Recently, many people have been asking whether AI stocks are worth buying. My personal view is: yes, but only if you understand what exactly you are buying.

Many people treat AI as a single industry to speculate on, but that’s actually a misconception. AI is not just one industry; it’s an entire supply chain, from upstream chip manufacturing, midstream cloud platforms, to downstream application software. Each segment makes money differently, and the logic behind their stock prices is completely different.

Let me start with the upstream. Companies like NVIDIA, TSMC, and AMD mainly see their stock prices driven by the supply and demand and pricing of AI chips. NVIDIA currently accounts for about 80% to 90% of the revenue share in the AI accelerator market. This moat isn’t just in hardware; it’s also in their software ecosystem built over more than a decade, making developers accustomed to coding on their platform. The cost to switch to another platform is prohibitively high. TSMC is similar, because all high-end AI chips must be produced on their advanced process nodes. Since January this year, they’ve been increasing prices for processes below 5 nanometers for four consecutive years. In the AI chip industry, foundries have no room for negotiation.

The midstream includes cloud giants like Microsoft, Amazon, and Google. They don’t sell chips directly but offer computing power services and APIs. This layer focuses on cloud revenue growth rates and the return cycle of capital expenditures. An interesting phenomenon here is: when upstream chip stocks surge too much, the costs for midstream companies get squeezed. As a result, some cloud companies have started developing their own chips to reduce costs, which in turn can impact the long-term profits of upstream suppliers.

Downstream are application software companies like Salesforce, ServiceNow, and Adobe. They embed AI into their products. This layer looks at whether enterprises are willing to pay extra for AI features and whether new AI tools are emerging to threaten their business. Usually, downstream stocks lag 1 to 2 quarters behind upstream stocks because it takes time to convert hardware investments into application-layer revenue.

If you ask me how to choose AI stocks now, I suggest different approaches depending on the situation. For stability, pick Microsoft, Amazon, and TSMC—these companies are solid, with AI just being part of their growth. Even if the AI hype cools down, their core businesses can still support them. To catch the mainstream capital, look at NVIDIA and Meta—they are tightly linked to AI, with strong growth momentum but also higher volatility.

AI stocks in Taiwan are also worth watching. TSMC is fundamental; Hon Hai and Quanta are involved in system integration; cooling component makers like Chih Hsin and Shuang Hong see rising demand because AI servers consume more power, making liquid cooling solutions essential. Their structural demand is increasing.

However, I must be honest: investing in AI stocks now carries significant risks. Valuations have already soared, and many companies’ stock prices have long reflected years of growth expectations. If growth slows or market sentiment shifts, the correction could be substantial. Plus, there’s the risk of capital rotation—markets might suddenly shift from hardware to software, or from AI to other themes.

Historically, Cisco’s stock peaked at $82 during the 2000 dot-com bubble. After the bubble burst, it fell over 90% and has never returned to that high. This lesson tells us that even infrastructure giants, AI stocks are not necessarily suitable for long-term, buy-and-hold without adjustments.

My approach is to adopt a phased investment strategy. Invest in batches, wait for pullbacks, and control the position size of individual stocks. Also, continuously monitor whether AI technology development is slowing, whether application monetization is meeting expectations, and whether individual companies’ profit growth is decelerating. As long as these conditions remain, AI stocks still hold investment value.

In the long run, AI’s impact on human life and productivity will be no less than the internet revolution. This big trend is certain. But in the short term, stock prices will definitely fluctuate—that’s normal. The key is to understand which part of the supply chain you are buying, why you are buying, and when to reduce your holdings. Only then can you truly profit from AI’s growth dividends instead of getting overwhelmed by volatility.