AI & stocks
Valuing AI stocks
A stock is not expensive because the price is high, but because the price assumes more future profit than the business is likely to deliver. That judgement is hard for AI names, where growth rates are extreme and capital spending is vast. Below we set out the measures you need.
P/E and PEG
The price-earnings ratio compares price with earnings per share. For fast growers it is almost always high; the forward P/E, based on the next twelve months of earnings, is more useful. Compare it with the company's own history rather than with the broad market.
PEG divides the P/E by expected earnings growth. A P/E of 40 with 40 percent growth gives a PEG of 1 and is defensible; the same P/E on 10 percent growth is not. The weak spot remains the growth estimate: it is an assumption, not a fact.
- Use the forward P/E, not just the trailing one
- A PEG near 1 is defensible when growth is durable
- Compare within the same layer of the chain
Capex and free cash flow
AI is capital intensive. A company can report strong profit while retaining almost no free cash flow because everything goes into data centres. Track capital expenditure as a share of revenue, and free cash flow after that spending.
Rising capex is not automatically bad; it can mean management sees demand worth locking in. It becomes worrying when capex grows faster than revenue year after year and management shows no concrete return on it.
- Capex/revenue shows how capital hungry growth is
- Free cash flow is the real profit test
- Ask what return management targets on the spend
Revenue quality and concentration
Recurring subscription revenue is worth more than one-off hardware sales, and a broad customer base is worth more than three large buyers. Annual reports must disclose customers above ten percent of revenue; read that section.
Watch gross margin too. Software layering AI features on bought-in compute can show falling gross margin while revenue grows. An early sign that pricing is not keeping up with cost.
Dilution and stock-based pay
Technology companies pay staff largely in shares. That flatters reported costs but raises the share count, shrinking your stake a little every year. Look at share count growth over five years and at buybacks offsetting it.
Use fully diluted earnings per share and treat stock-based compensation as a real cost, even when adjusted figures strip it out.
A simple personal checklist
Valuation is not an exact science; consistency helps. Ask the same questions of every AI stock and write your answers down so you can later see whether your assumption held. Being wrong for four quarters is normal; never testing the assumption is not.
- Is revenue growing faster than capex?
- Is gross margin stable or improving?
- How dependent is the firm on three big customers?
- What am I paying per euro of expected profit two years out?
- How much dilution am I willing to accept per year?
Frequently asked questions
Is a high P/E always a warning sign?
No. A high P/E is defensible as long as earnings growth delivers on it. The danger is the reversal: when growth halves, the multiple usually contracts too, and both effects compound.
Why is free cash flow more important than profit?
Accounting profit can stay high while all the cash goes into data centres. Free cash flow shows what is genuinely left for shareholders after that investment.
How often should I revisit my analysis?
Every quarterly report is a natural checkpoint. Test whether the growth, margin and capex assumptions you bought on still hold.
Read next
AI stocks explained
Which companies actually earn money from artificial intelligence? A walk through the AI chain: chips, data centres, cloud, models and applications, with the risks in each layer.
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.