Semiconductors took a hard hit today, and I’m sure many of you weren’t feeling great when you opened your portfolios—neither was I.

I went through JPMorgan’s latest *Flows & Liquidity* report and summarized it for everyone.
JPMorgan’s overall view is fairly optimistic.
JPMorgan’s newly released *Flows & Liquidity* is a macro strategy report that analyzes markets from the perspective of global capital flows, market positioning, and liquidity. The report mainly discusses the AI trade rotation, suggesting that the market is beginning to enter a phase of rotation within the industry chain. Over the past year, semiconductor, HBM, and memory companies have consistently outperformed cloud providers like Meta, Microsoft, and Google—a divergence that is difficult to sustain for long. JPMorgan believes the more likely scenario is that as AI commercialization advances, cloud providers’ profitability gradually improves, and their stock prices catch up with the hardware sector. What the market currently fears is another possibility: if cloud providers’ future earnings growth fails to keep pace with capital expenditures, they may slow down AI investment, thereby impacting demand for GPUs, HBM, and other hardware. To this end, JPMorgan tracked several indicators including short interest, GPU rental prices, and token prices, concluding that while the market has indeed begun discussing this risk, there is not yet sufficient evidence to suggest an inflection point in the AI industry. At the same time, U.S. liquidity continues to improve, with M2 expected to increase by approximately $1.8 trillion by 2026, banks more willing to lend, and the Fed’s balance sheet modestly expanding again. In the absence of major black swan events, ample liquidity should continue to support U.S. equities, so JPMorgan’s overall view remains optimistic.

Combining this with today’s market movement, I’ll also share my own understanding.
The biggest divergence in the market recently has been around how profits in the AI industry chain will be distributed in the future.
Over the past year, hardware companies like chips, HBM, and memory have led in stock performance, while cloud providers have significantly lagged. As this gap widens, the market is starting to worry whether cloud providers can sustain such high levels of capital expenditure.
Meta recently leaked news about renting out some of its GPU computing power, further amplifying these concerns. Many people are now speculating: if more and more large companies put idle computing power up for rent, could it slow down the pace of purchasing new GPUs, HBM, and memory across the industry?
There are also reports that OpenAI is seeking more funding to support AI infrastructure construction. This reflects that the development of large models still requires massive investment—training and deploying advanced models remains a very capital-intensive business.
So today, the market is essentially reassessing the profitability expectations of the AI industry chain over the next few years, recalculating how much money each link in the chain can make in the future.
As of now, I haven’t seen any long-term logic change. The U.S. hasn’t stopped investing in AI, and China is also continuing to ramp up AI infrastructure. No cloud provider has announced cuts in CapEx. Core drivers like tight HBM supply-demand, data center expansion, and growing AI inference demand are all still intact.
In the short term, Meta’s news did cool the semiconductor sector a bit, giving the market a chance to digest sentiment after consecutive gains. But for long-term investors, a couple of days of volatility doesn’t change a company’s value over the next several years.
What’s truly worth watching is the upcoming earnings season in July.
No amount of market discussion is as convincing as the CapEx guidance, order data, HBM supply-demand dynamics, and GPU delivery figures in the earnings reports. If these numbers remain strong, many of the current market concerns will naturally fade over time. The upward path won’t be smooth. Every bull market goes through skepticism, negative news, and FUD—shaking out the weak hands before moving higher.
Next week, I’ll be closely watching the SMH daily EMA50 zone around 575, and I believe it will likely hold.

Finally, let me share a bit of my own thoughts.
If you’re a long-term investor, I don’t think it’s necessary to disrupt your rhythm because of a few days of volatility.
Personally, I prefer holding common stock of quality companies, or LEAPS with maturities longer than one year, rather than focusing on short-term options. Short-term calls require not only getting the direction right but also dealing with time decay and volatility. Many times, even if you’re right on direction, you may not make money. It’s fine to play with short-term trades, but they should never be the main part of your portfolio—this bears repeating!
The big direction of AI hasn’t changed. The market is just starting to ask deeper questions, such as how profits will be split within the industry chain and whether cloud providers can maintain such high spending. These discussions are normal and are part of the process any emerging industry goes through.
I feel that the current stage is somewhat reminiscent of the second half of 1998 during the internet era. At that time, the internet had already begun to change the world, but the real explosion was still ahead. The analogy won’t be exact, and history never repeats itself exactly. But I believe we are still in the middle of a long-term industrial trend—perhaps only halfway through, still 1-2 years away from the end. I will patiently hold until the end of next year.
Currently, neither the S&P 500 nor the Nasdaq is outside historical valuation ranges. Corporate earnings are still growing, and AI CapEx has not shown any significant contraction. Against this backdrop, I prefer to focus on the long-term competitiveness of companies.
Every day brings new news and new doubts, but what ultimately drives stock prices higher over the long term is earnings and profitability.
Manage your position sizing. Don’t go all in easily, and don’t dismiss the entire trend because of one pullback. Every major dip can be an opportunity to buy a little according to your plan. As long as AI CapEx hasn’t clearly turned and the industry trend hasn’t changed, I will choose to continue holding and patiently wait for this AI wave to fully play out.