The semiconductor industry giant ASML, known for its "pinnacle of human technology," has experienced a rare earnings miss, triggering a global chip stock plunge. ASML's lackluster performance has doused the recent rebound of chip stocks from the "summer selling wave" with cold water. According to incomplete statistics, since Tuesday's opening of the US stock market until the close of the Japanese and Korean stock markets on Wednesday, the total market value of global chip stocks, including Nvidia, AMD, SK Hynix, and Samsung, has evaporated by more than $420 billion.
The process of ASML's earnings release was nothing short of dramatic. ASML was originally scheduled to release its third-quarter financial report on Wednesday, but due to a "technical glitch," the report was unexpectedly posted on the company's official website during the early morning hours of Tuesday's US stock market, a "bone-breaking" embarrassment. For a tech giant that controls the core lifeline of global chip production, such a significant error is a historically rare phenomenon.
Advertisement
Consequently, ASML's exceptionally poor performance has shocked global investors: although overall revenue exceeded expectations, the order size was only half of the market's expectations, while also lowering next year's sales target and gross margin guidance beyond expectations. In this "accidentally leaked" third-quarter financial report, ASML stated that the Q3 order size totaled only €2.6 billion, nearly half of the market's general expectation of €5.4 billion. At the same time, the company expects that by 2025, total net sales will grow to between €30 billion and €35 billion, however, this is in contrast to the previous quarter's guidance of €30-€40 billion. This latest forecast number is only in the lower half of the guidance provided at the 2022 Investor Day, thus the sales outlook can be said to have been significantly reduced by ASML's management.
After the financial report was released, ASML's US ADR fell by more than 17% at one point, eventually closing down 16.26%, at $730.43. The company's stock price on the Amsterdam stock market closed down 16%, marking the largest single-day drop since 1998, with multiple trading halts during the day. After ASML's stock price plummeted, the "highest market value technology company" in the European stock market also changed hands, with the long-dominant lithography giant ASML passing the title to the German software powerhouse SAP, which has greatly benefited from the AI boom.
In addition to the earnings miss, there are reports that the US government is considering restricting US companies from selling the most advanced AI chips to certain Middle Eastern countries, including the flagship data center AI chips from Nvidia and AMD, which are in high demand. This news further propelled ASML's stock price to plummet, as the demand for AI chips from chip giants like Nvidia may decline as a result, potentially leading TSMC to reduce the production capacity of Nvidia and other AI chips, and ASML's performance expectations have also been downgraded based on this logic.
However, from a rational investment perspective, ASML's earnings report, which has hit global chip stock prices, does not mean that the global craze for artificial intelligence is receding or cooling down. But this explosive financial report does reveal the latest dynamics of the global chip industry: the AI boom is still ongoing, especially the demand for all types of AI chips focused on the B-side data centers remains very strong, but areas unrelated to AI are still in a state of weak demand or even a significant decline in demand.
ASML's Chief Financial Officer, Roger Dassen, supported this market view in his performance statement. The ASML executive stated that the demand for AI-related chips is indeed surging, but the recovery process for other parts of the semiconductor market is weaker than expected, leading some logic chip manufacturers to postpone lithography orders.
Before ASML's earnings miss, there has long been an optimistic bullish view in the stock market that a new round of the entire chip industry's "prosperity cycle" has quietly begun in 2024. However, ASML's latest performance has made this bullish view highly questioned by the market, thereby promoting the sale of chip stocks. That is to say, the AI boom laid out by global enterprises and even global government agencies has not faded, but this AI boom has not yet driven the entire chip industry, including long-term weak demand for industrial chip, electric vehicle chip, power chip, components, and most consumer electronics core chips including smartphones, into a "prosperous development cycle."
After ASML's earnings miss, some analysts even exclaimed that even if ASML's performance is so ugly, it is actually good for the stock price trend of "AI shovel sellers" like Nvidia, after all, ASML's performance can show that the demand for data center AI chips is still strong. Looking at the stock price trend, ASML, which holds the lifeline of chip production capacity, has significantly underperformed Nvidia, and unknowingly, the market has already given an answer with real money on who is the biggest winner in chip stocks.
ASML's downward revision of expectations may mean a slight overcapacity in chip production, but it is by no means the end of the chip industry.For ASML's weak performance report, chip industry analysts have expressed that the weak performance outlook may reflect some overcapacity in chip factories. Large chip foundries, including TSMC, Intel, and Samsung, have accumulated a large number of expensive ASML EUV tools during the COVID-19 pandemic. These foundry giants are good at transforming and upgrading these tools to produce more high-end AI chips.
At present, in addition to AI chips, the inventory of other types of chips remains high. At the same time, after years of manufacturing process exploration, chip manufacturers such as TSMC have greatly improved the efficiency of using ASML photolithography machines, which also means they can produce more chips and are not in a hurry to order more photolithography equipment.
For example, due to the lack of significant increase in demand for smartphones and PCs, the "king of chip foundries" TSMC may transform and upgrade some EUVs for foundry 5nm and below smartphones and PC chips to increase the production capacity of NVIDIA's "endless" demand for H100/H200 AI GPUs, the latest Blackwell architecture AI GPUs, and AMD MI300 series GPUs, instead of choosing to continue to invest heavily in expensive EUVs.
Of course, from 2023 to 2024, during the extremely hot demand for NVIDIA AI GPUs, ASML's financial reports for several quarters can show that TSMC is continuously purchasing new EUVs, but the purchase scale is not as large as some analysts predicted, and the purchase pace may tend to stagnate in the third quarter. This fully highlights that as the demand for non-AI chips continues to decline, foundries such as TSMC will use the excessive photolithography inventory equipment accumulated since the epidemic for manufacturing high-end AI chips after transformation and upgrading, largely reflecting that the AI boom has not driven the entire chip industry into a growth cycle, and the chip factory's production capacity is slightly excessive.
Therefore, chip foundries such as TSMC can only order new EUV or DUV photolithography machines when the factory orders are full. Some analysts also said that ASML's latest forecast is a lagging indicator of the poor performance of these chip factories in recent months, revealing the latest situation of the current chip market's non-AI demand decline.
"Intel, TSMC, and Samsung may be withdrawing photolithography machine orders from ASML because they have realized that the production capacity is very sufficient." Dan Hutcheson, Vice Chairman of TechInsights, a technology-focused analysis company, said in the latest interview.
"The utilization rate of chip factories this year is about 81%, but manufacturers tend to purchase new photolithography tools when it reaches about 90%. In addition, Intel has slowed down the expansion of its U.S. factories, and Samsung and TSMC have also been cautious about chip capacity expansion recently." Hutcheson said.
Handel Jones, CEO of International Business Strategies, which tracks the chip manufacturing industry, said that some chip foundries have reduced the steps of using ASML's flagship photolithography machines, some by almost a third. He cited Samsung as an example, saying that Samsung may be able to use more cutting-edge chip etching technology in the future to reduce the steps of using ASML's flagship photolithography equipment from five or six to just one or two steps. He said that if it is really successful, Samsung may have a large amount of chip manufacturing overcapacity on these EUV photolithography machine devices.
However, Jones, who has been tracking the dynamics of the chip industry for many years, pointed out that he has not changed his forecast for the entire chip industry. That is, as AI large models are updated and iterated to completely innovate the productivity of various industries, the demand for AI chips and AI-specific storage chips will flourish, and the future will inevitably drive the entire chip industry, including industrial simulation chips, electric vehicle chips, and consumer electronics SoC chips, into a prosperous cycle. "This is just a short-term fluctuation in the chip industry. From a long-term perspective, everything will be fine."
The AI boom is still in full swing, but the entire chip industry has not fully benefited from it.ASML's performance results indicate a significant bifurcation in the fortunes of global chip companies: the surging demand for data center server-side AI chips capable of handling massive parallel computing patterns and high computational density matrix operations, such as NVIDIA AI GPUs, driven by artificial intelligence applications like ChatGPT and Sora, has overshadowed the extremely sluggish demand in other segments of the industry.
ASML did not elaborate on why its third-quarter order volume was less than half of analysts' forecasts, merely stating that some customers postponed factory construction. Analysts speculate that the "customers" ASML referred to are likely Intel and TSMC—two major chip manufacturers that have previously indicated they would slow down the global construction of chip factories. In the face of shrinking chip product sales and an increase in net losses from chip foundry businesses, Intel is significantly slowing down the construction of chip factories, having postponed plans to establish new factories in Germany and Poland last month.
Janardan Menon, an analyst from the Wall Street giant Jefferies, stated in a report on Wednesday: "ASML's financial report reflects that while demand for chips related to artificial intelligence remains very strong, the recovery in other areas is exceptionally lagging, a trend that could persist until 2025."
The analysis team from another major firm, Bernstein, stated in a recent report: "We are concerned that the time required for the recovery of end-demand will be longer, leading to a significant delay in chip capacity expansion, which is what we need to face as we enter 2025." "It seems we may need to wait patiently until the cyclical recovery of the chip industry becomes clearer."
The latest semiconductor industry outlook data released by the World Semiconductor Trade Statistics Organization (WSTS) shows that AI-driven memory chips and logic chips such as GPUs and CPUs are the main drivers of the recovery of the entire chip industry. In contrast, WSTS's expectations for the market size of analog chips and microchips covering a broader range of consumer electronics, including electric vehicles, industrial fields, IoT devices, and PS5, Switch, etc., are very pessimistic. Even the market size forecast for the analog chip sector this year is negative growth, with a weak recovery process expected next year.
WSTS stated that the 2024 forecast revision reflects the strong performance of the past two quarters, especially in the computing terminal market. After a significant market contraction in 2023, WSTS expects that in 2024, two core chip product categories will drive double-digit sales growth, with the total sales of the logic chip category, including the explosively demanded NVIDIA AI GPUs and AI ASICs, growing by 10.7%, and the memory chip category, which best reflects the chip cycle, is expected to surge by 76.8% in 2024—WSTS forecasts that the strong demand for memory chips will focus on HBM used in AI training/inference fields, as well as enterprise-grade DRAM and NAND storage systems that are crucial for the efficient operation of AI data centers.
This situation of AI-driven storage and logic chip demand is very evident from the inventory and export scale of Korean chips, as South Korea is home to the world's two largest memory chip producers—SK Hynix and Samsung.
Data released by the South Korean government shows that despite a slowdown in growth, semiconductor exports in September still increased by a significant 37% year-on-year, marking 11 consecutive months of growth, slightly lower than the 38.8% increase in August. Among the continuously growing chip export data, as much as one-third of the growth contribution comes from HBM storage systems. HBM storage systems are used in conjunction with NVIDIA's core hardware for AI chips—H100/H200/GB200 AI GPUs. HBM and AI GPUs are indispensable for driving major artificial intelligence applications like ChatGPT and Sora. The stronger the demand for HBM, the more intense the demand for AI chips.
Bank of America, a Wall Street commercial banking giant, recently released a research report stating that the global artificial intelligence boom is still in its infancy, essentially similar to the path of internet development in the 1990s, comparable to the "1996 moment" of the booming internet, implying that in the Bank of America analysis team's view, the AI boom is still in a very early stage.
"Investors may have underestimated the long-term impact of this technology while overestimating its short-term potential, but this is also a typical characteristic of technological prosperity. It is expected that in the coming years, capital expenditures related to artificial intelligence may exceed one trillion US dollars, but compared to the internet era, the development of artificial intelligence has just begun," Bank of America pointed out in the report.