ASML and Canon Stock Analysis

ASML has a stranglehold (but not a total monopoly) on lithography equipment for semiconductor manufacturing. Japanese conglomerate Canon (CAJPY) is doubling down on competing advanced chipmaking equipment it originally acquired in 2014, and recently announced a new breakthrough it thinks could compete with ASML’s most advanced machines. Canon certainly wants to be more than a PR-driven player in this department, but ASML’s roots are deep in the chip supply chain.

The semiconductor manufacturing process is getting increasingly complex and expensive. As a result of multi-decade efforts to “shrink down” features of the most advanced chips to make them more powerful and energy efficient, semi manufacturing complexity and costliness are poised to continue rising for the foreseeable future. 

As a result, numerous industry specialists have emerged, some of which tout defensible business models due to their deep engineering know-how and the massive amount of money and time it would take for peers to compete against them.

One of these specialist companies is ASML Holding (NASDAQ:ASML), the Netherlands-based lithography equipment business in the large semiconductor wafer fab equipment space. ASML is the only company that produces equipment utilizing extreme ultraviolet lithography (EUV) technology used to make the most powerful logic, networking, and memory chips. It also has a stranglehold on other types of lithography machines used in more “mature” chip manufacturing processes.

However, ASML isn’t a total monopoly as some believe. It has some small competitors. One is Japan’s Canon (OTC:CAJPY), which recently announced it has begun selling machines that could compete with ASML’s EUV tech. Are ASML’s days of dominating this crucial choke point in the computing technology supply chain about to come to an end?

A brief review of the lithography landscape

In order to shrink down microscopic features of chips, billions of dollars of investment have been poured into wafer fab equipment — the machines used to develop the thin discs made of silicon you’ll often see someone holding in a bunny suit. Once those silicon wafers enter a fab, one of the first of many dozens of steps is lithography. 

Lithography is a process in which a photoresist chemical (which reacts to certain types of light) is uniformly applied to the surface of the wafer, and then ultraviolet light is shown through a photomask (which allows some of the light through, and blocks other parts of the light source, to make patterns on the wafer coated with photoresist). After that, the photoresist that was not exposed to light is stripped away, leaving behind lines and channels on the wafer’s surface that later become circuits in a chip (once the wafer fab is finished and the wafer is cut into chips and packaged into a computing system).

This chart from Tech Insights back in 2021 shows the general steps and types of equipment involved in wafer fabrication in general chronological order. The relative amount of money spent on each type of equipment is shown by horizontal width, and individual company dominance of that equipment type indicated vertically. The companies shown on this chart are what we call the “fab five,” by far the largest participants in the wafer fab equipment market. Besides ASML in purple (the dominant force in lithography on the far left), there’s Applied Materials (NASDAQ:AMAT) shown in blue, Lam Research (NASDAQ:LRCX) displayed in green, Tokyo Electron (OTC:TOELY) in yellow, and KLA Corp (NASDAQ:KLAC) in orange. 

ASML makes the lithography machines that produce the ultraviolet light source that shines through the photomasks and onto those wafers. As you can see from the chart, ASML accounts for most, but not quite all, of lithography equipment revenue (sidepoint: ASML also dabbles a bit in metrology and inspection, as indicated by the big bar dominated by KLA Corp close to the far right of the chart, see our latest videos and articles on metrology companies KLAC, ONTO, NVMI, and CAMT). 

Who makes up the rest of the lithography market revenue, shown by the gray portion of the column above ASML? A small patchwork of litho equipment competitors — none of which exclusively focus on lithography — snap up the leftover crumbs, including Veeco Instruments (NASDAQ:VECO), Japan’s Nikon (OTC:NINOY), and fellow Japanese conglomerate Canon. 

Exchange rates as of Oct. 16, 2023. Data source: ASML Holding, Canon, Nikon, and Veeco Instruments investor relations. https://www.asml.com/en/investors/financial-overview https://www.nikon.com/company/ir/ir_library/fs/pdf/2023/fs2023.pdf pg 23 https://global.canon/en/ir/business.html https://ir.veeco.com/news-and-events/news-details/2023/Veeco-Reports-Fourth-Quarter-and-Fiscal-Year-2022-Financial-Results/default.aspx 

Canon about to cash in on ASML’s expensive machines?

Decades ago, the semiconductor industry came to a consensus that photolithography (using a light source) was the best way forward for crafting smaller featured chips. ASML’s research and development, with ample amounts of industry mindshare and financial funding along the way, now dominates with its equipment lineup. https://www.asml.com/en/products/duv-lithography-systems https://www.asml.com/en/products/euv-lithography-systems 

The machines that actually make up the bulk of ASML’s revenue utilize older deep ultraviolet (DUV) laser technology: dry (just a UV laser) and wet, or immersion (a UV laser shone through a droplet of water) ArF (argon fluoride lasers) and KrF (krypton fluoride) ultraviolet lithography systems. These machines can cost tens of millions of euro each. The latest and greatest machines, extreme ultraviolet (EUV), are fast approaching 200 million euro a pop. High-NA EUV (which Intel (NASDAQ:INTC) says it will be the first to use in production in 2025) will likely be up to 300 million to 400 million euro. But that’s a topic for another time. https://seekingalpha.com/article/4618131-asml-holding-n-v-asml-q2-2023-earnings-call-transcript https://www.anandtech.com/show/20044/asml-to-deliver-first-highna-euv-tool-this-year https://www.asml.com/en/news/press-releases/2022/intel-and-asml-strengthen-their-collaboration-to-drive-high-na-into-manufacturing-in-2025 

But as hinted before, photolithography isn’t the only possible solution. Nanoimprint lithography (NIL, an appropriate acronym as you’ll soon see) was another promising tech that involved uniform coating of photoresist on wafers, but then using a mold with all of the tiny peaks and valleys of the future chip’s circuitry pressed onto the surface. The photoresist coating would then squeeze into the channels, ultraviolet light is applied to solidify the photoresist into a resin, and the mold removed to reveal a 3D “stamp” of the chip behind. https://global.canon/en/technology/nil-2023.html https://semiengineering.com/what-happened-to-nanoimprint-litho/ 

(See video on Canon NIL article link above)

Canon sells equipment outside of lithography too (like deposition, etch, and various machine components). However, nanoimprint lithography is the tech that gets the most attention, as Canon has been trying to develop it to offer an alternative to ASML’s increasingly complex and expensive machinery. Various parties have been involved with NIL technology, and Canon has been a big contributor. It got more serious about it back in 2014 when it acquired the small Austin, Texas-based R&D outfit called Molecular Imprints (purchased for under $100 million, and now known as Canon Nanotechnologies). https://www.reuters.com/article/us-canon-molecularimprints/canon-says-to-buy-u-s-s-molecular-imprints-idUSBREA1D02A20140214 https://www.usa.canon.com/business/semiconductor 

Just like photolithography had to overcome obstacles in physics and chemistry to be a production-viable process, so has NIL. And up to this point, NIL just hasn’t received near the same level of financial or engineering attention. Challenges involving the proper application of photoresist, aligning the mold to the wafer properly, making sure photoresist doesn’t leak into areas it shouldn’t when the mold is applied to the wafer, the infiltration of contaminants, and ensuring no warping of the resulting photoresist resin all proved to be effective barriers to NIL gaining commercial adoption. The photolithography machines are costly to buy and operate, but photolithography has nonetheless been the big winner up to this point. https://global.canon/en/technology/nil-2023.html https://semiengineering.com/what-happened-to-nanoimprint-litho/ 

But Canon and other NIL developers (Nanonex, EV Group, SUSS MicroTec) still think NIL is ready especially as it can be far cheaper (and perhaps quicker, given patterns can be stamped onto a wafer in a single step) to operate versus photolithography. There are applications for NIL in non-chip manufacturing like in LED and OLED displays (known as FPD, or flat-panel display, manufacturing, you’ll often see this called FPD lithography or FPD nanoimprint lithography equipment on company presentations). 

DRAM and NAND memory chip manufacturing are another possibility, and one Canon has been excited about. https://global.canon/en/technology/nil-2023.html https://semiengineering.com/what-happened-to-nanoimprint-litho/ 

The ASML moat is wide and deep, but rising complexity and costs are a real risk

It’s in the extremely cost-sensitive memory chip manufacturing market that Canon hopes to finally begin making waves. After delivering multiple NIL machines for test and development over the years, Canon recently announced the availability of its latest FPA-1200NZ2C model. According to Canon:

“Canon’s NIL technology enables patterning with a minimum linewidth of 14 nm2, equivalent to the 5-nm-node3 required to produce most advanced logic semiconductors which are currently available. Furthermore, with further improvement of mask technology, NIL is expected to enable circuit patterning with a minimum linewidth of 10 nm, which corresponds to 2-nm-node.” https://global.canon/en/news/2023/20231013.html 

In other words, Canon’s FPA-1200NZ2C could compete with ASML’s EUV machines, which are currently working hard at Taiwan Semiconductor Manufacturing (NYSE:TSM), Intel, and Samsung fabs at 5-nm, 3-nm, and smaller manufacturing processes.

Over the years, NIL machines from Canon Nanotechnology have been delivered to Toshiba (OTC:TOSYY), which partners with memory chipmaker Western Digital (NASDAQ:WDC). Toshiba has also partnered with memory giant SK Hynix (part of South Korean conglomerate SK Group) to also work on NIL in DRAM and NAND memory chips. https://www.theregister.com/2023/10/13/canon_nanoimprint_litho/ https://news.skhynix.com/sk-hynix-and-toshiba-sign-joint-development-for-nil/ 

To be clear, I don’t think NIL developments make Canon a great investment. On the contrary, at just 8% of total sales last year, wafer fab equipment don’t move the needle much for Canon’s overall flat-and-volatile revenue and profits, and NIL won’t change that anytime soon. It could take a few years for NIL to ramp up to more widespread use, if it does at all. https://global.canon/en/ir/business.html https://app.tikr.com/stock/financials?cid=195225&tid=6207231&tab=is&ref=vd1u2u 

Nevertheless, in past analysis videos on ASML, we have called out alternatives to photolithography as a risk to ASML’s long-term dominance. The semiconductor manufacturing industry has lined up for DUV and EUV machines to chart a path on their future chip development roadmaps. That kind of multi-billion-dollar and multi-year development pipeline isn’t going to be quickly discarded. 

However, as ASML machines get more expensive, and more costly to operate, chip manufacturers might be open to having other conversations. For example, earlier this year, Applied Materials announced a new type of machine that performs a step called “pattern shaping” (the Centura Sculpta machine) that’s complementary to ASML’s lithography machines. Pattern shaping could reduce the number of lithography exposure steps (called multi-patterning, with some wafers getting exposed to UV lasers dozens of times), which could lower total cost of production with lower energy usage and quicker production times. https://www.youtube.com/watch?app=desktop&v=GSuTyOMq1Bg https://ir.appliedmaterials.com/static-files/50698544-1a56-4e66-a287-5fd9121f849e pg 16 https://www.appliedmaterials.com/us/en/product-library/sculpta.html 

Another example: It’s worth noting that EUV machine adoption has been no cakewalk. TSM has been having issues scaling its latest 3-nm process, perhaps due to EUV challenges, and ASML itself has said its customers have pushed out delivery of EUV. DUV machine sales, particularly to China, have offset that effect in 2023, but further EUV adoption delays could certainly happen.

Perhaps Canon’s NIL machines will finally start to pay off as these complexities mount. As for ASML, its stock has been highly volatile this year, with an impressive rally getting mostly reversed in recent months, though that’s due primarily to geopolitical risk (EUV, and now most DUV machines, sold to China having to go through export controls). Even so, the stock trades for a premium 31 times trailing 12-month earnings per share, and 40 times free cash flow (FCF), as of this writing. The market still thinks ASML is a dead ringer for long-term growth. https://app.tikr.com/stock/multiples?cid=388904&tid=20212083&tab=multi&ref=vd1u2u 

Semiconductors are a massive and still high-growth market (expected ~7% CAGR through 2030), and there will be plenty of need for more ASML. We have a long position in the stock, which we initiated last autumn. But at that time, the stock carried a depressed valuation. These days, that premium warrants more caution, and risks should be kept in mind. We believe this can be a long-term market-beating investment from here, but ASML isn’t cheap. Tread carefully, and perhaps scale into a position over time given the high valuation.