Huaibei Mining: Director and General Manager Qiu Dan Resigns Due to Job Transfer
Why are Zhipu AI and MiniMax in such a hurry to return to the A-share market? This question reveals the most pressing anxieties faced by China's AI startups. While their Silicon Valley counterparts are discussing how to implement large language models for profit, China’s leading players are struggling to determine the right venue and timing for their IPOs—a situation that is, in itself, thought-provoking.
Analysis
Why are Zhipu AI and MiniMax in such a hurry to return to the A-share market? This question reveals the most pressing anxieties faced by China's AI startups. While their Silicon Valley counterparts are discussing how to implement large models for profit, China’s leading players are struggling to determine the right venue and timing for their IPOs—a situation that is, in itself, thought-provoking.
News of MiniMax planning to list on the STAR Market, combined with recent rumors of capital operations by Zhipu AI, serves as a striking footnote in the current AI landscape. Their choice is quite clear: amid tightening overseas listing channels and a retreating tide of US dollar funds, the A-share market—especially the STAR Market—has become one of the few places where a “hard tech” narrative can still be told. But the question remains: can large models still be considered “hard tech” today? Even OpenAI is struggling to recoup costs through ChatGPT Plus subscriptions. So where exactly does the confidence of our companies come from when rushing into the capital market with just a concept?
Look at another trending story: Silicon Valley giants have begun limiting employees’ token usage. This is practically dark humor. On one side, we are still burning money on computing power and celebrating sky-high valuations; on the other, they are already fine-tuning operational costs and starting to “use sparingly.” This contrast signals that the industry is entering a deeper phase—the era of storytelling based solely on parameter scale and API calls is over. Now, you have to prove you can make money sustainably and efficiently. And the A-share market, especially retail investors, often lacks patience for such complex tech-business logic, preferring straightforward narratives like “technological breakthroughs” or “domestic substitution.” Companies choosing to return to the A-share market are, to some extent, opting for an audience that is both easier to understand and more easily swayed.
The news about the viral rise of “one-person companies” is quite interesting. Some are earning millions a year, while others have seen their income shrink by 90%. Isn’t this the inevitable outcome after the widespread adoption of AI tools? The democratization of production tools never leads to universal wealth, but rather a more extreme Matthew effect. Top talent uses AI to multiply efficiency; most others find that lowered barriers only intensify competition. AI hasn’t eliminated differences in work—it has likely amplified the variance in individual capabilities. The claims that “AI will replace everyone’s jobs” can take a rest; AI is more about redefining the nature of “work” itself, pushing some to the center of the stage while pushing others further to the margins.
Then there’s Microsoft’s “3.3k stars in a week” self-evolving skill project. A giant like Microsoft researching “how to train skills like training neural networks” is laying the groundwork for the next platform paradigm. When AI evolves from “generating content” to “executing actions” and “completing tasks,” the real productivity revolution will have begun. Meanwhile, some of our companies may still be preoccupied with the conversational fluency of their next version. The gap in strategic vision is fully exposed in the emphasis placed on foundational versus applied research.
Ultimately, the rush of Zhipu AI and MiniMax to return to the A-share market is, on the surface, a capital choice, but at its core, it is a survival strategy. Pressured from multiple fronts—being strangled at the frontier of technology, facing arduous commercialization overseas, and dealing with price wars and homogenized competition domestically—quickly accessing the capital market for ammunition has become the most realistic option. But this is a stimulant, not a cure. Will the funds raised from going public be used to burn through the next quarter’s computing bills, or will they be truly invested in foundational research and differentiated implementations that can build long-term competitive moats? This is the key to determining whether they become the next “true tech giants” or just another “valuation bubble” in the capital market.
The restoration of access to Tianya Community feels poetically out of place in this context. A relic of an old-era BBS being reopened at the peak of the AI wave is like a delayed digital archaeological dig. It reminds us that technology is always sprinting forward, but the human desire for communication, expression, and the need to mark the passage of time remains an unchanging backdrop. AI can generate everything, but it cannot generate a “sense of history” or “community memory.”
In the final analysis, the choices made in the capital market are a mirror reflecting the collective subconscious of the industry—its fears and desires. The rush to go public stems from fear of missing the window; choosing the A-share market comes from having no other options. The bill for this AI feast is being transferred from global USD funds to domestic retail investors. Going forward, it remains to be seen whether these highly anticipated companies can use real technological prowess to forge a moat, or if they will repeatedly explain during earnings calls why their gross profit margins remain so low.
Disclaimer: The above content is generated by AI and is for reference only.