A fabless semiconductor company designs chips but owns no fabrication plant; it pays a third-party foundry to manufacture the wafers and pays separate subcontractors to package and test the finished devices. A foundry is the manufacturing partner that owns the fabs and sells wafer capacity to fabless customers. The two models split the chip value chain: the fabless firm carries design, intellectual property, and demand risk but little factory depreciation, while the foundry carries the multi-billion-dollar capital base and the utilization risk that comes with it.

The model is visible directly in the filings. NVIDIA (NVDA), a fabless designer, states in its fiscal 2025 Form 10-K that it relies on outside foundries and packaging — it does not own a fab — and discloses the specific manufacturing partners it depends on. That reliance is both the source of the fabless model's capital efficiency and the reason supply concentration appears as a risk factor in the same document.

"We utilize foundries, such as Taiwan Semiconductor Manufacturing Company Limited, or TSMC, and Samsung Electronics Co., Ltd., or Samsung, to produce our semiconductor wafers. We purchase memory from SK Hynix Inc., Micron Technology, Inc., and Samsung. We utilize CoWoS technology for semiconductor packaging."— NVIDIA FY2025 Form 10-K, SEC, source

How the economics differ

For the fabless company, manufacturing is largely a variable cost: it pays a per-wafer price and an advanced-packaging fee, so its cost of goods sold rises and falls with volume rather than with idle-factory depreciation. Its balance sheet is lighter, its capital expenditure is modest, and its margin lives in design value and pricing power. The corresponding exposure is supply: the fabless firm depends on foundry capacity allocations, foundry price increases, and packaging bottlenecks, all of which it must disclose as risk factors. For the foundry, manufacturing is largely a fixed cost: it has financed the fab and depreciates it regardless of how full it runs, so its gross margin is driven by utilization and wafer pricing, and its capex cycle is the central financial event of the business.

An integrated device manufacturer (IDM) such as Intel (INTC) sits between the two, designing and manufacturing in-house and therefore carrying both the design risk of a fabless firm and the fixed-cost depreciation of a foundry. The contemporary twist is the foundry-services model, in which an IDM also sells spare capacity to outside customers, blending the two structures in a single set of financial statements and segment disclosures.

Why the split matters for the filings

The model determines where the money and the risk show up. A fabless company's financial statements feature low property, plant and equipment, large inventory-purchase and capacity commitments to foundries, and customer- and supplier-concentration risk factors. A foundry's statements feature enormous property, plant and equipment, heavy depreciation in cost of goods sold, and capex and utilization as the dominant variables. Reg S-K Item 101's business description and Item 105's risk factors are where each company is required to disclose its manufacturing dependence, so the cleanest way to confirm whether a company is fabless, a foundry, or an IDM is to read those sections rather than to infer it from product marketing.

Where the value chain splits the profit

The foundry-versus-fabless split also determines how profit is distributed along the chain. A fabless designer captures the value of the architecture and the software ecosystem around it, which is where pricing power tends to concentrate; the foundry captures the value of being able to manufacture at the leading edge, which is scarce and capital-intensive. When leading-edge capacity is tight, the foundry's bargaining position strengthens and wafer pricing can rise, which the fabless customer must absorb or pass on. When capacity is ample, the fabless designer's leverage improves. The same dynamic plays out in advanced packaging, which has become a distinct bottleneck: a fabless company that needs scarce packaging capacity discloses that dependence the same way it discloses foundry dependence, and the packaging supplier captures value accordingly.

These structural differences are why the two models read so differently in a downturn. A fabless company facing weak demand can reduce wafer starts and let much of its manufacturing cost fall with volume, but it may still be exposed through the purchase commitments it made to secure capacity. A foundry facing weak demand cannot shed its depreciation; its fixed asset base keeps generating cost regardless of how full the fabs run, so utilization becomes the dominant driver of its margin. An IDM carries both pressures at once. Reading a company's property footnote — large for a foundry or IDM, small for a fabless firm — alongside its commitment and risk-factor disclosures is the most direct way to confirm which model a company runs and how a cycle will move its results.

Neither model is presented as superior in the disclosure; each carries a different risk profile that the company must state. The fabless firm trades capital efficiency for supply dependence; the foundry trades capital intensity for the leverage of owning the bottleneck. For a reader, the durable takeaway is that the income statement, the property footnote, and the risk-factor section together identify the model, and the model in turn predicts which line items will swing the company's results.

What the record shows: a fabless company designs chips and outsources manufacturing to a third-party foundry, carrying light property but large foundry purchase commitments and supply-concentration risk; a foundry owns the fabs, carrying heavy depreciation and utilization risk; an IDM such as Intel carries both. NVIDIA's fiscal 2025 Form 10-K names its foundry and packaging dependence directly, and Item 101 and Item 105 of Regulation S-K are where each company is required to disclose which model it runs.