China’s AI Models Thrill Global Users and Investors
Recently, China’s affordable artificial intelligence models have been gaining traction worldwide, driving a surge in stock market activity related to these technologies.
As the focus of AI development expands to encompass a variety of tasks, a new economy is taking shape around “tokens,” which are the fundamental data units used in large-scale language models. China’s edge stems from its plentiful, inexpensive electricity and a multitude of AI model developers competing in the space, leading to an increase in global token usage.
Niche startups are drawing investment attention away from established tech giants, as the potential for growth in the AI sector seems promising. Companies like MiniMax Group Inc. and Knowledge Atlas Technology Joint Stock Co. (Zhipu) just went public and have already surpassed $40 billion in market value, eclipsing more entrenched players like Baidu and Kuaishou.
Larger firms like Alibaba have experienced significant stock gains despite facing operational challenges. Investors like Victoria Mio from Janus Henderson see this as a pivotal shift, suggesting that the current rally is reflective of a deeper market revaluation rather than just temporary trading excitement.
“You can really see how consumers are engaging with tokens and how AI is becoming something that can be monetized,” Mio noted. It seems investors are optimistic that China’s cost advantages will secure the country’s position as a long-term leader in this field. The Chinese government is also actively promoting this trend to enhance its tech industry’s global competitiveness, making “token exports” a hot topic in the media.
The Rise of Tokens
A token represents a small amount of data, roughly equating to four characters. While many AI firms offer limited free or fixed monthly plans, users with greater needs, like app developers, often find themselves paying per token used—sometimes up to ten times more for output tokens compared to input tokens.
Tokens have really become essential in the AI economy. They facilitate transactions and are also utilized by employers to gauge developer productivity, showcasing their significance in various contexts.
Recently reported data from OpenRouter indicates that all three top models for token consumption this month are Chinese, including Xiaomi’s MiMo, Alibaba’s Qwen, and DeepSeek. MiniMax, a smaller player, has two models in the top ten, and StepFun has one as well.
“China is really in a position to compete globally when it comes to the provision of tokens and models,” said Bush Chew from Aberdeen Standard Investment. “Building a data center in China costs significantly less than in the U.S.”
Interestingly, MiniMax is now fourth in overall market share according to OpenRouter, trailing only major American firms like Alphabet and OpenAI. Their most sought-after models charge around $1 per million output tokens, which is a stark contrast to Google’s $3 or over $15 for some of Anthropic’s models.
In the stock market, Zhipu surged by 16% in Hong Kong recently, while MiniMax climbed 7.1%. Moreover, excitement has been generated around DeepSeek’s plans to secure at least $300 million in funding.
Goldman Sachs regards MiniMax as a strong contender among Chinese AI companies, equipped with solid multimodal products and competitive pricing. Both MiniMax and Zhipu are part of what’s known as the “Six Tigers,” a group of startups that provide more specialized AI services, often with backing from larger corporations like Alibaba.
Both firms went public in January, with Zhipu’s shares soaring around 750% since then and MiniMax up 440%. In contrast, Alibaba has seen its stock drop by 3.6% this year, and Tencent is down 12%.
The Market Landscape
Amidst a flurry of investments in new startups, several companies like StepFun and Moonshot AI are preparing to enter the market. However, the rapid rise in stock prices of unprofitable newcomers has raised concerns for some investors.
“Valuations in some areas have increased too quickly compared to actual earnings growth,” noted Nicholas Chui from Franklin Templeton. He emphasized that how well these companies model AI solutions will be pivotal going forward.
The market is highly competitive, with both new and established players vying for attention. Tencent, which has vast potential due to its WeChat platform, has not made a significant mark in the recent AI advancements.
On a related note, affordability remains crucial for users. The success of OpenClaw, which automates various functions in daily life, has intensified the demand for competitively priced tokens from smaller models. There’s also a trend towards “token maximization,” where engineers demonstrate their productivity by maximizing token usage.
While the U.S. retains a leading position technologically, China’s budget-friendly AI applications are meeting a substantial part of the global demand. Recent state media reports show that China’s daily token consumption jumped to over 140 trillion in March, a significant increase from 100 billion at the start of 2024.
“For investors, it’s important to evaluate China’s AI landscape based on how effectively resources are utilized rather than just relying on the headlines,” suggested Janus Henderson’s Mio. “China is enhancing AI services at a low cost and quickly integrating them into real-world operations.”
