Stock Soared 20 Times... But Profitability Remains Uncertain [Weekend Money]

Zhipu’s GLM 5.2 Ranks Fourth Globally in Performance
Zhipu’s Market Cap Soars 1,958% Compared to IPO Price
Profitability Only at 0.5% of Anthropic’s Level

Chinese artificial intelligence (AI) models are rapidly improving in performance and catching up with leading U.S. companies, but there remains a significant gap when it comes to monetization, according to recent analysis. While model capabilities and user adoption are advancing quickly, it will take more time for Chinese firms to match their American counterparts in key commercialization areas such as a global enterprise client base, recurring revenue, and robust cloud partnerships.

Most recently, research analyst Choi Seolhwa of Meritz Securities noted, “Silicon Valley in the U.S. has maintained its dominance for decades in terms of revenue scale, market capitalization, and sustainable business cycles. Chinese AI companies now need to look beyond the domestic market nurtured by government support and build a verified, paying global customer base on their own.”

Chinese AI Models on the Rise... Narrowing the Performance and Market Share Gap

Last month, Chinese AI startup Zhipu released its new model GLM 5.2, which ranked fourth globally in performance. This placed it right after the top U.S. models—Anthropic's “Fable 5,” “Opus 4.8,” and “GPT 5.5”—demonstrating a rapid closing of the performance gap with leading American models.

In particular, the model set off what some are calling a “second DeepSeek moment” because its pricing per 1 million output tokens is only $4.4, about 15% that of GPT 5.5’s $30. As a result, Zhipu, which is listed on the Hong Kong Stock Exchange, saw its market capitalization surpass 1 trillion Hong Kong dollars (approximately 189 trillion won), with its stock price soaring 1,958% compared to the IPO price.

The momentum of Chinese AI models is also apparent in data from OpenRouter, which tracks models used by developers. Since February of this year, token usage for Chinese AI models has rapidly increased, and in terms of API call volume, their market share has climbed from 23% at the end of last year to 34% currently.

Choi explained, “This shows that Chinese AI models are quickly dominating large-scale data processing and background task segments with overwhelming cost-effectiveness and an open-weight strategy.”

“Monetization Gap Remains... China’s State Capital Model Faces Limitations”

However, significant limitations remain in terms of profitability. While Anthropic recorded annual recurring revenue (ARR) of $47 billion and a valuation of $965 billion (about 1,430 trillion won), Zhipu’s ARR stands at just $250 million—only 0.5% of Anthropic’s level. When using ARR as a proxy for annual revenue and calculating the price-to-sales ratio (PSR) based on company valuation, Anthropic stands at 20.5x, while Zhipu’s PSR is a staggering 467.6x. Although revenue is growing, increased spending on research and development has pushed Zhipu’s net loss last year to $691.8 million.

Choi commented, “Model performance only determines the supply ceiling—whether that supply can be sustained as long-term profit is a separate matter. When discussing the substitutability of a model, sustainable profitability is just as crucial as technical performance. From a profitability standpoint, it appears Chinese models still need time to close the gap with the U.S.”

The structural advantage of heavy state funding, which enabled such rapid growth, is also seen as a long-term limitation. Zhipu generates 42% of its revenue from the Chinese government and public institutions—a strong foundation for early growth, but its ability to stand alone in global commercial markets remains uncertain. Furthermore, governments in the U.S., Taiwan, and other Western countries have recently banned the use of Chinese AI models over security concerns, which is also a major headwind.

Choi emphasized, “The government and public (B2G) sector may provide stable demand, but by its nature, it is fundamentally different from the global commercial (B2B) market. Unconditional optimism about Chinese AI models should be avoided.”

She added, “The current valuation of Zhipu is largely based on future expectations. For both related stocks and potential beneficiaries within the open ecosystem, validating the sustainability of this valuation will be a core risk factor going forward.”

This content was produced with the assistance of AI translation services.