You spot a big gap in the 10‑K
The filing is open, you’re in the cash flow statement, and the line for “capital expenditures” is doing something that doesn’t match the income statement. Depreciation and amortization looks steady, almost comforting. CapEx is meaningfully lower, or lurches higher, and the difference is too large to ignore if you’re underwriting free cash flow. The friction is that the 10‑K doesn’t hand you the answer in one place: the cash number is aggregated, the asset detail is scattered across footnotes, and management’s narrative is usually framed to defend whatever the number happens to be this year.
So the first pass becomes mechanical: circle the gap, jot the ratio, and note whether it’s a one-year move or a multi-year pattern. Even before you decide what it means, you’re already constrained by timing—CapEx is a cash outflow, D&A is an allocation—and by classification choices that can shove spend into acquisitions, leases, or “software.” The gap is real, but what it’s measuring is still unsettled.
The first instinct: “CapEx below D&A means trouble”
The reflex is to treat depreciation as a proxy for “maintenance CapEx,” so when CapEx prints below D&A the brain jumps to a simple story: they’re consuming the asset base faster than they’re replacing it. On a spreadsheet that’s tidy. Underwrite D&A as the reinvestment floor, haircut the reported free cash flow, and you’ve protected yourself from the rosy case. The problem is that this shortcut feels objective, so it tends to harden into a verdict before you’ve checked what’s actually being depreciated and how the business is changing.
It also fits the kind of risk you don’t want to miss: deferred plant repairs, aging fleets, systems that limp along until a forced catch-up year wrecks margins and cash conversion. If you own the stock through that, the “cheap” valuation wasn’t cheap; it was borrowed time. Still, even in the moment you can feel the constraint: one year of CapEx is a timing snapshot, while D&A is a blended run-rate across many vintages, policies, and acquisition layers. The instinct flags a real downside scenario, but it doesn’t tell you the probability yet.
When reality breaks the shortcut you wanted
Then you hit the first contradiction: the companies that “should” be in trouble keep printing stable service levels, stable gross margins, and no obvious operating friction. A distributor runs a mostly leased fleet and expensed repairs; D&A is still bloated from an old acquisition. A software-heavy business capitalizes implementation costs, so the depreciation line is partly amortization of intangibles that doesn’t map to replacing physical capacity. The shortcut starts to wobble because the accounting expense you’re using as a floor is often a blend of things that don’t fail on the same schedule.
The timing mismatch shows up fast once you line up a few years. CapEx can come in pulses (a plant expansion, a data-center refresh, a retail remodel cycle) while D&A smooths across the entire asset base. If management pulled spend forward last year to lock in equipment lead times, this year can look “underinvested” even if the operating assets are newer than the depreciation schedule implies. The constraint is practical: you’re trying to make a forward statement from backward allocations, and one period of gap can be nothing more than calendar math.
What usually breaks the spell is realizing you’re not actually asking, “Is CapEx less than D&A?” You’re asking, “What cash burden is required to keep revenue, unit economics, and compliance intact?” That answer sits in the boring places—maintenance backlog hints, capacity utilization, lease mix shifts, warranty trends—and it rarely collapses into a single proxy. The shortcut was useful as a warning light, but it isn’t the instrument panel.
Separate growth spend from keeping lights on
Once you’re thinking in “required cash burden,” the next constraint is motive: some of that CapEx is optional growth, and some of it is the toll to keep throughput, uptime, and regulatory posture from drifting. The filing rarely labels it cleanly, so you end up triangulating. If revenue is flat and management is still talking about “capacity adds,” that’s a hint the spend is aimed at mix, geography, or new product lines—not simply replacing worn assets. If volumes are up but unit costs and service levels are stable, the same dollars might be doing double duty: expansion plus refresh.
The practical move is to force a split, even if it’s crude. Start with what must be true for the current run-rate to hold: replacement cycles, store remodel cadence, fleet age targets, data-center refresh norms, environmental compliance. Then treat the remainder as growth CapEx and ask whether it’s earning: is incremental revenue showing up without a matching step-up in maintenance expense, working capital drag, or downtime? The gap stops being “underinvestment” and turns into a question about returns versus optionality.
Asset age and inflation can fake “underinvestment”
A different kind of doubt shows up when the business looks fine operationally, yet the math keeps screaming “CapEx is too low.” The reason can be boring: the asset base is older than the depreciation schedule suggests, or newer than it suggests, and either case can break the proxy. If a company ran a heavy build-out years ago, today’s D&A can be dominated by those legacy vintages even though the current replacement need is lighter. The opposite happens when useful lives are long on paper but the real cycle is shorter because of wear, safety standards, or uptime expectations.
Inflation adds another layer of optical distortion. Depreciation is based on historical cost, but replacement is paid at today’s prices. In a high-inflation equipment category, “maintenance CapEx equals D&A” becomes a moving target, and a flat CapEx line can quietly mean fewer units replaced. The constraint is that you can’t fix this with a single ratio; you have to check asset age signals—gross versus net PP&E, accumulated depreciation as a percent of gross, and any language about refresh cycles—then stress what replacement would cost at current price levels.
Acquisitions and accounting timing distort the picture
Then a deal drops into the timeline and the “gap” stops being about reinvestment and starts being about purchase accounting. The acquired assets often come in at fair value and get depreciated on a fresh schedule, so D&A can step up even if the acquired business already had its heavy replacement cycle behind it. Meanwhile, the cash outflow that created the capacity sits in “acquisitions, net of cash acquired,” not in CapEx, so the cash flow statement can show low CapEx alongside rising depreciation. The constraint is awkward: the economics may be fine, but the line items have stopped describing the same thing.
Timing inside a normal year can do similar damage. A large project placed in service late in the year can pull cash into the current period while contributing only a partial year of depreciation, making CapEx look “too high.” Flip it around and a lull in builds can make CapEx look “disciplined” even as depreciation keeps running from prior puts and takes. If the company is active in tuck-ins, the tell is that D&A moves with deal cadence while “CapEx” doesn’t; that’s when you stop judging a single year and start mapping spend and depreciation across the combined base.
Turn the gap into a range, not a verdict
By this point, the gap is still useful, just not as a binary signal. Treat it like a pricing band for reinvestment risk: a low case where true maintenance sits below D&A (older acquisition amortization, leased/expensed upkeep, lumpy refresh already done), and a high case where maintenance sits above D&A (replacement at inflated costs, shorter real lives, catch-up deferred by timing). The constraint is you rarely get enough disclosure to pin a point estimate, so pretending you do is the mistake.
So the model gets two free-cash-flow paths, not one. In the base case, hold reported CapEx but stress margins for minor upkeep creep; in the downside, lift CapEx toward a replacement-cost view and assume some working-capital drag when projects finally run. If the valuation only works in the low case, it’s not “cheap,” it’s conditionally cheap, and now you can decide whether that condition is actually knowable.
