Text/Yang Jianyong

Large AI models stand as one of the most sought-after sectors today, deeply embraced by the market. Among them, Zhipu has witnessed an astonishing cumulative surge of 1,974% since its market debut, with its valuation soaring to HK$1.07 trillion.

AI large model companies are currently in high demand in the market, primarily because artificial intelligence has transitioned into the era of large models, presenting boundless opportunities. Meanwhile, global leaders like OpenAI and Anthropic boast valuations nearing a trillion dollars. As a rare Chinese contender in the large model arena, Zhipu has become a hotly pursued target.

It's worth noting that artificial intelligence has evolved over more than 70 years, with rapid advancements in the past decade, from neural networks to deep learning, and now to large models, propelling AI to unprecedented heights. Various industries are actively harnessing the transformative power of AI large models to drive intelligent upgrades.

In particular, large models are being continuously optimized for efficiency, enabling the deployment of compact models in AI toys, AI glasses, wearable devices, and various end-user terminals such as smart vehicles.

Simultaneously, the valuations of companies targeted by large model firms are approaching the trillion-dollar mark.

Since OpenAI introduced ChatGPT, it has ushered in the era of generative AI, becoming the world's foremost AI large model startup with a valuation of $850 billion. Anthropic's valuation is equally impressive, with its latest $65 billion funding round catapulting its valuation to $900 billion.

In the market, this benchmarking has had a profound impact, spurring market enthusiasm. Zhipu, as China's most prominent large model startup, has surpassed the trillion-dollar market value milestone.

However, amidst the awe at capital's might, concerns about a speculative bubble have also surfaced. After all, the current market value far exceeds what can be justified by revenue scale and profitability.

In terms of revenue scale, Zhipu's revenue in 2025 was merely RMB 724 million, yet it incurred losses amounting to RMB 4.718 billion, up from RMB 2.958 billion in the same period last year, marking a 59.5% increase in losses.

Crucially, large models belong to a technology and capital-intensive industry, heavily reliant on AI infrastructure. Profiting from large models is exceptionally challenging, even for OpenAI, which also struggles to break even. Data reveals that in 2025, OpenAI's revenue was approximately $13 billion, with losses reaching a staggering $38.5 billion.

Despite the prominence of AI large models, generating profits remains elusive due to the substantial computational resources invested in AI model training. In 2025, Zhipu's R&D expenditure was RMB 3.18 billion, a 44.9% year-on-year increase, with 70% allocated to computing power.

Consequently, Zhipu's gross margin declined from 56.3% in 2024 to 41% in 2025. This was attributed to the investment in additional computing power resources to meet demand, resulting in an expansion of losses last year.

Faced with the rapid evolution of Agentic AI and the subsequent surge in Token consumption, computing power is undergoing significant transformation. Zhipu asserts that its focus lies not on profitability but on supporting the exponentially rising curve of high-quality Token consumption.

Finally, with the swift iteration of AI large model technology, the commercialization process of large models is showing signs of growth. Nevertheless, it remains imperative to view the long-term development potential of AI large models with a rational perspective.

For large model vendors, it is crucial to foster the robust development and commercialization of large models, thereby positively influencing performance, with the anticipation of achieving profitability in the future.

Yang Jianyong, a contributor to Forbes China, expresses views that are solely his own. He is committed to providing in-depth interpretations of cutting-edge technologies such as AI large models, artificial intelligence, the Internet of Things, cloud computing, and smart home appliances.