Semiconductor stocks compared: Nvidia, Broadcom, AMD, Micron, Intel and more

"Semiconductors" covers businesses as different as a software platform and a commodity factory. Some companies design chips and pay others to make them; one designs and manufactures its own; one makes memory, which trades almost like a commodity; one builds servers around other people's chips. Put seven of them side by side on the same filings data, as this study does, and the industry stops looking like one trade and starts looking like seven different bets. All figures are trailing twelve months to each company's latest filing, with valuations at the 14 October 2026 close.

The numbers side by side

By market value the group is just as uneven: Nvidia at $5,333.3bn, Broadcom $1,652.9bn, Micron $1,034.3bn, AMD $931.8bn, Intel $494.9bn, Qualcomm $181.3bn and Super Micro $23.5bn.

Who captures the money: designers, makers and assemblers

Nvidia designs graphics and AI processors and the software ecosystem around them, and outsources manufacturing. Revenue growth of 100.05% a year for three years — roughly eight-fold in total — combined with a 65.21% operating margin is a pattern almost without precedent at this scale. It is what happens when demand runs far ahead of any competitor's ability to supply.

Broadcom earns a 48.05% operating margin from a mix of networking chips, custom processors designed for individual cloud customers, and infrastructure software. That software business, built largely by acquisition, is part of why its margins look more like a software company's than a chipmaker's.

AMD, also a designer, competes with Intel in processors and with Nvidia in AI accelerators. Its 15.71% operating margin and 8.49% ROIC are far below Nvidia's: it is earning a modest return while investing to win share.

Intel both designs and manufactures its chips, and is spending heavily to rebuild its manufacturing lead and sell factory capacity to other designers. That strategy carries the cost of the factories before the revenue arrives: an operating margin of −0.14% on revenue that fell 5.71% a year. Our AMD vs Intel comparison shows the contrast between the two models directly.

Qualcomm sells smartphone processors and licenses wireless patents. It is profitable — a 23.19% operating margin and ROIC — but revenue has been flat for three years, growing 0.06% a year.

Super Micro assembles servers around Nvidia's and AMD's chips. It grew revenue 76.34% a year by riding the same AI wave, but a gross margin of 10.82% leaves little room: it captures volume, not pricing power. Fast growth also tied up cash in inventory, leaving free cash flow negative.

Micron and the memory cycle

Micron is the most surprising line in the table. A memory maker earning a 74.59% operating margin — higher than Nvidia's — and an 80.45% ROIC would have been unthinkable a few years ago. Memory chips are close to a commodity: when supply is short, prices soar, and when new factories come online, prices fall. Micron lost money in its 2023 fiscal year; it is now in a strong phase of the cycle, driven by demand for the high-bandwidth memory that AI accelerators need.

That is why Micron's P/E of 12.3 and EV/EBITDA of 9.2 are the lowest in the table after Super Micro's 10.98 and 7.31. A multiple that low on earnings like these is commonly read as investors doubting that the earnings will last at this level. With cyclical companies, the multiple tends to be lowest when profits are highest — see the cyclical trap in our P/E guide. Our Nvidia vs Micron page ties the two on quality: by our rules their returns are too close to separate.

What the valuations are betting on

The spread of multiples says the market sees very different futures:

• AMD at 146.36 times earnings is priced for a much larger share of the AI market than it holds today. Its free cash flow yield of 0.9% offers little cushion if that share does not arrive.
• Nvidia at 27.98 times earnings is cheaper than AMD on earnings, sales and cash-flow multiples despite far higher margins and growth; only on price-to-book, a weak measure for chip designers, does AMD look cheaper. The market's question about Nvidia is not whether today's earnings are real but how long growth of this kind can continue.
• Broadcom at 44.16 is valued for the durability of its custom-chip and software revenue.
• Intel has no meaningful P/E, but trades at 8.68 times sales — a valuation that rests on its turnaround and manufacturing plans rather than current profits.
• Qualcomm is the cash-return case: an EV/EBITDA of 16.11 and a 5.75% free cash flow yield, with a 2.08% dividend yield and 5.1% buyback yield on top. It is priced for little growth.

How the head-to-heads come out

Our six-factor scores reflect all this. Nvidia beats AMD 5.5 to 0.5 — every factor except momentum, which is not scored — and beats Broadcom and Intel 4.5 to 1.5 each. AMD loses to Qualcomm 2.5 to 3.5: Qualcomm is cheaper, earns higher returns and hands back far more cash, while AMD wins only on growth and balance sheet. Explore the full set: Nvidia vs AMD, Nvidia vs Broadcom, Nvidia vs Super Micro, AMD vs Qualcomm, Broadcom vs Micron and Intel vs Micron.

The risks common to all seven

Three risks run through the whole group. First, concentration: a large share of current AI-related revenue comes from a small number of very large cloud and internet companies, whose spending plans can change quickly — see our big tech study for what that spending is doing to their own cash flow. Second, cyclicality: semiconductors have always moved in booms and busts, and the current boom is unusually concentrated. Third, geopolitics: export controls and the concentration of advanced manufacturing in a few locations affect every company here. None of these shows up in a trailing twelve-month figure, which is precisely why the figures need reading with them in mind. For how margins and returns compare with other industries, see operating margin by business model.