I've been thinking lately, instead of blindly following AI themes, it's better to first understand how this industry chain is divided. Because AI is not a single industry at all, but an entire supply chain, and different segments make money in completely different ways.

I realize that many people, when discussing AI stocks, often overlook the most critical point — understanding the industry structure first, in order to judge whether current stock prices are cheap or overvalued.

The entire AI industry chain can be roughly divided into three layers. Upstream is computing hardware, like NVIDIA, TSMC, Foxconn, which are directly driven by the supply and demand and prices of GPUs and AI chips. Midstream is the platform and model layer, where cloud giants like Microsoft, Amazon, Alphabet, and Meta operate, monetizing through cloud AI services and large model APIs. Downstream is the application software layer, where companies like Salesforce, ServiceNow, Adobe embed AI capabilities into their products.

I notice many only see NVIDIA's surge and follow suit, but the logic of upstream and midstream is completely different. Upstream is driven by chip supply and demand and capital expenditure, while midstream depends on cloud revenue growth and capital return cycles. That’s why sometimes NVIDIA’s rapid rise can actually suppress midstream stocks — because costs are increasing.

Regarding which AI stocks are worth paying attention to, my view is this: if you want the most certain targets, NVIDIA is absolutely unavoidable. It holds 80% to 90% of revenue in the AI accelerator market, and just data center GPUs generate over $100 billion annually. Even more impressive is its software ecosystem, with millions of developers accustomed to programming on NVIDIA platforms, making switching costs extremely high — that’s the real moat.

TSMC is also worth key attention. Almost all AI chips from NVIDIA, Apple, AMD are produced using TSMC’s advanced processes. In Q1 2026, its combined revenue will grow 35% year-over-year, with high-performance computing accounting for 58%, growing at 48% annually — this is the strongest engine driving growth. More critically, TSMC has started a continuous four-year price increase on all advanced processes below 5nm since January 1 this year, with AI chips seeing a 10% increase. Customers know they’ll face price hikes for four years but still rush to order.

I’ve also been watching Microsoft. Through its exclusive partnership with OpenAI, along with Azure AI platform and Copilot integration, it has successfully embedded AI technology seamlessly into global enterprise workflows. As Copilot is deeply integrated into Windows, Office, Teams — products with over a billion users — monetization capabilities will continue to unfold. Many institutions believe Microsoft is the most certain beneficiary of the enterprise AI adoption wave.

Amazon’s strategy I find very interesting. It acts as Anthropic’s main cloud partner via AWS, while also supplying self-developed AI chips. The invested money flows back through infrastructure fees, forming a complete closed loop. When the market re-focuses on the monetization potential of AI infrastructure, Amazon’s advantages are likely underestimated.

Meta is different — it’s a representative of the application layer. By optimizing advertising AI and open-sourcing Llama models, it directly monetizes AI. The precision of Facebook and Instagram’s ad targeting has greatly improved thanks to AI, which directly reflects in revenue. This is the most straightforward AI monetization case I’ve seen.

If you want to participate in AI but avoid too much volatility, Microsoft, Amazon, and TSMC are more stable choices. These companies have solid fundamentals, with AI being just one growth driver. Even if the AI hype cools down, their core businesses can still support their stock prices.

To capture mainstream capital flowing into AI, focus on NVIDIA and Meta. These companies are highly tied to AI, with strong growth momentum, but also more volatility. Suitable for investors who can tolerate some fluctuations and are willing to hold long-term.

I think the most important thing is to realize that AI concept stocks may still experience short-term turbulence, but the long-term trend remains upward. According to Gartner’s latest report, global AI spending is expected to reach $2.53 trillion in 2026, climbing further to $3.33 trillion in 2027. This number clearly shows AI is not just a fleeting theme.

However, I must also be honest — the risks in this field are indeed significant. Overvaluation, capital rotation, geopolitical issues, intensified competition — these are all factors to consider. Many stocks have already reflected years of growth expectations; if growth slows or market sentiment shifts, the pullback could be substantial.

Therefore, my advice is to adopt a phased investment approach. Instead of going all-in at once, stagger your positions, wait for dips, and control your single-stock exposure. Also, continuously monitor the pace of AI technology development, application monetization, and whether individual companies’ profit growth shows signs of slowing. Only if these conditions remain favorable can the investment value of AI concept stocks continue to be supported by the market.