Firmly holding onto the AI belief for ten years without wavering, but we need to stay vigilant about AI stocks.

Let's start with my view: Long-term, there is no need to waver in faith in AI; this wave of technological revolution will completely disrupt social productivity. Computing power is in short supply, and AI concept stocks are bullish in the long run. Many A-share tech stocks and US semiconductor stocks that have risen too much in the early stage face pressure for weekly-level adjustments; any rebound is a selling opportunity, and funds will rotate to other low-position AI concept stocks.

Stock prices are driven by capital flows, not fundamentals

As many AI concept stocks have risen to this stage, what truly affects their prices is no longer fundamentals (profits), but rather the inflow and outflow of market capital.

In March-June this year, Nvidia's performance and outlook were far ahead of the rest, yet it underperformed most small-cap semiconductors (such as MRVL, AMD, ON, AMAT). This is because market capital was simultaneously going long on other semiconductors (CPU, memory concepts, analog chips) while shorting Nvidia to hedge systemic risk. However, when memory, CPU, and WFE started to fall, Nvidia instead became the main force supporting the market. Moreover, at this stage, apart from Nvidia, the stock prices of most semiconductor companies have already fully reflected optimistic expectations for 2027-2028 (especially MRVL, which once surged to 300).

On the other hand, in the Hong Kong and A-share markets, various tech stocks skyrocketed before July, completely overdrawing growth for the next few years. In contrast, since July, Xiaomi, Nio, XPeng, and Li Auto—stocks that should have been laggards—have risen (considering there have been many bad news items regarding smartphones and electric vehicles). This indicates that capital is covering its short positions. These rises have nothing to do with fundamentals; they are purely capital-driven behaviors.

Parabolic rallies require a painful deleveraging process

Since March, US semiconductor stocks have undergone an extremely crazy rally, identical to gold and silver earlier this year. In the A-share market, tech stocks like Yizhongtian also seem to be following a similar pattern.

When gold stood above 5500, everyone thought this time would be different because the logic behind gold seemed flawless. Yet, it still adjusted for over half a year, and gold hasn't even sent a signal of stabilizing at the weekly level hhh.

I believe US semiconductors and certain A-share tech stocks are the same. When everyone thinks this time is different, it might actually be the same. When something rises, the market looks for reasons to justify its rise; when it falls, the market finds reasons to justify its fall. The news we see is what the market wants us to see.

From a capital perspective, institutions have no reason to pull US semiconductors and A-share tech directions to new highs just to let retail investors break even.

Therefore, I believe that in the second half of the year, A-share tech and US semiconductors will likely experience a painful oscillation process similar to gold this year.

The next opportunity may lie with cloud providers

With the semiconductor track (especially semiconductor equipment in A-shares) being extremely crowded, capital flowing out of semiconductors must seek new targets for long positions (building narratives). I believe that, optimistically, the major cloud providers in China and the US are well-positioned to absorb market capital.

First, Alibaba, Google, Amazon, etc., have relatively low valuations, and their cloud business prospects are good under the push of AI. Second, their gains have lagged far behind those of semiconductors during the same period; they were even shorted to hedge systemic risk during the semiconductor rally. Now that semiconductor capital is unwinding, short covering combined with positive narratives makes a price surge quite hopeful.

However, from a seasonal perspective, liquidity is poor in summer, and VIX tends to rise. Currently, only individual stock volatility is high, but correlation is weak (e.g., if semiconductors crash hard, consumer goods, software, and M7 rise to support the market, causing the index to remain relatively stable). If hyperscalers' earnings miss expectations, it could trigger a rise in volatility across the entire index level, leading to a dump of the broader market. I expect at least one global stock market panic sell-off in the second half of this year, with QQQ dropping 5% in the short term and China's STAR 50 dropping 15% in the short term.

Outlook for Chinese Assets

When the big A-share index broke below 3800, the "National Team" entered to support the market, seemingly stabilizing sentiment. However, this administrative measure to ease downward pressure is merely a temporary fix; it cannot solve the root problem: stock prices need to be stabilized through profitability.

Looking back at March-April this year, many tech enterprises, especially AI-related ones, had reasonable valuations. For example, JCET (a leader in semiconductor packaging and testing) was one I favored. But within just two months, JCET's stock price more than doubled. Similar tech stocks are everywhere. These gains are clearly due to capital clustering, excessively overdrawing future fundamentals and stock prices. It is now obvious that clustered capital is beginning to exit. Policy can only slow down selling pressure. My personal view is that the A-share tech sector is about to undergo a deep correction and oscillation, lasting at least three months.

Furthermore, apart from AI-related tech stocks, the profit prospects of other stocks in the market are hard to be optimistic about, accompanied by CN's sluggish economy. Simply put, consumption is insufficient, and economic growth relies solely on exports and AI computing power capital expenditure. Therefore, I assert that the bull market at the index level of the Chinese stock market has basically ended (the end point could even be moved up to March, near 4200). There will be no broad-based index bull market (including CSI 300, CSI 500, CSI 1000) in the future. Specifically, CSI 300, CSI 500, and CSI 1000 will not simultaneously reach new highs in the next two years.

So where are the opportunities?

One is the return of dividend assets. There is a possibility of capital shifting from the AI sector to the dividend track, including the National Team stabilizing indices by supporting large central and state-owned enterprises.

Second, some low-position Hong Kong stocks have the potential for short-term rebounds, such as Alibaba, which I mentioned. Benefiting from domestic computing demand, its cloud business has growth potential, and its food delivery service is reducing losses.

Third, the domestic computing power sector. Such as SMIC (a leader in semiconductor foundry services) and JCET (a leader in semiconductor packaging and testing). It is known that the shortage of domestic chips will persist for a long time. However, one must wait for the semiconductor sector to fully adjust and valuations to become reasonable before entering. These enterprises benefiting from domestic capital expenditure growth will have more long-term development potential than those benefiting from overseas cloud giants' capital expenditure.

The first can be participated in by buying dividend ETFs. For the second and third, I think it is difficult to participate via ETFs.