Semiconductor Comparison Using Two Exact Annual Filings

The conclusion comes after the business-model adjustment

NVIDIA and Intel both sell semiconductor products, but they do not own the same production system. NVIDIA primarily designs accelerated-computing platforms and relies on foundries and other suppliers to manufacture, assemble, test, and package its products. Intel is an integrated device manufacturer that owns a large fabrication network while also developing a foundry business. The accounting follows that economic difference: NVIDIA's manufacturing capacity is largely expressed through inventory, purchase commitments, and supplier dependencies, while Intel reports far more property, plant, and equipment, depreciation, construction spending, and factory-utilization risk.

This is therefore a comparison of two business systems, not a declaration that every dollar of gross margin or capital expenditure is interchangeable. It is useful for learning which adjustments a semiconductor comparison needs. It is not, by itself, an investment recommendation or a current valuation. Market prices, diluted share counts, and forecasts are intentionally excluded.

1. Lock the filings, periods, and units

NVIDIA's fiscal 2025 ended January 26, 2025. Intel's 2024 fiscal year ended December 28, 2024, only twenty-nine days earlier. That is much closer than comparing two filings with six-month-apart year ends, but the companies still use different fiscal labels and week-based calendars. All figures below are USD millions from the respective filings unless stated otherwise.

The worksheet is reproducible. NVIDIA's operating margin is $81,453m / $130,497m = 62.4%, and Intel's is -$11,678m / $53,101m = -22.0%. Simple free cash flow is $64,089m - $3,236m = $60,853m for NVIDIA and $8,288m - $23,944m = -$15,656m for Intel. Intel's capital spending was $23,944m / $53,101m = 45.1% of revenue, versus 2.5% for NVIDIA.

2. A fabless model still carries manufacturing commitments

NVIDIA's low reported property additions do not mean manufacturing capacity is free or riskless. Its filing disclosed approximately $30.8 billion of inventory purchase and long-term supply and capacity obligations. Those commitments include components, manufacturing capacity, and related supply arrangements. They are economically important even though they do not appear as NVIDIA-owned fabrication plants on the balance sheet.

An analyst should therefore pair the 2.5% capex ratio with supplier concentration, prepaid supply, inventory, noncancelable commitments, advanced-packaging availability, and the time needed to qualify manufacturing changes. Outsourcing converts part of fixed asset ownership into contractual and supply-chain exposure. It does not eliminate capacity risk.

3. Intel's owned-factory cash spending needs a funding bridge

Intel's $23,944 million of property, plant, and equipment additions cannot be analyzed only as an expense against the current year's revenue. New fabrication assets can support future internal products and external foundry customers, but construction timing, process yields, utilization, and depreciation determine the eventual return. The balance sheet also shows how much productive capital remains exposed if demand or execution falls short.

Intel reported $12,714 million of partner contributions and $1,936 million of government incentives in the investing section of its cash-flow statement. Those inflows help fund expansion, but they do not turn gross capital additions into ordinary operating cash flow. A good comparison presents gross additions, partner or government funding, and Intel's retained economic obligations separately rather than netting everything into one opaque capex number.

4. Separate operating economics from disclosed reset costs

Intel's reported loss contains information that a normalized case should identify without pretending the costs never happened. The filing discusses approximately $3.3 billion of impairment and accelerated depreciation charges and $922 million of Gaudi inventory-related charges. A scenario may show results before selected items to study ongoing earning power, but cash consequences, repeated restructuring, and future depreciation still belong in the investment record.

NVIDIA's 75.0% gross margin also needs a scenario rather than a permanent label. Its mix was heavily influenced by data-center demand and accelerated-computing products. Product transitions, competition, export controls, supplier costs, customer concentration, and inventory provisions can change margin. The honest comparison includes a base case and downside case for each company instead of normalizing Intel upward while freezing NVIDIA at a peak reported margin.

5. Use different denominator tests before valuation

NVIDIA reported $72,880 million of net income, while Intel reported an $18,756 million net loss attributable to Intel. A conventional P/E comparison is therefore not meaningful: Intel's denominator is negative. Substituting enterprise value to sales would make both denominators positive, but it would ignore their enormous margin and capital-intensity gap. Enterprise value to current operating profit also fails for Intel until the analyst builds and defends a normalized profit.

A valuation worksheet should first model revenue by segment, mid-cycle gross margin, operating expenses, depreciation, taxes, working capital, gross capital needs, and dilution. Then compare market values measured on the same date. For Intel, include a range of utilization and foundry outcomes. For NVIDIA, include a range of demand, mix, supplier cost, and export-control outcomes. If a conclusion depends on one precise normalized margin, show how quickly it changes under nearby assumptions.

6. A reusable semiconductor comparison checklist

• Use exact filings and show each fiscal year-end rather than relying on a vendor's “2024” label.
• Classify each company as fabless, foundry, integrated manufacturer, memory producer, equipment supplier, or licensor.
• Reconcile owned capex with leases, prepayments, purchase commitments, partner funding, and government incentives.
• Compare inventory composition and provisions, not just inventory divided by revenue.
• Separate current product mix from a defensible mid-cycle margin range.
• Test customer concentration, geographic restrictions, suppliers, process transitions, and packaging constraints.
• Use one market-price date and a denominator that remains economically meaningful for every peer.
• Keep reported results visible beside every adjusted or normalized scenario.

The apparent 42.6-point difference in reported operating margin is a starting observation. The investment question is how much of that gap persists after demand, mix, utilization, production ownership, and required reinvestment are modeled consistently.