TSMC is the world's largest
independent semiconductor foundry. It has a 50% share in the market. They
supply fabless chip designers like Apple, Nvidia, Qualcomm and AMD. There has been a lot
of press and interest recently in its work. But TSMC has its own suppliers too. These are the
people who make the super cutting-edge machines that TSMC operates 24/7. Dutch multinational
ASML is TSMC’s most important supplier and the maker of the EUV equipment
critical to its N7+ and N5 processes. In this video I am going to profile this rather
obscure company and examine its relationship with TSMC. I’ll also help answer the question of
who is “more important” so that the fanboys can better score points against each other on
internet boards. Face-off! Who matters more!? Spoiler alert: They both matter
a whole lot to each other. ASML is a leading provider of advanced technology
systems for the semiconductor industry. The company was founded in 1984 as a joint venture
between Dutch firms ASM International and Philips. ASML's first machine, the PAS 2000 stepper, was released in 1985. A stepper is a device, kind
of like a slide projector, that projects a design onto a silicon wafer. ASML would subsequently
make over 50 versions of this machine. For the first ten years of its existence,
the company struggled to compete in the semiconductor lithography space. The semiconductor
supplier industry faced immense challenges as the economy went through various boom and busts. ASML
lost money for the first six years of its life and in 1990 it would be spun
out as an independent company. Two years later in 1991, ASML comes out with
its first hit product - an i-line stepper. I-line meaning that the stepper
uses the 365 nm light wavelength. This lets foundries print smaller designs
and continue to push Moore's Law forward. The product's success allowed ASML to go IPO
in 1995, raising money for further product development. After acquiring a few American
lithography companies, the company grew into an oligopoly alongside Japanese firms
Nikon and Canon. Then came the mid-2000s, where a pair of big product launches
allowed ASML to take the lead. In 2001, ASML developed and released its
TWINSCAN system. This machine offered a process optimization feature that helped foundry customers
etch more wafers in the same amount of time. Nikon and Canon could not replicate
this technology and it gave ASML its first market share lead over the Japanese. The second big move came with a generational shift
in photolithography machinery techniques called "Immersion Lithography". Roughly speaking, it
uses water as a lens to shrink the wavelength of the wafer etching laser. Yeah it sounds nuts but
this technique, discovered by Taiwanese scientist Burn Lin in the 1980s, helped keep Moore's Law
alive for a little longer without needing to further shrink the light wavelength. This
helped cement ASML's 83% market share. But the industry and ASML continued
to think about what could be the next great revolutionary technology in
the semiconductor production space. They started an R&D office and began
investing millions of euro into R&D. I discussed a bit of ASML's work in Extreme
Ultraviolet Lithography in a previous video of mine. I am not going to tell the same
story a second time so you can go watch that sometime. Instead I want to tell the same story
but from a different perspective. Like Tenet. In 2003, the industry moved forward with the
idea of using EUV for the next generation of lithography tools. This decision at the
time was controversial and heavily debated. EUV at the time did not have
any theoretical blockers. It wasn't a time travel Delorean or a
faster-than-light energy drive, after all. But there were significant practical challenges
of all sorts to getting it up to speed. For instance, the light source would have to
be entirely redesigned from the ground up. End user customers like Intel, Samsung
and TSMC would have to retool their fabs. Some argued that current techniques like
Immersion Lithography and another thing called multi-patterning could
carry the industry forward. So many things are not working out when
it came to EUV. Developing it could turn out to be a money pit. Why shoot for a
unicorn? Make what already works, better. ASML was fortunate to be
in a position to try and do both. They continued development in EUV but
invested to retain a leadership position in current techniques. From 2008 to 2014, ASML as
a company invested over 4.6 billion euro in R&D. Over time by 2010, the industry's big
players like TSMC, Intel and Samsung came to believe that EUV could be brought to
fruition. ASML saw an opportunity to steal Intel from its previous main
supplier Nikon. Thus in 2012, ASML announced a customer co-investment scheme
where Intel received ownership of 15% of ASML in exchange for $4.1 billion. The money
would help develop future tools like EUV. For some reason, Intel did not ask for product
exclusivity - something that raised a few eyebrows at the time. It allowed TSMC and Samsung to also
buy in later on. The money and the customer buy-in gave ASML breathing room to work out
all the kinks. Of which there were many. In 2016, the first EUV devices were
shipped out of the Netherlands - 30 years after its concepts were first
explored. The technology had been delayed for over ten years after its 2007
prototypes, but it was now ready to go. I think what is interesting is how
the semiconductor industry's major players banded together to invest
and collaborate towards a final, finished technology that they would all share in.
These companies, ordinarily fierce competitors, saw the threat of not having anything ready
for the future and decided to work together. ASML could not have finished EUV without
intimate participation from its customers. These customers invested billions of euro
into the product development process, aware that they would not see the results
of their investment for many years. Without this buy-in, ASML would have
ended up like Canon - unable to sustain this decades-long odyssey without going
bankrupt. We should acknowledge that. On a side note. Recently, the
US government has intervened to prevent the export of ASML's EUV machines
to China. I have seen some people asking on what grounds the United States has in
restricting the export of EUV. After all, ASML is a Dutch company and last time I checked,
the Netherlands is not an American state. Additionally in 2012, then ASML CEO had said: > We welcome Intel as the first
customer to agree to contribute to these [investments in semiconductor equipment], the results of which will be available to every
semiconductor manufacturer with no restrictions Read my lips. No restrictions. So what gives? This is because EUV is literally an
American-founded technology. Think of EUV as a startup and the US is its controlling
shareholder - its Mark Zuckerberg, so to say. EUV LLC was founded by American semiconductor
firms Intel and Micron in 1997 to partner with the US Department of Energy to commercialize
its EUV studies. EUV LLC invested $250 million over 3 years at the earliest stages and as a
result receives royalties on all EUV revenues. In February 1999, ASML negotiated for and received
a license from EUV LLC for its EUV technologies. Back in 1999 - long before ASML would
come out with the products that won it the market - Japanese companies
Nikon and Canon still had 54% share of the semiconductor
lithography market. ASML had 34%. There was no American player. Had there been,
I believe they would have received this license first. The EUV LLC participation allowed
ASML to leapfrog the Japanese's progress in EUV and get that critical first buy-in from
major customers like Intel, TSMC and Samsung. Canon and Nikon did not receive a license
and eventually declined to develop EUV to the same degree. Canon
in 2007 due to financial issues. Nikon because they pursued the continued evolution
of older techniques that ended up petering out. Yes, we all agree that ASML invested
immense resources in making EUV the final, usable product that it is today. They took
a huge risk and swung at the right pitch, ending up to be the right company at the
right time. And we agree that Europeans should not always go along with what America says.
Maintaining national sovereignty is important. But the EUV example to me is more clear
cut than most. Europeans, South Koreans, Indians, Taiwanese. A whole bunch of people
contributed to the effort. But America founded and led it. (Side note to the side note. I
am aware of an EU-backed EUCLIDES program pursuing EUV but it apparently did not seem
to be as successful as EUV LLC.) And if you have worked in a tech startup before, you would
know that the founder's say matters the most. I got the idea for this video largely because of
the nationalist internet food fight that tends to flair up every time TSMC appears in the news. You
have a choir of comments which immediately sing “But they all depend on ASML ... which is Dutch.” Which in turn triggers talk of ASML’s
American subsidiaries and so on. Setting immature nationalism aside, the
actual question is a good one to ask. In my opinion, good questions help you get a
better understanding of the situation at hand. Let us use a metaphor. Semiconductor foundry
work is like baking a pound cake. You have raw materials, you need to process them
together in special ways over many steps for which you use specialized equipment
(an oven by Middleby, for instance, or a blender by Hamilton Beach) to process
those materials into a final product. You can liken TSMC to the baker in the kitchen.
ASML is like Middleby here, providing the high tech oven. The baking of a cake is just one
step out of many. A critical one for sure but if you focus too much on it then you ignore the
importance of stirring together the flour and eggs as well as applying the creamy frosting to give
it that extra deliciousness. Your cake might suck. Same with microchips. Lithography is a
critical step in the final development, but there’s many more that need to be finished off before the final chip is ready to be
put into a computer or graphics card. Packaging and testing, for instance.
Otherwise it’s still just a chunk of silicon. This makes intuitive sense when you think
about the progress made by other companies using ASML's same EUV machines. TSMC has the
most EUV machines in the world, but Intel and Samsung also got the same machines. Intel was
literally the lead investor in making EUV happen. Neither of those foundries are
at the same level of yield and output. Cooking skills and experience
matter - as my sucky cakes can attest. When the customer sits down to the final product,
they see and enjoy a cake. Can you attribute a percentage “credit” of the cake’s deliciousness
to the eggs and the flour from which it was made? To the heat from the oven in which it was
baked? To the skill of the cook who baked it? You can try but it all ends up in a ridiculous
exercise of numbers pulled out of your butt. Because the reality is they all came
together in a single harmonious moment to give you this amazing thing.
