, citing autonomous driving’s decade-long journey to early commercialization. On the timeline for achieving AGI (Artificial General Intelligence) or ASI (Artificial Superintelligence), he regarded it as an open scientific question, noting it is difficult to judge whether the current Transformer-based Scaling Law will suffice. He suggested two areas for potential breakthrough.

 

First, the evolution of the technical paradigm—specifically whether architectures like Transformer still offer room for significant improvement. Second, the role of data: current language models are built primarily on compressions of existing and past data, including visual information. He further noted the recent emergence of synthetic data and new approaches based on reinforcement learning. Given these developments, he concluded that reaching AGI or ASI will likely depend on novel and substantial breakthroughs in these domains.

 

From an investment and entrepreneurship standpoint, Li shared that NIO Capital concentrates on three key dimensions: technology leverage, model boundaries, and the data flywheel. Technology leverage examines whether a product can achieve exponential gains as foundational models advance. Model boundaries assess whether an application-layer consumer-facing startup risks being absorbed by future foundational models. Reflecting on these dimensions can help startups better define their products and businesses, while NIO Capital will also continue to seek promising investment targets based on these three criteria.

 

Li concluded by reflecting that AI has pressed the fast-forward button on everything, creating a widespread sense that time is passing unusually quickly. This acceleration has led many in R&D—and even some investors—to worry about potential unemployment. Yet for a field measured in decades, he noted, the year-on-year changes may in hindsight appear less abrupt. He expressed hope that all stakeholders can seize the opportunities of this era to contribute meaningfully to the industry and society.