"AI stocks" is one of the hottest searches, but there's no single way to invest in artificial intelligence. Behind the boom sits a chain of very different companies: the ones that make the chips, the ones that build the data centers, the ones that deploy AI in the cloud, and even the ones that generate the electricity all of this consumes. This is the map of that ecosystem —and how to judge, with the numbers, whether they're already expensive.
There's no single "AI stock"
AI isn't a company or a sector: it's a wave that cuts across several industries. Thinking in "layers" of a chain helps you understand who earns what. We go from the bottom (the silicon) up to the top (the applications), with the most representative companies of each. Every name links to its fundamental analysis so you can see its numbers, not just the story.
1. The chips: the engine of AI
Training and running AI takes enormous computing power, and that means specialized chips.
- NVIDIA: the undisputed king of AI GPUs; its chips are the standard for training models. Its moat isn't just the chip, but its software (CUDA).
- AMD: the main challenger in data-center chips.
- Broadcom: designs custom AI chips for big tech and dominates the networking that connects them.
- Micron: makes the high-speed memory (HBM) that AI chips need so they don't run "out of breath".
2. Making the chips: the suppliers to the fabs
Someone has to physically manufacture those chips. The big fabs (such as Taiwan's TSMC or Europe's ASML, which aren't U.S.-listed) buy their equipment from a handful of companies:
- Applied Materials and Lam Research: make the machines that etch and deposit circuits onto silicon. They're the "picks and shovels" of the sector.
3. The infrastructure: inside the data center
Chips don't work on their own: they're mounted on servers, cooled, connected and powered inside giant data centers.
- Vertiv: power and cooling for data centers, a physical bottleneck of the boom.
- Dell and Super Micro: assemble the AI servers.
- Arista Networks: the ultra-fast networking that connects thousands of chips to each other.
4. The cloud: who deploys AI at scale
Big tech ("hyperscalers") builds the data centers and rents AI power to the rest of the world. They're both the largest buyers of chips and the ones monetizing AI in their own products:
- Microsoft (Azure, Copilot), Alphabet (Google Cloud, Gemini) and Amazon (AWS).
- Meta doesn't rent cloud, but it's one of the biggest investors in AI for its own products.
5. The software: AI applied
The layer closest to the user: companies that embed AI into tools they sell to other businesses.
- Palantir, ServiceNow and Oracle: each builds AI into its enterprise software (and, in Oracle's case, into its cloud too).
6. The hidden layer: energy
This one escapes almost everyone. AI data centers consume brutal amounts of electricity, and that has turned utilities into unexpected beneficiaries of the boom:
- Vistra, Constellation Energy and NRG Energy: generate the electricity (including nuclear) the new data centers demand.
The step almost nobody takes: are they expensive?
Here's what matters, and what separates investing from speculating. A company being "in AI" doesn't mean its stock is a good buy: many of these already trade at prices that assume years of perfect growth. A great story can be a mediocre investment if you overpay.
That's why the last step is to look at the fundamentals: is the P/E reasonable for its growth? Do margins and ROE back up the story? Type the ticker of any of these companies into the analyzer and you'll get its valuation rating in seconds. And if in doubt, start with the how to tell if a stock is cheap or expensive guide.
This article is educational and describes the AI ecosystem. It is not a buy recommendation or a prediction of which companies will win. Always do your own analysis.