Intel Or NVIDIA: Who Is Paying For The AI Build-Out?

If you own Intel (INTC) or NVIDIA, you own the same shortage. Both said the same thing at their latest reports: buyers want more compute than either can ship, and supply rather than demand is the limit. The fork is what each does about it. NVIDIA sells into the shortage and keeps most of what it collects. Intel has to build the capacity first, and it is paying for that now.

Which One Guided A Year Of Revenue, And Which Years Of Spending?

NVIDIA guided the money coming in. Management put fiscal 2028 revenue growth at roughly 70%, called that outlook supply-constrained, and said customer forecasts point to demand doubling. Supply is the ceiling, so that revenue is close to spoken for.

Intel guided the money going out. It lifted 2026 capital spending to more than $20 billion and said 2027 would run significantly above that. Weeks later it sold $20 billion of common stock. Intel’s trailing twelve-month revenue was about $57 billion, so one year of capital spending at that scale is committed years before the capacity earns anything.

How Does NVIDIA Get Paid For The Same Shortage?

NVIDIA sells the whole rack. Each gigawatt of AI data center it equips is worth roughly $40 billion to NVIDIA on the Vera Rubin generation, against $25 billion on Blackwell and about $18 billion in the Hopper era, because the build now buys its CPUs and networking too. That is why the same demand converts so differently: NVIDIA runs an operating margin of 65.2% against 7.6% at Intel, and a net margin of 63.7% against -19.8%.

That position has a rising price. NVIDIA guided gross margin down from 75% in fiscal Q2 2027 to a trough in the 71% to 72% range in fiscal Q4 2027, blaming extreme pricing conditions in memory, before settling at 72% to 73% in fiscal 2028. It has also put nearly $50 billion into the frontier AI labs that buy its compute, which management admits some will call circular financing. So NVIDIA is not only collecting from the build-out. It is helping fund the buyers.

Does Intel’s Cheaper Multiple Settle It?

Intel screens cheaper on exactly one measure, valuation, at a price-to-EBIT multiple of -52.2 against 23.6 for NVIDIA. A negative multiple is not a discount. It is a negative denominator: Intel’s EBIT margin over the trailing twelve months is -15.3%.

NVIDIA takes every other measure: growth, both margins, and leverage. Intel’s counterweight is real. Its AI-driven businesses grew more than 70% year over year in fiscal Q2 2026 and made up about 70% of revenue, management calls Xeon 6 one of the fastest-ramping products in Intel’s history, and the stock returned 315.6% over the past twelve months against 23.4% for NVIDIA, though it is down 12.0% over the past three months.

But Intel Foundry, the reason for the spending, sold just $293 million to outside customers in fiscal Q2 2026 and lost $2.1 billion, $348 million less than the quarter before. The business evidence here favors NVIDIA clearly, and a multiple you cannot read is no argument against it. Both can be scored on the same five measures: growth, profitability, stability, resilience and valuation

How Do You Actually Decide Between These Two?

Intel is rebuilding and NVIDIA is harvesting, and both can stay true for years. Two things help. Put Intel and NVIDIA on one page with the rest of the chip sector before you choose. Then stop making this call one pair at a time, which is what the Trefis High Quality Portfolio is for. That portfolio has a track record of outpacing the three major indices.