Demand for artificial-intelligence chips is booming. Unfortunately for those who make semiconductor-manufacturing gear, that isn't always enough.
That dichotomy has become extremely clear this week. Taiwan Semiconductor Manufacturing posted strong third-quarter results Thursday and said it expects revenue from "server AI processors" that include Nvidia's red-hot chips to more than triple this year.
That was preceded by a rather bleak report from ASML Holding NV. The Dutch equipment maker said net bookings that represent orders were 2.6 billion euros ($2.8 billion) for the third quarter -- less than half of what Wall Street was expecting and a major miss for a company whose lithography gear is a must-have for the most advanced chip manufacturing. The company also said it expects 2025 revenue to come in at the low end of its previous forecast.
The dueling reports resulted in whiplash for investors. Stocks of major U.S. equipment makers Applied Materials, Lam Research and KLA-Tencor slid between 13% and 18% over the two days since ASML's report was accidentally released Tuesday morning -- a day earlier than planned. ASML itself shed 20% of its value over that time. TSMC's results revived chip stocks to an extent on Thursday, sending the PHLX Semiconductor Index up about 1% in Thursday trading.
Analysts still widely expect total sales of chip-manufacturing gear to top $100 billion globally this year for the first time ever. But it is a business dominated by a few big spenders -- TSMC said Thursday that its capital expenditures will be slightly higher than $30 billion this year and "very likely" higher next year.
The other two big spenders are struggling. Intel's market share in data centers and PCs has been sliding just as it attempts an ambitious turnaround plan, forcing the storied chip giant to slash capital spending to conserve cash. Wall Street expects Intel's capex to fall 8% this year and 15% next year, according to consensus estimates on Visible Alpha.
And Samsung actually apologized to investors last week for its own missteps -- in which the world's largest memory-chip maker has fallen behind its rivals in the specialized type of memory needed for AI. Analysts are expecting a 4% drop in Samsung's chip-related capex this year, though it is expected to bounce back to 2023 levels next year, according to Visible Alpha's estimates. Samsung may give further updates on its spending plans when it reports its full third-quarter results later this month.
And then there's China, which has been spending ferociously to build up its domestic chip industry as the U.S. tries to limit the country's access to the most advanced chips and manufacturing gear. Sales to China have historically accounted for 14% to 18% of ASML's annual system sales but jumped to 29% last year and have averaged 48% in the first three quarters of this year. Analysts have long worried that this burst in spending wasn't sustainable, and indeed it isn't.
Christophe Fouquet, ASML's chief executive officer, said on the company's earnings call Wednesday that "we expect the China business to go back to a more normalized percentage of our business," with China projected to account for 20% of the company's revenue next year.
Investors will thus still need to exercise caution -- no easy feat in a market still dominated by AI hype. TSMC's report Thursday predictably gave a further boost to stocks like NVIDIA Corp, Micron Technology and Broadcom that have been benefiting directly from AI demand.
But the market is more complicated for equipment makers, whose customers also have to contend with slowing PC demand, uneven smartphone sales and a sluggish auto market. Applied, Lam and KLA saw their stock prices more than double between the launch of ChatGPT in late 2022 and July 10 of this year -- when all three set record highs. And yet, their combined trailing 12-month revenue actually fell 9% over the same period, according to data from S&P Global Market Intelligence.
Even after this week's drubbing, those stocks are still averaging a premium of 10% to their average price-to-forward earnings ratios over the last five years, according to FactSet data.
"The mistake that many investors and inexperienced analysts have made is to take the huge, rip-roaring success of AI and translate it as an indicator of the entirety of the chip market," Robert Maire of Semiconductor Advisors noted Wednesday.
Nvidia's rising tide doesn't lift all boats equally.
Comments