Kai-Fu Lee argues Chinese AI growth is driven by profit and frugality
The AI pioneer explains how U.S. chip limits and a focus on monetization are shaping a new, resource-efficient Chinese AI landscape.
The short version
- Lee estimates Chinese open-weight models are now only six to 12 months behind top American models like those from OpenAI.
- Chinese firms are prioritizing profitability over the pursuit of artificial general intelligence (AGI).
- Market competition and venture funding, rather than just government policy, drive Chinese AI development.
- Chinese companies can operate models for as little as one-tenth the cost of their American competitors.
- A Stanford study shows U.S. companies still lead in top-tier models and high-impact patents.
Why it matters
This shift suggests that the AI landscape may bifurcate into high-cost, high-performance American models and lower-cost, highly efficient Chinese alternatives. For businesses, this could mean more affordable AI options as Chinese firms optimize for cost-effectiveness.
The closing gap in AI
U.S. export restrictions on advanced chips were intended to maintain a technological lead over China. However, Kai-Fu Lee, a former president of Google China, suggests these limits have actually forced Chinese developers to innovate in different ways. Instead of relying on massive hardware hauls, they are building competitive models with fewer resources.
Lee estimates that Chinese open-weight models—AI systems where the underlying weights are shared publicly—are now just six to 12 months behind major American models. This includes models from companies like OpenAI and Anthropic. The rapid progress has shifted the sentiment among Chinese developers who previously felt they were years behind.
The emergence of DeepSeek in early 2025 served as a turning point for this sentiment. By releasing a highly capable model trained with significantly less computing power, the startup proved that Chinese developers could compete with the leaders. This success has boosted national pride and provided a roadmap for other firms to follow.
Profit over AGI goals
A common misconception in the United States is that Chinese AI companies are primarily focused on winning the race to achieve artificial general intelligence (AGI). Lee argues that this view ignores the fundamental nature of Chinese business. He states that Chinese companies are born to make money and that profit is their primary purpose.
Lee compares the current wave of Chinese AI developers to established tech giants like Pinduoduo, Meituan, and ByteDance. These companies were initially questioned regarding their profitability before they became some of the largest entities in China. He suggests that while the U.S. may lead in breakthroughs, the ability of Chinese firms to find a way to monetize technology is well-established.
This profit motive creates a different set of priorities for Chinese developers. While American companies may spend heavily to attract the world's best researchers and provide them with massive computing power, Chinese firms are often more concerned with survival. They aim to create products that are attractive to businesses that do not want to pay premium prices for AI services.
Market forces and funding
Many observers assume that China's embrace of open-source technology is a coordinated government effort to compete with the West. Lee argues that market competition is the true driver of this trend. He points out that Chinese companies are venture-funded just like their American counterparts and are primarily motivated to serve their investors.
While the Chinese government does play a role, the influence is multifaceted. According to a Pitchbook analyst note cited by Fortune, state-linked investors were involved in over 140 Chinese AI transactions during 2025. Additionally, President Xi Jinping has called for the development of an independent and controllable AI software and hardware system.
Despite these influences, the underlying mechanics of the industry remain tied to commercial success. Chinese startups are navigating a landscape where they must prove their models can generate revenue. This creates a dynamic where market needs dictate the pace of development as much as, if not more than, state policy.
The efficiency advantage
Lee identifies a third misconception regarding why Chinese companies have shifted toward models that require less computing power. He argues that they are naturally parsimonious rather than just reacting to U.S. export controls. This frugality is a result of different priorities and a strong survival instinct that emphasizes prudence over the pursuit of AGI.
Due to their ability to be resourceful, Chinese firms can run models at a price point that is as little as one-tenth the cost of their American rivals. This cost advantage makes their products highly attractive to a wide range of commercial users. It allows them to offer high-quality AI services at a fraction of the price of the American alternatives.
However, the U.S. still maintains a lead in certain high-level metrics. Research from Stanford University indicates that American firms are still producing a higher volume of high-impact patents and top-tier models. Furthermore, OpenAI and Anthropic remain the world's most valuable AI companies, with Anthropic seeking a valuation of up to $1.4 trillion in a potential IPO.
What to watch
- Whether the cost-efficiency of Chinese models leads to a wider adoption of cheaper AI services globally.
- The outcome of Anthropic's potential IPO and its impact on the valuation of Chinese competitors.
- How U.S. companies will react to the rise of lower-cost, high-performing open-weight models.
Try it
Look up the DeepSeek model to see how it compares to American models in terms of training efficiency and cost.
Sources and links
- OfficialStanford University study hai.stanford.edu
- ReportingFortune: the article we sourced
First reported by Fortune. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.