The semiconductor capital-expenditure (capex) cycle is the pattern of large, lumpy investment in fabrication plants, equipment, and advanced-packaging capacity that chip makers commit years before the revenue those facilities will generate. Capex appears in a company's cash-flow statement as “purchases of property and equipment” (or additions to property, plant and equipment), and it is the single clearest disclosed signal of how much future demand a manufacturer is building toward. Because a leading-edge fab costs tens of billions of dollars and takes years to equip and ramp, capex is committed in advance and then depreciated over many years, which is why it drives both factory utilization and the fixed-cost base that sets gross margin.

The disclosure backbone is Item 303 of Regulation S-K, which requires registrants to analyze, in the liquidity and capital resources section of MD&A, their material cash requirements from known contractual and other obligations — specifying the type of obligation and the time period — and to describe known trends and uncertainties in capital resources. Capacity-expansion plans, equipment commitments, and construction-in-progress are surfaced there and in the property footnotes, not in press releases alone.

"Identify any known trends or any known demands, commitments, events or uncertainties that will result in or that are reasonably likely to result in the registrant's liquidity increasing or decreasing in any material way."— 17 CFR 229.303 (Item 303), eCFR, source

Why the cycle is lumpy

Capex in chips is cyclical because capacity decisions are discrete: a foundry either builds a new fab line or it does not, and the increment is enormous. When AI-accelerator or memory demand runs ahead of supply, manufacturers raise capex sharply; when a glut forms, they cut it just as sharply, and the depreciation from the prior build keeps flowing through cost of goods sold even as utilization falls. That timing mismatch — capex committed at the top of demand, depreciation peaking into a downturn — is the mechanism behind the sector's margin swings. Taiwan Semiconductor Manufacturing Company (TSM), in its fiscal 2024 Form 20-F filed April 17, 2025, frames its capital expenditures as central to its forward plans, noting that more than half of its capital expenditures are denominated in currencies other than the NT dollar, which adds a foreign-exchange layer to an already large spend.

Reading capex from the filings

The primary figure is the investing-activities line of the cash-flow statement; the property, plant and equipment footnote then breaks the asset base into land, buildings, machinery, and construction-in-progress, the last of which signals capacity still being built. Rising construction-in-progress with rising purchase commitments is the disclosed shape of an up-cycle. The MD&A is where management is required to tie those numbers to demand and liquidity. A reader should pair the cash-flow capex line with the contractual-obligations and purchase-commitment disclosures to see the full forward cash claim, because much capacity is locked in through multi-year equipment and supply commitments that sit beside reported capex rather than inside it.

Capex versus depreciation, and the lag between them

The most useful way to read the capex cycle is against depreciation, because the two move on different clocks. Capex is recognized when the company buys and installs equipment; depreciation of that equipment is then spread across its useful life and flows through cost of goods sold for years afterward. In an up-cycle, capex can run far above depreciation as a company builds capacity; in a down-cycle, capex falls quickly while depreciation from the prior build stays elevated, weighing on gross margin even as new spending stops. A reader comparing the investing-activities capex line with the depreciation disclosed in the cash-flow statement and property footnote can see where a company sits in that cycle: capex well above depreciation signals expansion, capex below depreciation signals a company harvesting an existing base rather than growing it.

The disclosure also separates maintenance capex from growth capex in substance even when the filing does not label them. Some spending replaces worn or obsolete equipment to keep existing capacity running; some adds genuinely new capacity or a new node. Construction-in-progress and disclosed capacity-expansion plans point to the growth portion, while steady replacement spending tracks the installed base. For semiconductor companies, the node transition adds a further wrinkle: leading-edge equipment for a new process is far more expensive than the prior generation, so a flat unit-capacity plan can still require rising capex simply to advance a node. Reading the MD&A's stated rationale alongside the numbers is the only way to tell which kind of spending a given year's capex represents.

Capex is a confession of expected demand, but the filing states what was spent and committed, not what the facilities will earn. The disclosure shows direction — expansion or restraint — and the obligations behind it; it does not guarantee the revenue that motivated the build. For analysts, the durable reading is to track the investing-line capex alongside construction-in-progress and disclosed purchase commitments over several periods, because the cycle only becomes visible across quarters, and the depreciation it creates outlives the demand that triggered it.

What the record shows: semiconductor capex is the purchases-of-property-and-equipment line of the cash-flow statement, committed years ahead of the revenue it serves and depreciated long afterward; Item 303 of Regulation S-K requires registrants to analyze the related material cash requirements and capital-resource trends in MD&A; and the cycle is read by comparing capex with depreciation, construction-in-progress, and disclosed purchase commitments across several periods rather than from any single quarter.