By Jack Hough

The U.S. stock market wouldn't be riding so high without chip designer Nvidia, which is at the center of a flood of spending on artificial intelligence. Nvidia, in turn, is deeply dependent on Taiwan Semiconductor Manufacturing, which runs the world's most sophisticated chip factories. And Taiwan Semi would be lost in cutting-edge chips without a Dutch company called ASML Holding, the leader in lithography, or using lasers and mirrors to print impossibly dense patterns on silicon wafers.

All three companies have done wonders for investors over the past decade: Nvidia shares have returned 29,000%; Taiwan Semi's American depositary receipts, 1,100%; and ASML's ADRs (which trade under the ticker ASML), 800%. But Nvidia and Taiwan Semi are trading close to their highs. ASML, after slashing its revenue guidance in October, is down by nearly a third from its peak last summer.

Is this an early warning for chips, and by extension, the U.S. stock market? Or is ASML losing its monopoly hold on high-end lithography? Bet neither -- and that ASML will bounce back.

For background, start with Roy G. Biv, the old grade-school anagram for the colors of the rainbow, from red to violet. Colors are determined by wavelengths, and these are the ones that human eyes are tuned to see, from 700 nanometers down to 380, but there are others. For example, beyond red is the infrared used by remote controls, and beyond that are broadcast signals, because long waves travel far. Old-fashioned television, and now cellphones, can use waves as long as meatball heroes, and AM radio waves can be the size of football fields.

In the other direction, past violet, are extremely small waves that are handy for passing through solids, like the ultraviolet used in tanning beds, and below 10 nm, the X-rays that can reveal broken bones. This is the neighborhood where ASML plays. Its current extreme ultraviolet, or EUV, machines can print 13-nm resolutions, which chip makers can turn into even smaller features to achieve their current 7-nm, 5-nm, and 3-nm "nodes," or generations. A nanometer is roughly how long a fingernail grows in a second; a human hair is 80,000 to 100,000 nanometers wide.

Transistor density tends to double every two years or so with a minimal increase in cost, making computers ever more powerful -- a trend known as Moore's law, and kept alive today by ASML. Its room-size machines are the most complex thing humanity has ever created, says Didier Scemama, who has studied ASML for decades and covers it for BofA Securities. For example, they fire lasers at molten tin droplets falling at a rate of 40,000 to 50,000 a second -- two hits per droplet, without a miss. A stray particle could ruin a wafer, or take down the machine.

That technology took decades to develop. The first EUV test chips were made at SUNY Albany in 2008. The first commercial product to use EUV-enabled chips was Samsung Electronics' family of Galaxy Note 10 smartphones in 2019. China, which is banned from buying ASML's most sophisticated machines, would surely like to make its own. "I think the odds of them being very successful in EUV over the course of the next 10, 15, 20 years are fairly low," says Scemama. Beyond complexity, there are ASML's supply deals with Germany's Carl Zeiss, which makes the world's flattest mirrors, and Trumpf, which makes the world's most powerful pulsed industrial lasers.

ASML's current top machines are estimated to cost more than $220 million. The next generation, called high numerical aperture, or NA EUV, can print 8-nm resolutions, supporting nodes starting at 2 nm. Intel is an early recipient. Broader uptake could start in a year or two, at prices estimated at $380 million to $400 million per machine. Apple, which first brought 3-nm chips to market in its top 2023 iPhone, could move to 2-nm production around next Christmas. It currently leads the miniaturization push; Nvidia is content to lag a nanometer or two behind, while differentiating its chips with software and other add-ons.

Back to our original question. Why is ASML slumping? Three reasons. First, China has rushed to stockpile lower-end ASML equipment ahead of what might be tighter trade restrictions starting soon. The shift back to lower demand from China will temporarily sap growth. Second, Intel made a bold push to expand in chip manufacturing, with little financial success so far. Intel and Samsung have delayed equipment orders for their foundries. Third, despite strong growth in AI, there is a slump in consumer devices and their chips. Two-thirds of the companies in the Philadelphia Semiconductor index have trailed behind the S&P 500 index over the past year.

A return to growth for ASML seems likely. Timing will depend on demand for phones, personal computers, cars, and gaming machines -- as well as whether Intel and Samsung are able to lure Taiwan Semi customers to their foundries. One bear case is that lithography intensity is rising more slowly than in the past. Taiwan Semi's 5-nm node uses 10 to 12 EUV layers, and its 3-nm node, about 20. But with its 2-nm node, this is believed to rise to only 22 layers. On the other hand, older nodes will remain relevant for many years; the shift to high NA EUV is coming; and computer memory is following the same rising EUV demand path as processors, adding to growth.

BofA's Scemama reckons that earnings for ASML will grow at a compounded average of 16% a year over the next five years. Shares trade at 37 times this year's slashed earnings estimate -- a premium, but one that is close to a 10-year low relative to U.S. chip-equipment companies, J.P. Morgan points out. My colleague Andrew Bary recently named ASML one of his top 10 stock picks for 2025. And Scemama's price target for the ADRs implies 26% upside.

Write to Jack Hough at jack.hough@barrons.com Follow him on X Subscribe to his Barron's Streetwise podcast

To subscribe to Barron's, visit http://www.barrons.com/subscribe

(END) Dow Jones Newswires

January 10, 2025 21:30 ET (02:30 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.