Involution and dependence: the lives of the three major semiconductor equipment suppliers in the United States

A month ago, the Semiconductor Industry Alliance of America (SIAC) was established. According to the official website of SIAC, the alliance has 66 members, including 10 semiconductor companies outside the United States.

The formation of this alliance is not surprising given the introduction of the US chip subsidy bill and the Biden administration's desire to strengthen the "onshore" semiconductor supply chain self-sufficiency system. However, it is worth noting that not all major US semiconductor manufacturers are willing to join this loose profit-sharing group alliance that is largely driven by geopolitics and whose form outweighs its content. For example, Lam Research, one of the three largest semiconductor equipment manufacturers in the United States, did not participate.

In line with this, a few days ago, Robert Maire, a senior contributor to the famous semiconductor community Semiwiki, wrote a short commentary, which triggered heated discussions among industry insiders. He believes that it is better to replace government subsidies with equity with a bet agreement to stimulate the development of the US semiconductor industry. He also defined Intel and GlobalFoundries as "losers" in the past decade, and believes that instead of investing in such companies with low expansion prospects, it is better to focus on more low-key and more basic semiconductor equipment companies.

There is no doubt that to this day, as the U.S. semiconductor industry continues to examine its own strengths and weaknesses, it has characterized semiconductor equipment as an advantageous area in the upstream of the global chip value chain that needs to be consolidated. According to data released by SIA (Semiconductor Industry Association), simply looking at the share and ratio of the value chain, U.S. semiconductor equipment ranks third after EDA and core patent fields, and occupies half of the world; but if you change your perspective, that is, consider market size and profit margins, then U.S. semiconductor equipment will undoubtedly occupy the first place in the global semiconductor market segment. After all, the global EDA's total annual sales are only about 70% of the annual revenue of Applied Materials (US$10.3 billion vs. US$14.6 billion). The 2019 data is as follows:

So, in the context of the global chip shortage crisis showing no clear signs of slowing down and raw material market prices rising sharply, it is necessary for us to carefully study the layout, direction and concerns of these major equipment manufacturers in the United States.

The "involutionary" battle between Applied Materials and Lam Research

VISL Chairman Risto Puhakka said in an interview with "Jiwei Interview" that this year, global WFE (wafer fab equipment) will grow by 12% year-on-year to $70.4 billion, which is almost equivalent to the total sales of global semiconductor equipment in 2017. It is also worth noting that the compound annual growth rate of WFE in the past five years has outperformed the growth rate of semiconductor device sales and chip company capital expenditures, which also shows that the equipment sector still has huge growth potential. In 2018, the industry generally estimated that the WFE market size would exceed $70 billion by 2023. Now it seems that this judgment has been realized two years ahead of schedule.

Global semiconductor WFE market share: Applied Materials, ASML, Lam Research, and Tokyo Electron rank in the top four

Looking at the WFE global market growth curve, it had long hovered around US$30 billion before 2017, and began to surge in 2018, but then fell slightly in 2019. We can basically conclude that WFE follows a moving average growth curve rather than a linear mean growth, so we often need to use four years as a unit to judge the performance trend of WFE suppliers.

Lam Research has ranked among the top three in the world in terms of WFE expenditure and revenue over the past decade. Lam Research's performance growth rate has been twice the average growth rate of the entire semiconductor industry over the past decade. The company's long-term goal is to increase its global equipment market share from 20% to 25% by 2024, with revenue reaching US$14 billion to US$17 billion.

In the abstract, it is reasonable to maintain a sufficiently optimistic attitude towards the company's trend, but a closer analysis will capture some seemingly inexplicable phenomena, such as Lam Research's share of the entire global semiconductor equipment market has dropped from a peak of 13.2% in 2017 to 10.6% in 2020, and this is when the company's revenue share from foundry customers suddenly increased. The reason behind this is not difficult to explain - Lam Research is too dependent on the sales of memory semiconductor equipment.

As we all know, the price fluctuations of DRAM and NAND products are larger than those of other semiconductor products, and these products account for 60%-70% of Lam Research's total revenue. In 2019, the global memory chip market was in a digestion period caused by excessive capital investment in the early stage, and equipment procurement slowed down.

If we look at the growth of memory semiconductors over the past decade, we will find that the prices of DRAM and NAND (generally weighted calculation) have a more obvious increase every four years. The last rising cycle of NAND prices ended in 2018. Therefore, Lam Research's financial situation in 2019 was relatively difficult. Compared with its direct competitors Applied Materials and Tokyo Electron, it can be seen that the company's financial curve is relatively more consistent with the price curve of memory chips:

Of course, taking a step back, DRAM and NAND are indeed important components of WFE. In 2018, NAND manufacturing costs once accounted for more than 40% of the total wafer equipment expenditure. Generally, this figure is around 20%-25%:

Data source: informationnet

Almost the same situation has happened to Applied Materials since the second half of last year. After the company announced its new quarterly results on May 20, the stock market did not rise as expected by the outside world. The main reason was that the NAND of semiconductor systems, which accounted for two-thirds of the company's business, fell by 1.5 percentage points instead of rising.

For a long time, Lam Research's main customers have been storage semiconductor companies such as Micron, Samsung, SK Hynix, and wafer foundry TSMC, which account for more than 10% of Lam Research's business; while Applied Materials' largest customers are TSMC and Intel. In 2019, Samsung's share of the company's business once fell to 10%.

The overlap of the two parties' customers does not seem to be that high, and last year the two parties achieved a "win-win" situation in terms of revenue and gross profit margin. However, Applied Materials' business was affected by two non-repeatable incidental factors. One is that SMIC placed a large number of orders with Applied Materials before hearing about the US ban (September 15, 2020), which also caused the company's equipment shipments to China in the third quarter of last year to surge by 22.5% month-on-month. Almost in the same period, Intel, another major customer of Applied Materials, outsourced its 7nm chips to TSMC, which also caused the company's equipment shipments to Taiwan, China in the third quarter of last year to increase by about 36.2% month-on-month.

However, the above facts cannot conceal the fact that the business scope of Applied Materials and Lam Research Semiconductor Equipment has a high degree of overlap, and they often compete fiercely in the same sub-segment, such as thin film deposition, plasma etching and wafer cleaning processes.

In recent years, the ratio of WFE allocated to NAND and DRAM has been roughly maintained at around 6:4. However, many foundries with cutting-edge chip processes need the "Leviathan" of lithography machines - EUV technology provided by ASML to improve more advanced processes. The shipment volume of EUV is often directly related to the expenditure of other equipment supporting the wafer fab, which objectively leads to the "zero-sum game" between Applied Materials and Lam Research. Let's take a closer look at the two companies' pursuit of each other in certain tracks:

Data source: Gartner

Both parties rank in the top three in market share in the fields of CVD, photoresist cleaning, dry etching, etc., and each has its own "killer weapon", such as Lam Research's photoresist edge repair (Edge Bead Removal) and Applied Materials' high-precision ion implantation technology.

KLA-Tencent and Advanced Foundry

If Applied Materials and Lam Research often face a cruel "winner takes all" situation, KLA, another important US semiconductor equipment supplier, has a more subtle sense of pattern. The company is obviously different from the first two. In the field of metrology inspection, its market share has been further consolidated from 55% to 62% in the first quarter of this year in the past five years. Applied Materials, which ranks second, has only 12%, and ASML has 6%, which is a gap in grade from KLA. If we list the sectors of semiconductor equipment laid by these three companies one by one, we can see that KLA is the only company that relies heavily on the advanced process of foundries. Two-thirds of their revenue comes from the foundry/logic department, which is 10 percentage points higher than ASML (this is the data after ASML acquired the inspection company Hermes Microvision). As advanced foundries like TSMC move towards chip processes with smaller process nodes, the difficulty of monitoring and inspecting chips has become more important and more difficult, and competitors are becoming more and more concentrated. In the process of technological breakthroughs from 7nm to 3nm, only KLA, Applied Materials and ASML are left.

KLA dominates the wafer front-end yield inspection equipment market (@informationnet)

In the overall layout of front-end inspection, Advanced Materials chose key breakthroughs. For example, the key market share in CD SEM (CD-SEM nanoline width measurement experiment) increased from 17% in 2019 to 20% in 2020; and the competition between KLA-Tencor and ASML is more interesting, with a jagged and interdependent approach. Last year, ASML sold 12 EUV lithography systems to TSMC and 9 to Samsung. This year, ASML estimates that 17 EUV systems will be shipped to TSMC, and Samsung is expected to still obtain 9 devices. As an important customer of KLA-Tencor, TSMC can only increase the production capacity of advanced process chips and thus improve its procurement rate from KLA-Tencor if it obtains enough EUV lithography systems. KLA-Tencor can only improve its own inventory management level by relying on ASML to continue to ship goods to TSMC stably and smoothly. It can be said that one suffers and one prospers, all prosper. TSMC and Samsung contributed 27% and 17% of KLA-Tencor's revenue respectively. The capital expenditures of these two companies this year have increased by an average of more than 60% year-on-year. Especially for TSMC, the 7nm and higher process wafer front-end equipment is very dependent on KLA-Tencor. Based on this, Wall Street generally believes that the company's outlook for a 15% year-on-year revenue growth this year is still somewhat conservative.