Capex-to-Sales: The AI-Era Spending Gauge for Stocks

Meta (META) lost roughly 7% of its market cap on April 29, 2026, after raising full-year capex guidance by ~$10 billion. The single number on the buy-side spreadsheet that triggered the selloff was capex-to-sales — and most retail investors do not even watch it.

What is the capex-to-sales ratio?

Capex-to-sales is one of the simplest yet most underused valuation diagnostics. The formula:

Capex-to-Sales = Capital Expenditures ÷ Revenue

Both numbers come from the same period (typically trailing twelve months or the current fiscal year). The output is usually expressed as a percentage. A capex-to-sales ratio of 30% means the company is reinvesting 30 cents of every revenue dollar into property, plant, equipment, or capitalized software — that money is leaving the income statement and showing up on the balance sheet as long-lived assets.

Why it matters: capex hits free cash flow, not earnings. A company can post strong EPS while burning all of it in the capex line. Investors who anchor only on the P/E ratio miss this entirely.

For a complete picture of how capex distorts net income, see our breakdown of free cash flow yield, the metric that adjusts for exactly this gap.

Why does capex-to-sales matter more in 2026 than in past cycles?

Because the AI build-out has pushed software-adjacent companies into industrial-scale capex profiles. For ~30 years, software companies enjoyed capital-light economics. Microsoft, Oracle, and SAP ran at single-digit capex-to-sales. The asset on the balance sheet was the developer team, not the data center.

That is over. Microsoft (MSFT), Alphabet (GOOGL), Meta (META), and Amazon (AMZN) are now collectively spending more than the entire US oil & gas industry on capex. When Alphabet (GOOGL) guided to roughly $185 billion in 2026 capex on revenue tracking near $475 billion, the implied ratio crossed 38%.

That number used to belong to railroads. Now it belongs to ad-supported search. The whole framework for how investors value these stocks has to adjust.

How do you calculate capex-to-sales for a stock?

Three steps.

First, find capex. On the cash flow statement, look for "Purchase of property, plant and equipment" or "Capital expenditures." Some companies break out capitalized software separately — include both for the cleanest comp.

Second, find revenue. Top line of the income statement, same period. Use trailing twelve months for stability or the most recent fiscal year for current-state.

Third, divide. Annualize if mixing quarters — do not compare a Q1 capex number to a full-year revenue number. A common mistake is using forward revenue with trailing capex, which produces a misleadingly low ratio.

Most stock data providers will display the ratio directly. If you want to verify, the 10-K and 10-Q filings have everything you need.

What does a high capex-to-sales ratio actually mean?

It depends entirely on the industry baseline.

The same 30% capex-to-sales ratio is normal for Nvidia (NVDA) building chip capacity through partners, alarming for Costco (COST), and a sign of a company building a moat for Microsoft (MSFT). Context is everything. The number alone tells you nothing — you have to compare against industry baseline AND the company's own three-year trend.

Why is Meta's capex-to-sales spiking the alarm bells?

Because the rate of change matters more than the absolute level. Meta (META) was running at roughly 18% capex-to-sales in 2023. The new FY2026 guide of ~$125-145 billion against approximately $200 billion in expected revenue puts the ratio in the 60-72% range — more than triple the historical level.

That kind of step-change forces a re-rating. At ~18% capex-to-sales, Meta could trade like a software company. At 60%+, it has to trade like a hybrid software-infrastructure company, which carries a different and lower multiple. The market's 7% selloff on April 29 was not really about Q1 — it was the buy-side recalculating multiples for the new capex profile.

The bull case: if AI revenue ramps quickly, the ratio normalizes by 2027-2028. The bear case: if AI revenue lags and component costs keep rising, the ratio stays elevated and free cash flow stays compressed for multiple years.

Real examples: 5 stocks ranked by 2026 capex-to-sales

A snapshot of where major names sit going into Q2 2026:

Amazon (AMZN) looks better than the others because so much of its revenue is low-margin retail; the AWS capex is high but absolute revenue is huge. Costco (COST) is the textbook capital-light comp — almost all reinvestment goes into inventory turns, not buildings or chips.

For a primer on how Buffett evaluates capital allocation across very different capex profiles, see our Charlie Munger profile.

Common mistakes investors make with capex

Three traps.

The first trap: assuming low capex is always good. A company under-investing in capex can look attractive on free cash flow today and uncompetitive in five years. Intel (INTC) under-spent on advanced node capacity for years, and the result was being lapped by competitors on chip technology.

The second trap: comparing across industries without adjusting. A 25% capex-to-sales ratio is alarming for a software company and unremarkable for a railroad like Union Pacific (UNP). Always normalize against industry peers.

The third trap: ignoring capex efficiency. Two companies can spend the same capex and get very different revenue out of it. The ratio of incremental revenue to incremental capex (sometimes called "capex productivity") is the real read-through. Hyperscalers right now are being judged not on capex level but on whether each new dollar of capex shows up as revenue — that is exactly what punished Meta and rewarded Alphabet on April 29.

Pro tips: when capex-to-sales actually predicts returns

Three setups where the ratio is the cleanest signal.

First, when capex-to-sales is rising AND revenue growth is decelerating. This is the classic value-trap signal. The company is spending more to grow less. Treat it as a yellow flag.

Second, when capex-to-sales is falling AND ROIC is rising. This is the "operating leverage kicking in" signal. The capex investments from prior years are now generating returns. Amazon (AMZN) AWS in 2018-2020 was the textbook case.

Third, when capex-to-sales spikes due to a specific project. If a company is building a new fab, a new pipeline, or a new content library, the ratio temporarily rises. Look at the asset depreciation schedule — if the project depreciates over 15-20 years, the spike is rational; if it depreciates over 3-5 years and competitors can copy, the spike is dangerous.

When NOT to use capex-to-sales

In two cases.

First, for early-stage growth companies. A startup growing 80% YoY with high capex looks bad on the ratio but may be exactly the company you want to own. Use revenue growth + path to profitability instead.

Second, for asset-light businesses with high stock-based compensation. Many software companies that look "low capex" are actually compensating with stock instead of cash. The real capital intensity is hidden in dilution. For these, total cash + SBC ÷ revenue is a better metric. See our piece on stock-based compensation for that adjustment.

For a complementary metric on judging capital allocation discipline, our free cash flow yield guide pairs naturally with capex-to-sales.

Ready to analyze these stocks yourself? Search any ticker on MainRatios to see valuations from 6 legendary investors - free.