AI & stocks
AI stocks explained
AI stocks is a catch-all label for listed companies that make money from artificial intelligence. It is not one homogeneous group: a chip designer, a data centre builder, a cloud provider and a software vendor have very different margins, risks and sensitivity to the investment cycle. Below we break the chain down layer by layer so you know what you are actually buying.
Layer 1: semiconductors and equipment
At the bottom of the chain sit the companies that produce compute: accelerator designers, memory makers and the firms that supply the machines used to manufacture chips. This layer earns the highest margins when demand peaks, but it is also the most cyclical: if a few large buyers delay capital spending for a couple of quarters, revenue drops abruptly.
Watch order books, inventory levels and customer concentration. When a handful of hyperscalers account for most of the revenue, one postponed build-out is enough to move the share price hard. Export controls add a second risk that is unrelated to demand itself.
- Highest margin, highest cyclicality
- Customer concentration is a structural risk
- Export rules can block revenue by region
Layer 2: data centres, power and networks
AI runs on physical infrastructure: halls full of servers, cooling, fibre and above all electricity. Data centre operators, network builders, cooling specialists and utilities benefit indirectly, often through long-term contracts that make revenue more predictable than for chipmakers.
Here the risk shifts from demand to capital: these companies borrow heavily to build. Rising rates, permitting delays or grid congestion feed straight through to returns. Look at leverage, contract duration and counterparty quality.
- Long contracts give steadier cash flows
- Interest rates and debt load dominate returns
- Grid access is a real constraint on build-out
Layer 3: cloud and models
The large cloud platforms rent out compute and offer models of their own. They sit in a unique position: they are the biggest buyers of AI silicon and simultaneously resell AI to millions of business customers. Their cloud growth is the sector's most watched indicator.
The flipside is capital intensity. Every euro spent on data centres today depresses free cash flow and has to be earned back in future cloud margin. Investors therefore track two numbers at once: cloud revenue growth and the capex required to produce it.
Layer 4: software and applications
At the top sit companies embedding AI in products: search and advertising platforms, enterprise software, security vendors and specialised applications in healthcare, logistics or financial services. They buy compute and sell outcomes.
The question here is not whether AI works, but whether the company can charge more for it than the compute costs. Check pricing, retention and whether AI features actually convert into new subscription revenue rather than being given away.
- Look for evidence of higher prices or new subscription revenue
- Compute costs can erode gross margin
- Proprietary data is often the real moat
How European investors get exposure
Most pure AI names are listed in the United States. European investors buy them directly through a broker or indirectly through a broad technology or world index fund, where the largest AI names already carry heavy weight. Thematic investors can pick an AI or semiconductor ETF, accepting that it is far more concentrated than a world index.
Mind currency risk: dollar-denominated prices move with EUR/USD. And remember that a thematic ETF often holds the same ten names you already own through a world index; double exposure is the most common mistake in this theme.
Frequently asked questions
Is an AI ETF better than individual stocks?
An ETF spreads company risk but not theme risk: if the AI investment cycle stalls, nearly every holding falls together. Single stocks give more control but require you to follow each company's numbers.
What is the biggest risk in AI stocks?
That spending on data centres and chips runs ahead of the revenue AI applications actually generate. When buyers delay budgets, the lower layers of the chain are hit first and hardest.
Do AI stocks overlap with my index fund?
Almost certainly. Global equity indices already give the largest technology firms a substantial weight. Add up your exposure before buying a thematic fund.
Read next
Valuing AI stocks
From P/E and PEG to capex, free cash flow and customer concentration: how to judge whether an AI stock is expensive or justifiably highly valued.
AI stocks alongside crypto
AI stocks and crypto move together more often than investors assume. How to set an allocation, manage correlation and avoid doubling the same risk.
What is blockchain?
A clear explanation of blockchain: how it works, why it creates trust without a middleman, and what its real-world risks and uses are.