Text/Yang Jianyong

Since OpenAI unveiled ChatGPT at the end of 2022, it has not only ushered artificial intelligence into the era of generative AI but also catapulted large models into the global technology spotlight. Consequently, OpenAI has emerged as the world's most prominent AI large model unicorn, boasting an astonishing valuation of $850 billion.

AI Giant Suffers Over $260 Billion Loss

However, the large model industry is both technology-intensive and capital-intensive. The cornerstone lies in constructing a vast AI computing power infrastructure, which is a key factor hindering AI large models from achieving short-term profitability.

It can be said that companies offering large model services have yet to turn a profit, including OpenAI, which is also grappling with financial challenges, with annual losses surpassing $30 billion.

According to reports, by 2025, OpenAI's revenue is projected to reach approximately $13 billion, but its losses are expected to soar to a staggering $38.5 billion (around 260 billion yuan). The primary culprit is the substantial investment in artificial intelligence infrastructure, including spending billions of dollars to procure high-performance chips from NVIDIA.

Data reveals that this AI giant's total expenditures in 2025 will amount to $34 billion. Among these, $19 billion will be earmarked for research and development, with a core focus on pouring massive funds into large model development and AI infrastructure construction, causing costs to skyrocket. It is estimated that computing power investments will reach $600 billion by 2030.

It's crucial to note that in the era of generative AI, global tech behemoths are embarking on a new round of AI computing power races, ramping up their investments in artificial intelligence. Goldman Sachs predicts that the AI spending of just four tech giants—Meta, Microsoft, Amazon, and Google—is expected to hit $5.3 trillion by the end of 2030.

The AI Race Drives NVIDIA's Profitability to New Heights

This global trend of expanding capital expenditures underscores that these cloud giants possess abundant cash and massive free cash flows to bolster their competitiveness in the artificial intelligence market. Consequently, vendors selling computing power services are reaping enormous profits, with annual profits reaching $120 billion. NVIDIA's Jensen Huang claims that the company will achieve cumulative revenues of $1 trillion from 2025 to 2027.

The global demand for NVIDIA's AI infrastructure is unprecedentedly robust, reflecting that the development of AI large models across various sectors cannot proceed without NVIDIA. This includes deploying at least 10 gigawatts of NVIDIA systems for OpenAI's next-generation artificial intelligence infrastructure. Collaborating with Google Cloud, Microsoft, Oracle, and xAI to construct AI infrastructure using hundreds of thousands of NVIDIA GPUs.

AI Large Models Command Astonishing Valuations, with OpenAI and Anthropic Eyeing Trillion-Dollar Market Caps

Overall, AI large models are heavily reliant on AI infrastructure and necessitate substantial computational resources for model training. Of course, large models are undoubtedly the hottest sector currently, deeply favored by the market. Despite massive losses, the market still exhibits a strong preference for them. OpenAI's market valuation is anticipated to exceed $1 trillion upon going public. Meanwhile, Anthropic's valuation is equally remarkable, reaching $900 billion in its latest $65 billion funding round.

Zhipu and MiniMax, as the most representative large model startups in China, have successively listed on the Hong Kong Stock Exchange. Fueled by capital, their combined market value exceeds 850 billion Hong Kong dollars. Zhipu's market value has even soared to 720 billion Hong Kong dollars, with a cumulative year-to-date increase of 1277%, achieving a 12-fold surge in market value, demonstrating the market's confidence in the commercialization potential of AI large models.

Finally, information technology is advancing at an astonishing pace, and with each technological transformation, dazzling companies emerge. Among them, OpenAI and Anthropic are undoubtedly the world's most prominent large model unicorns, and numerous large model companies have also surfaced in China.

With the rapid iteration of AI large model technology, the commercialization process of large models is exhibiting a growth trajectory, with AI becoming ubiquitous from the cloud to the edge. For individual C-end users, large models are profoundly reshaping the way information is accessed, serving as a new entry point for users to obtain information.

At the same time, large models are injecting vitality into innovation across various industries, sparking a new wave of industrial intelligence. Amidst this trend, the commercialization of large models is rapidly being implemented.

However, compared to their revenue scale, their market values have far outstripped the support provided by revenue and profitability. It is particularly crucial to view the long-term development potential of AI large models rationally.

Yang Jianyong, a contributor to Forbes China, expresses views that represent only his personal opinions. He is dedicated to providing in-depth interpretations of cutting-edge technologies such as AI large models, artificial intelligence, the Internet of Things, cloud computing, and intelligent hardware.