Recently, I’ve been looking into which AI stocks are worth paying attention to, and I’ve found that many people in the market are still blindly chasing themes. In reality, they don’t really understand the logic of the AI industry chain.

To be honest, AI is not an industry but an entire supply chain. The upstream is computing power (like NVIDIA, TSMC), the midstream is cloud platforms (Microsoft, Amazon), and the downstream is application software. Each layer’s driving factors are completely different, and the logic behind stock price movements also varies.

The most direct upstream factor is supply and demand for AI chips and their prices. NVIDIA currently controls about 80-90% of the global AI accelerator market, earning over $100 billion annually just from data center GPUs. Its moat isn’t just in hardware but also in a software ecosystem built over more than a decade, making developer switching costs extremely high. TSMC is also critical because almost all high-end AI chips rely on TSMC’s advanced process technology and CoWoS packaging. This year, TSMC has initiated four consecutive years of price increases, with AI chip prices rising by 10%, and customers are still eager to buy.

In the midstream, Microsoft, Amazon, and Google profit from cloud AI services. Microsoft, due to its exclusive partnership with OpenAI, has deeply integrated Azure AI platform and Copilot, which has already penetrated the product ecosystem of 1 billion users worldwide. Amazon has tied up with Anthropic; AWS is both a cloud partner and a supplier of self-developed chips, forming a complete closed loop.

On the downstream application layer, software companies like Salesforce and ServiceNow are developing AI assistants and automation tools. Meta is the most direct monetization example—by optimizing advertising AI, which directly reflects in revenue.

The question is, can you still buy AI stocks now? My view is yes, but with stratification. For stable investments, choose Microsoft, Amazon, and TSMC—these companies are solid, and AI is just one of their growth drivers. To capture mainstream capital, NVIDIA and Meta are more volatile but have strong growth momentum. If you’re willing to accept high risk, second-tier AI chip companies and innovative startups in applications offer the greatest flexibility.

However, be aware of the risks. Valuations are already very high, and many companies’ stock prices have long reflected years of growth expectations. Competitors like AMD and Google’s self-developed TPU are catching up, and geopolitical export controls could impact supply chains. Capital rotation is also an issue; the market might shift from hardware to software, or from AI to other themes.

Looking back at the internet bubble era, Cisco’s stock once surged to $82 but later fell more than 90%. Even though Cisco’s fundamentals remained solid, its stock price has yet to return to its high point. This reminds us that infrastructure-type companies, even if stable, might be better suited for phased deployment rather than holding long-term without action.

Therefore, I recommend adopting a phased investment approach. 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, opportunities for strategic positions in AI stocks still exist.

In terms of trading strategy, staggered entry, waiting for dips, and controlling position sizes are more pragmatic. You can also consider investing through AI ETFs to diversify and reduce single-stock volatility risk. If you want to catch sector rotation opportunities, flexible trading tools can help you enter and exit more quickly.

Overall, from 2026 to 2030, the AI investment landscape is likely characterized by “long-term optimism with short-term volatility.” Prioritizing chipmakers, server accelerators, and infrastructure providers, or selecting companies with tangible applications, is a more prudent strategy.