Qunhe Technology rose 4 times in two days and fell 2.54% yesterday—this is not an AI bottom-fishing opportunity, but a trap of buying at a high price.

Core Conclusion: The collective decline in Hong Kong stock AI newcomers yesterday is not a trend reversal for most stocks, but a profit-taking after excessive previous gains—the real danger is not the drop, but their current valuation levels. Key data: Qunhe Technology (00068.HK) IPO price HK$7.62, cumulative gain of 391% over two days of listing (PS ~60-70x); Zhipu (02513.HK) cumulative gain of nearly 6x, PS ~400x; MiniMax cumulative gain over 475%. Southbound capital net inflow exceeded HK$220 billion in Q1, with over HK$90 billion flowing into the information technology sector. Why it matters: The Hong Kong stock AI sector has followed a chip-driven trajectory of listing, surging, rising further, and being chased from January to April this year. Now, for all declining AI stocks, one must first see how much they rose before and then judge whether to act.

First, clarify the facts—what exactly happened to Qunhe yesterday
Qunhe Technology closed at HK$18.60 on its first listing day, April 17, up 144%; on April 20, it surged from HK$19.46 to HK$37.44, a single-day gain of 101.29%; yesterday, April 21, it surged another 30% to HK$43.62 at market open, then retreated from highs and closed down 2.54%.
That is to say—yesterday's 2.54% drop was a 2.54% fall from a position that had risen 4x.
If you only read hot news headlines saying Hong Kong stock AI concept stocks fell, Qunhe down 2.54%, you see a declining stock; but open the K-line chart, and you see a stock that has pulled back after rising 4x in three days.
Misunderstanding these two things is the difference between losing principal.

Where is the valuation of Hong Kong stock AI now
Qunhe Technology's 2025 revenue was RMB 820 million, adjusted net profit RMB 57.12 million (first-time turnaround), intraday market cap HK$63.7 billion, PS ~60-70x. Traditional software in the same Hong Kong stock sector, such as Kingdee International and Mingyuan Cloud, typically have PS ratios of 2-5x.
Zhipu (02513.HK) broke through HK$1,000 intraday on April 13, cumulative gain of nearly 6x from IPO price, market cap over HK$300 billion, corresponding to 2025 revenue of RMB 724 million, PS ~400x. MiniMax-W reached a high of HK$1,330 on March 18, up over 475% in three months of listing. Tianshu Zhixin hit a new high of HK$372.60 on April 13.
Southbound capital net inflow exceeded HK$220 billion in Q1, with over HK$90 billion flowing into the information technology sector. Hong Kong stock 2026Q1 IPO fundraising totaled HK$109.9 billion, nearly 5x year-on-year, topping the global ranking.
Lots of money, few stocks, concentrated new share supply—this is why a company with 2025 revenue of RMB 820 million could be pushed to a market cap of HK$63.7 billion in two days. This is not driven by fundamentals, but by chips.

In 2015, I stepped into the exact same pit with a stock in a similar position
In 2015, I chased a VR concept stock in A-shares. I didn't chase it when it rose from 20 to 150; when it fell to 80, I thought it was cheap enough, and bought a batch.
It then fell from 80 to 30 before stopping.
What I learned then: A correction in a high-valuation stock is never a buying point; it's the first pressure valve of a bubble starting to release. The cheapness you see is cheap relative to its own previous high; not cheap relative to fundamentals.
Qunhe is not exactly the same as that VR stock—Qunhe is a good company with real customers, real technology, and a first-time turnaround. But they share one thing in common: the valuation has already fully priced in the story several years ahead.

Hong Kong stock AI now needs to be viewed in at least four tiers
Tier 1·Performance Delivery Type: Kingsoft Office's main business grew 25%, ASMPT rose over 7% after earnings yesterday, Q1 profit +190% YoY, Changfei Fiber's optical module capacity sold out across the industry. The ups and downs of this tier are supported by fundamentals, and PE/PS ratios are calculable.
Tier 2·Internet Giants: Tencent, Alibaba, JD.com, Meituan. Real cash flow, historically low valuations. This tier is for defensive positioning.
Tier 3·Scarce AI Leaders: Model companies like Zhipu and MiniMax. Big story, true scarcity, but valuations at PS 400x have already priced in expectations up to 2027-2028.
Tier 4·New Stock Speculation: Qunhe Technology, Shenghong Technology listed yesterday (up 57% on first day, market cap HK$315.4 billion), A-share Shenghe Jingwei (up 289% on first day). Essentially chip speculation rather than AI logic.

What you should do now
If you see Hong Kong stock AI falling today and want to bottom-fish, please do one thing first—open this stock's K-line chart for the past month. If its gain in the past month exceeds 50%, or its gain in the past week exceeds 100%, today's drop is not a buying point, it's profit-taking. In this case, at least wait for it to fall back near the starting point of the rise a month ago before considering, otherwise you're not fishing the bottom, you're fishing the mountainside.

If you hold AI newcomers like Qunhe, Zhipu, MiniMax, Tianshu Zhixin with PS ratios above 60x—my suggestion is not to increase holdings. If the position is already heavy, consider taking profits in batches during the next surge. Trigger conditions: if it surges back to above 95% of the previous high, or if it surges on high volume in a single day and then retreats to close with a negative candle, reduce by 1/3 first. This is not bearishness, it's acknowledging that the money you've made has far exceeded the valuation your research can justify.

On April 14th, in the article "Kingsoft Office Q1 Surges 400%", I gave a judgment: Hong Kong stock tech is splitting, the emergence of AI application layer performance is a dislocation opportunity, and I put the Kingsoft system, Kingdee, and Weimob into the watchlist.
One week later, a correction—that judgment was half right. The AI application layer is indeed splitting out companies delivering performance (Kingsoft Office main business 25%, ASMPT profit +190%). But the wrong half is: the AI application layer label still includes newcomers like Qunhe, Zhipu, and MiniMax that have risen 5x in three months of listing, which cannot be equated with old companies like Kingsoft that have been around for over a decade.
Today, adjust the watchlist: retain Kingsoft Software, Kingdee, Weimob, only watch Q2 main business data. Remove Qunhe, Zhipu, MiniMax, Tianshu Zhixin from the watchlist—not because they are bad, but because the valuation is no longer something I can calculate.

My positions
Hong Kong stock core holdings Tencent + Alibaba, from the first day of establishment in September last year to the 30th trading day today, not a single share moved. Last night, the three major US stock indices closed down 0.6%; today, the Hang Seng Index fell 1.16%, I still won't move.
I hold no AI new stocks or AI leaders. Not planning to add today, nor next week.
This is not conservatism. This is a rule learned from losses on a VR concept stock 11 years ago.

$MANYCORE TECH(00068.HK) $KNOWLEDGE ATLAS(02513.HK) $MINIMAX-W(00100.HK)