Text/Yang Jianyong

Since the launch of ChatGPT, artificial intelligence has transitioned into the era of generative AI. As a result, AI chips and large AI models have become the most sought-after sectors, with investors showing strong preference for companies operating in these fields. Among them, Cambrian and Zhipu AI were once highly sought-after companies, with their market valuations repeatedly reaching new highs.

From a 24-Fold Surge to a Trillion in Market Value Vanished: A Reassessment of Worth

Driven by the growing demand for AI chips for model training, along with multiple factors such as substitution dividends, Cambrian's performance surged exponentially, and its market value repeatedly hit new highs, at one point surpassing the 1 trillion mark, making it one of the biggest winners in China's AI computing chip sector.

In the large model sector, MiniMax and Zhipu are representative domestic large model companies and the only large model stocks listed on the Hong Kong stock market. Zhipu's stock price once peaked at 2,980 HKD, a 24-fold increase from its issuance price of 116.2 HKD, with a peak market value of 1.33 trillion HKD. MiniMax's peak market value also reached 417.6 billion HKD, a 7-fold increase.

The market enthusiasm for AI large model companies stems from the vast prospects of artificial intelligence entering the era of generative AI. Meanwhile, the valuation of companies that large model companies are benchmarked against is in the trillions of dollars. OpenAI and Anthropic, as global benchmarks for large models, have a combined valuation of 2 trillion USD, with Anthropic alone valued at that amount.

Unfortunately, Zhipu's current market value stands at only 300 billion HKD, a reduction of 1 trillion HKD from its historical peak, representing a 77% decline. MiniMax's market value is 84.7 billion HKD, with over 330 billion HKD in market value having evaporated, an astonishing 80% decline.

Although MiniMax faces pressure from share unlocking, its market performance is primarily due to the fact that its revenue scale and commercialization potential cannot support a trillion-dollar market value, leading to a reassessment of its worth and a market reevaluation of AI implementation and profit expectations.

It should be noted that large models are capital-intensive industries, and it is challenging to generate profits from large model services to achieve break-even, necessitating continuous financing to enhance AI large model capabilities.

Persistently High Computing Costs: Zhipu Raises 75.6 Billion HKD in Financing This Year

Competition in large models centers around AI infrastructure, requiring substantial computational resources for model training. Technology giants have increased their investments in artificial intelligence, aiming to enhance model competitiveness. This year alone, major global cloud giants like Microsoft and Amazon have increased their AI spending to over 810 billion USD.

OpenAI and Anthropic have also invested heavily in large model development and AI infrastructure construction. OpenAI's computing investment is expected to reach 600 billion USD by 2030, while Anthropic has secured 518 billion in computing expenditure contracts, including hundreds of billions in purchases from Amazon AWS and Google Cloud services.

It is worth mentioning that large models have evolved from initial knowledge question-and-answer to Agent AI, beginning to perform tasks and further increasing Token usage. Consequently, model companies still need to invest substantial funds to support this, making AI large models a veritable money-consuming industry from research and development to computing power.

For Zhipu, lacking the abundant cash and large free cash flow of cloud giants, it has chosen to raise funds through market placements and bond issuances to develop models and meet demand by investing in more computing resources.

This year, Zhipu has raised a total of 75.6 billion HKD in the market, including 4.9 billion HKD from its IPO in January 2026, 31.4 billion HKD from a placement in July, and 39.3 billion HKD from another placement and bond issuance in September. Approximately 60% of these funds will be used for the research and development of the next-generation GLM foundational model and a fully self-trained system, as well as for upgrading computing infrastructure.

Zhipu's Revenue Surges by 399%, Accelerating Commercialization of Large Models

With persistently rising computing costs and industry competition, it is exceptionally challenging for the entire industry to profit from large models. Zhipu is no exception, with multiple factors leading to a significant decline in gross margin, from 56.3% in 2024 to 26.38% in the first half of 2026; losses have also continued to widen. In the first half of 2026, adjusted losses reached 1.964 billion yuan, an increase of 12.1%.

Of course, with the rapid iteration of AI large model technology, the commercialization process is showing growth. Zhipu primarily relies on Model as a Service (MaaS) and commercialization platforms to provide general large model services to institutional clients and individual users. In the first half of 2026, the MaaS platform had over 7.4 million users, and the commercialization of AI large models is accelerating, with revenue also showing explosive growth, reaching 954 million yuan in the first half of the year, a year-on-year increase of 399.7%.

Among them, Zhipu's revenue from open platforms and API businesses reached 825 million yuan in the first half of the year, a 27-fold increase year-on-year, with the revenue share increasing from 15.2% in the same period last year to 86.5%, becoming a crucial business for Zhipu. Most importantly, the rapid growth in revenue scale confirms the enormous potential for the commercialization of AI large model technology.

Finally, artificial intelligence has developed for over 70 years, especially rapidly in the past decade, from neural networks to deep learning, and now to large models, propelling artificial intelligence to new heights.

The sweeping wave of large models has made AI large models ubiquitous, bringing about a new wave of industrial intelligence. The large model market has enormous potential, and the commercialization process of large models is showing growth. For large model companies, only by abandoning market hype, promoting the development and commercialization of large models through solid and steady efforts, and creating technological differentiation + commercial sustainability to achieve long-term value can they navigate through technological cycles.

Yang Jianyong, a contributor to Forbes China, expresses views that represent his own. He focuses on in-depth observations of technology sectors such as AI large models, artificial intelligence, industrial robots, IoT, cloud computing, and smart hardware.