Recently, I’ve noticed many people discussing AI concept stocks, especially in the U.S. stock market, but most don’t really understand the logic of the AI industry chain—they’re just blindly following the trend. I’ve spent quite some time researching this myself, and today I want to share some insights.

First, the conclusion: AI is definitely worth paying attention to, but don’t treat it as a single industry for investment. It’s actually a complete supply chain, with completely different playbooks for upstream, midstream, and downstream.

I divide it into three layers. The top layer is computing hardware, namely chips and servers. NVIDIA currently accounts for about 80% to 90% of the revenue share in the AI accelerator market, earning over $100 billion annually just from data center GPUs. TSMC is the key manufacturer of these chips; almost all high-end AI chips are produced by TSMC. Recently, TSMC has started raising prices for advanced processes below 5 nanometers for four consecutive years, with AI and high-performance computing chips seeing a 10% increase. Customers are still rushing to buy, which shows how tight the supply and demand are.

The midstream is cloud platforms and model APIs, mainly giants like Microsoft, Amazon, and Google. They don’t sell chips directly but offer computing power services. The most critical indicators here are cloud revenue growth rates and the return cycle on capital expenditure. The four major cloud giants’ combined capital expenditure is projected to approach $600-700 billion by 2026, which is the most direct leading indicator driving upstream chip demand. But here’s an interesting phenomenon: when upstream prices rise too sharply, midstream costs also increase. Google and Amazon are starting to develop their own chips to reduce costs, which could impact NVIDIA’s long-term market share.

The downstream layer is application software, like Salesforce, Adobe, and ServiceNow. They embed AI capabilities into their products. This layer mainly depends on enterprise adoption speed and the premium rate for AI features. Usually, downstream lags upstream by one to two quarters because after chips are shipped, it takes time to build infrastructure.

If you want to invest in AI concept stocks in the U.S., I would recommend based on your risk tolerance. If you prefer less volatility, Microsoft, Amazon, and TSMC are good choices; these companies are solid, with AI being just one of their growth drivers. If you want to capture the mainstream capital flow, NVIDIA and Meta are core targets—strong growth momentum but also higher volatility. Meta is particularly interesting: their AI-driven ad optimization directly monetizes, and the precision of Facebook and Instagram advertising has significantly improved thanks to AI, reflected very directly in revenue.

There are also many opportunities in Taiwan. TSMC, of course, is fundamental infrastructure. Hon Hai (Foxconn), as NVIDIA’s main server manufacturer, has recently seen some weakness in stock price, mainly because its gross margin improvement was below expectations. MediaTek is actively deploying AI chips; their Dimensity series already includes built-in AI computing units. Cooling solutions are also worth noting—companies like Chicony and Shuanghong are benefiting as AI servers’ power consumption continues to rise, making liquid cooling solutions a necessary feature. Their profit flexibility could also expand.

However, honestly, both U.S. and Taiwan AI concept stocks’ valuations have already been significantly inflated. I recall the Cisco example from the Internet era: during the dot-com bubble, its stock soared to $82, then fell more than 90%. Even though the company’s fundamentals remained decent, the stock never returned to its peak. This reminds us that infrastructure-type companies, even with solid fundamentals, are better suited for phased positioning rather than holding onto blindly.

My advice is to adopt a phased investment approach. Invest in batches, wait for pullbacks, and control individual stock positions. Also, keep an eye on key factors: whether the pace of AI technological development begins to slow, whether application monetization improves as expected, and whether individual companies’ profit growth slows down. As long as these conditions hold, the investment value of AI concept stocks can continue to be supported.

According to Gartner’s latest report, global AI spending is expected to reach $2.53 trillion by 2026, climbing further to $3.33 trillion in 2027. In the long run, AI’s impact on human life will likely be no less than the internet revolution, but short-term fluctuations are still possible. The most pragmatic approach is to stay focused on industry dynamics and enter or exit at the right moments.