AI Stocks: The Data-Driven Overview of the Sector
“AI stocks” is not a sector but a cross-section of four very different business models — from the chipmaker with 70% gross margins to the insurer using AI merely as an efficiency lever. Treating the space as one block means buying risks you don’t see.
This overview delivers the taxonomy, puts valuation levels in realistic perspective, and shows where the concentration risk in your own portfolio is usually bigger than you think.
The sector taxonomy: four layers of AI value creation
The space splits sensibly into four layers that respond to the AI boom in completely different ways:
- Semiconductors & hardware: Nvidia, AMD, Broadcom, TSMC, ASML. Earning real money from the boom today — and carrying the full capex cycle risk. The infrastructure side is covered in depth in AI data center stocks.
- Cloud & platforms: Microsoft, Alphabet, Amazon, Meta. Funding the build-out from their own cash flows; the open question is the return on those AI billions.
- AI software & models: Palantir, ServiceNow, specialized vendors. Valuations are most aggressive here and business models least proven.
- Adopters: companies across industries cutting costs with AI. The least hyped — and long term possibly the broadest — group of beneficiaries.
| Layer | Examples | Earnings today | Valuation risk |
|---|---|---|---|
| Semiconductors & hardware | Nvidia, TSMC, ASML | Real, high margins | High (cycle) |
| Cloud & platforms | Microsoft, Alphabet, Amazon | Strong, AI share unclear | Medium |
| AI software | Palantir, ServiceNow | Growth, often expensive | Very high |
| Adopters | Across all industries | Efficiency gains, hard to measure | Low to medium |
AI-powered stock and portfolio analysis with sentiment, risk score and research assistant – try it for free.
Valuation realism: what is priced in
The sober assessment: most major AI beneficiaries trade well above their historical valuation averages and above the broad market. That is not automatically a bubble — unlike 2000, the earnings are real. But it means prices embed the assumption that capex growth continues for years and the platforms’ AI investments monetize. If either assumption disappoints, the drop is substantial.
Practically: expected return does not come from the story, but from the gap between expectations and reality. Buying today requires a reason why already-high expectations will still be exceeded.
The underestimated concentration risk in your own portfolio
The most common mistake is doubled exposure: the MSCI World is roughly 70% US stocks, and its largest holdings are precisely the AI platforms and chip names. Stacking Nvidia, Microsoft and a tech ETF on top of a world ETF can quickly tie €30,000 or more of a €100,000 portfolio to the same factor.
Before adding AI exposure: measure what you already have. A bundled alternative to single names are AI ETFs — with their own pitfalls around overlap and costs.
Frequently asked questions
What actually counts as an AI stock?
Four layers make sense: semiconductors/hardware (Nvidia, TSMC), cloud platforms (Microsoft, Alphabet), AI software (Palantir), and adopters using AI as an efficiency lever. The layers differ fundamentally in margins, valuation and cycle risk.
Are AI stocks currently overvalued?
Most major AI beneficiaries trade well above historical averages — but unlike 2000, with real earnings. Priced in are continued capex growth and successful monetization; if either disappoints, the downside is significant.
How much AI exposure do I already have via an MSCI World ETF?
A lot: the index is roughly 70% US stocks, and its top holdings are the big platform and chip names. Additional single AI names often increase concentration risk more than expected.
AI-powered stock and portfolio analysis with sentiment, risk score and research assistant – try it for free.