# How to Evaluate AI Stocks: Metrics That Matter Beyond Hype

Source: https://cluenex.com/learn/technical-analysis/ai-stocks-valuation/
Publisher: Cluenex (https://cluenex.com)
Updated: 2026-10-15

## Short answer

Judge an AI stock by whether AI spending turns into reported revenue, margins and free cash flow. Check capital spending as a share of revenue, the revenue line tied to AI, operating margin trend, customer concentration and the price paid per dollar of earnings or sales.

## Key takeaways

- Capital spending (capex) is a cost for the buyer and revenue for the seller, so the same number reads differently depending on which side the company is on.
- Revenue that AI customers actually pay for, shown in a filing, matters more than the word AI in a press release.
- Operating margin and free cash flow show whether growth is profitable or just bought.
- Customer concentration, often a few big cloud companies, is the main hidden risk, and the 10-K discloses it.
- Valuation still matters: a great business can be a poor investment at a high enough price.

## What does it mean to evaluate an AI stock?

It means checking whether the AI story shows up in the financial statements. A story is a claim; a filing is a record. Public US companies must report revenue, costs, capital spending and risks to the SEC, so those documents are the best place to test the hype.

The checks below work for any company that says AI drives its future, from chip designers to software firms to cloud providers.

## Is the company selling AI or buying AI?

Start by sorting the company into one of two roles, because the same metric means opposite things in each.

| Role | Example type | What rising AI spending means |
|---|---|---|
| Seller | Chip makers, networking, some software | Revenue, if customers keep ordering |
| Buyer | Cloud providers, large platforms | A cost that must earn a return later |
| Both | Large cloud companies | Revenue from customers and cost of their own build-out |

A chip company benefits when a cloud company raises its capex. The cloud company has to earn that money back through AI services. Read how the AI Capex chain works in [How the AI Boom Turns Into a Stock Boom](/learn/markets/ai-capex-chain/).

## How do you read capex for an AI company?

Capex is cash spent on long-lived assets such as data centers and servers. It appears in the cash flow statement under investing activities, and the 10-K or 10-Q gives the exact figure.

Compute capex intensity: capex divided by revenue. Example with made-up numbers: capex of $20 billion on revenue of $100 billion is 20%. If that ratio rises each year while revenue growth slows, the company is spending faster than it is earning.

For a buyer, higher capex is not automatically good. It is a bet. The test is whether revenue and operating income grow enough to cover it. See [Capex Today, Revenue Tomorrow](/learn/markets/ai-capex-vs-revenue-payoff/) for that test in detail.

## Is the AI revenue real?

Look for revenue the company reports in a segment or names in its filings, not a phrase in a call. Some companies break out data center or cloud revenue; others only mention AI in passing.

Questions to ask:

- Is there a reported line item, or only a verbal estimate?
- Is it growing faster than total revenue?
- Does the company define what counts as AI revenue, and has that definition changed?

If a company does not disclose AI revenue, treat any outside estimate as a guess, not a fact. A company that grows total revenue 10% with a reported AI segment growing 40% is showing something. One that only says AI is a priority is showing nothing yet.

## What do margins and free cash flow say?

Operating margin is operating income divided by revenue. It shows how much of each sales dollar remains after running costs. Free cash flow is operating cash flow minus capex, and it shows cash left over. Learn more in [Free Cash Flow Explained](/learn/fundamental-analysis/free-cash-flow/) and [Gross Margin vs Operating Margin vs Net Margin](/learn/fundamental-analysis/gross-operating-net-margin/).

Worked example with made-up numbers: revenue of $50 billion and operating income of $10 billion is a 20% operating margin. If next year revenue is $65 billion and operating income is $11.7 billion, margin falls to 18%. Revenue grew 30%, but each dollar of sales is less profitable. That is growth bought with cost.

Compare against peers. For context, 15.9% (Cluenex data, https://cluenex.com/learn/data/sector-benchmarks/) is a useful yardstick, and 61.5% (Cluenex data, https://cluenex.com/learn/data/sector-benchmarks/) shows how much room the sector usually has.

## How concentrated are the customers?

A seller whose sales depend on a handful of buyers carries a risk that the average hides. The 10-K must disclose customers that make up 10% or more of revenue, usually in the notes to the financial statements or the risk factors.

Example with made-up numbers: if two customers account for 40% of revenue and both cut orders by a quarter, revenue falls about 10% from that alone. Check how many large customers there are and whether they build their own chips or software, which could replace the seller. For how a durable advantage looks, see [How to Evaluate a Company's Moat](/learn/fundamental-analysis/moat-analysis/).

## What price is too high for an AI stock?

A high price-to-earnings (P/E) ratio means investors pay a lot for each dollar of profit. It is not wrong by itself, but it leaves little room for disappointment. The Technology sector median P/E is 33.0x (Cluenex data, https://cluenex.com/learn/data/sector-benchmarks/), against 22.2x (Cluenex data, https://cluenex.com/learn/data/sector-benchmarks/) for the whole group.

If profits are small or negative, use price-to-sales, but remember it ignores costs. See [Price-to-Sales Ratio Explained](/learn/fundamental-analysis/price-to-sales-ratio/) and [Forward P/E vs Trailing P/E](/learn/fundamental-analysis/forward-pe-vs-trailing-pe/).

Simple reverse check: if a stock trades at 40 times earnings, it needs earnings to grow enough to make that multiple look reasonable later. If growth slows, the multiple can shrink even when profit still rises. [Why a Company's Profit Can Rise While Its Stock Price Falls](/learn/fundamental-analysis/pe-multiple-compression/) explains that.

## Where does this approach fail?

No checklist predicts prices. These limits matter:

- **Capex can be wasted.** Spending is not proof of returns. Past tech build-outs left some firms with idle capacity.
- **Demand can shift fast.** Orders can be pulled forward, then drop.
- **Disclosure is uneven.** Many firms do not separate AI revenue, so some checks rely on judgment.
- **Good business, bad price.** A strong company can still deliver weak returns if bought at a very high valuation.
- **Backward-looking data.** Filings describe the past quarter, not the next one.

Also avoid rules such as a fixed win rate for a pattern. No source supports a guaranteed hit rate for AI stocks.

## What this means for you

If you own or are looking at AI-related stocks with a regular job and a brokerage account, build a short routine.

1. Open the latest 10-K or 10-Q and find revenue, operating income, capex and customer concentration. [What's Actually Inside a 10-K](/learn/fundamental-analysis/10k-filing-explained/) shows where.
2. Write down capex divided by revenue and operating margin for the last four quarters.
3. Note whether AI revenue is a reported line or only a talking point.
4. Compare valuation with the sector median and with the company's own history.
5. Check how much of your portfolio sits in one theme. [Concentrated Stock Positions](/learn/portfolio-management/concentrated-stock-position/) covers sizing.

Cluenex gives each of about 1,000 large US stocks a free public page with business quality, the main risk and insider activity at [cluenex.com/stocks/](https://cluenex.com/stocks/), which can speed up the first pass. Treat everything here as education, not a call on any security.

## Questions people ask

### How do I find out how much revenue comes from AI?

Check the segment note in the 10-K or 10-Q and the earnings release for a named line such as data center or cloud. If the company only mentions AI in a call without a figure, treat the AI share as unknown.

### Does the P/E ratio work for AI stocks?

It works for profitable companies, and should be compared with sector peers and the company's own history. For firms with little or no profit, price-to-sales is a rough substitute, but it ignores costs.

### Is high capex good or bad for an AI company?

It depends on the role. For a chip or equipment seller it is a customer's spending and can become revenue. For the spender it is a cost that must be earned back through higher revenue and profit.

### What is customer concentration and why does it matter?

It is the share of sales coming from a few buyers. US companies disclose customers above 10% of revenue, and losing one of them can cut revenue sharply.

### Can AI stocks fall even if AI demand keeps growing?

Yes. If the price already assumes very high growth, slower growth can lower the stock even while revenue and profit still rise.

## Sources

- [Beginners' Guide to Financial Statements](https://www.sec.gov/about/reports-publications/beginners-guide-financial-statements)
- [EDGAR Full Text Search and Company Filings](https://www.sec.gov/edgar/search/)
- [Price-Earnings Ratio](https://www.investor.gov/introduction-investing/investing-basics/glossary/price-earnings-ratio)

Cite as: Cluenex. "How to Evaluate AI Stocks: Metrics That Matter Beyond Hype." Cluenex Learn, updated Oct 15, 2026. https://cluenex.com/learn/technical-analysis/ai-stocks-valuation/