ASML is the most important company you've never heard of.
The $300B+ Dutch firm makes the machines that make semiconductors. Each one costs $150m and access to them are a huge geopolitical flashpoint.
Here's a breakdown 🧵
1/ What *exactly* does ASML sell?
Its key product is an extreme ultraviolet lithography (EUV) machine, which uses advanced light technology to "print" tiny circuits onto Silicon wafers.
Only ~50 are made a year and ASML has a near monopoly on the machine technology.
2/ You def know ASML's main clients: Intel, Samsung and Taiwan Semiconductor Manufacturing Co. (TSMC).
They need EUV to keep Moore's Law ("# of transistors on microchips doubles every 2 years") alive and continue to advance computing.
Total ASML sales in 2020 = $16B+.
3/ The ASML story begins in 1984, as a joint venture between Dutch conglomerate Philips and an electronics maker called Advanced Semiconductor Materials Int.
The project had a very humble start: it was launched in a shed behind a Philip's building in Eindhoven, Netherlands.
4/ The venture's first product was called the PA 2000 stepper (think a slide projector "projecting" designs on silicon).
For years, the product failed to make headway against leading Japanese competitors (Nikon, Canon) and, in 1990, ASML was spun out as its own company.
5/ ASML scored its 1st hit product in 1991, giving it momentum to IPO in 1995.
Soon after, it acquired a # of US lithography firms and -- by the end of 1990s -- it had comparable market share to Nikon and Canon.
From there, ASML made 2 big bets that separated it from the pack.
6/ Both bets were made to keep up with Moore's law.
First: In 2006, ASML released its TWINSCAN system using immersion lithography (it utilizes water as lens to shrink the laser's wavelength = more circuits on chip).
It was ASML's first market-leading product.
7/ Also in the mid-2000s, ASML started spending huge R&D on EUV technology.
It was a massive risk, though: EUV lithography would require Samsung, Intel and TMSC to completely rebuild and redesign their fabrication plants.
From 2008-14, ASML put $5B+ into EUV research.
8/ The science behind EUV was established in the late 1980s. It was a US-led effort between the Dept. of Energy and industry (e.g. AMD, IBM, Intel).
ASML licensed EUV tech in 1999. Canon elected not to pursue it due to financial problems while Nikon chose to develop older tech.
9/ How EUV works today:
◻️ A tin droplet drops into a vacuum
◻️ It's pulsed by a high-power laser
◻️ Tin atoms are ionized, creating plasma
◻️ A mirror captures EUV radiation emitted by plasma
◻️ Mirror transfers EUV to wafer (wavelength=13.5 nanometers, basically X-ray level)
10/ The potential of EUV was so great that Intel, Samsung and TMSC -- all competitors -- jointly acquired 23% of ASML.
Intel put up the most: €2.5B for a 15% share (today, the firms have sold down most their stakes).
The first production-ready EUV machine was released in 2016.
11/ Why are EUVs so expensive?
ASML plays a role similar to Boeing for airplanes (also $100m+ products): it's an integrator of 4750 global high-value parts suppliers:
12/ Why can ASML only produce 50 EUV machines a year?
◻️ Co-ordinating 1000s of suppliers is very difficult (just like an aircraft)
◻️ Each machine is custom (30+ variables to choose from)
◻️ Lead-time are long (speciality parts like the Zeiss lens takes 40 weeks to produce)
13/ The delivery process is nuts, too:
◻️ Each EUV weighs 180 tons
◻️ A disassembled EUV takes up 40 shipping containers
◻️ Shipping it (mostly to Asia) takes 20 trucks and 3 Boeing 747s
◻️ ASML teams must be on-the-ground to maintain them
◻️ The min spend to house EUVs is $1B
14/ Today, ASML has a 90% share in semi lithography (EUV and Deep UV).
EUV tailwinds are huge:
◻️ Semi CAPEX >$120B+ in 2021 (similar spend in following years)
◻️ Key sectors (esp. auto AKA Tesla chips) will see growth for years
◻️ Transition to 5nm process requires more EUV
15/ Even if machine sales slow, ASML's business is increasingly shifting to system maintenance, relocation and upgrades.
Over a 20yr lifespan of an ASML machine, services-based sales may reach 50% of the initial machine price (w/ high margins)...across a growing installed base.
16/ With chips needed in everything (data centres, AI, autos, mining), semis are the OIL of the 21st century.
The US has even blocked Dutch exports of EUV-licensed tech to China. As the Tech Cold War heats up, expect to hear more of ASML: the $300B+ giant that started in a shed.
17/ If you enjoyed that, I write threads breaking down tech and business 1-2x a week.
Def follow @TrungTPhan to catch them in your feed.
Here's a related one that might tickle your fancy:
When Iron Man came out in 2008, Robert Downey Jr. was *not* a marquee star.
He was rebuilding his career and paid a below market rate of $500k.
But the deal terms set him up for one of the great acting comebacks ever (while earnings $450m+ as Tony Stark).
Here’s the story 🧵
The Marvel Cinematic Universe (MCU) we know today was a long shot in the early 2000s.
Marvel was a public co. coming off bankruptcy in 1996 and had sold rights to its best IP (Spiderman, X-Men, Fantastic 4)
From 2000-07, films based on the IP minted cash but Marvel made little:
In the early-90s, Downey Jr. was one of the brightest young stars in Hollywood, receiving a Best Actor nomination for "Charlie Chaplin" in 1992 (@ 27yrs old).
In the 2nd half of the decade, though, he dealt with drug addiction, arrests and jail stints before going clean in 2003.
Masayoshi Son does the craziest investment swings:
▫️In mid-90s: invested $1.7B into 100+ internet firms (including a ~30% stake in Yahoo! for $100m)
▫️In 2000: was worth $78B at peak Dotcom and was the richest person in the world for 3 days (ahead of Gates)
▫️The bubble burst and he lost 99% of wealth
▫️In 2000, puts $20m into Alibaba for a 34% stake (sold out almost entirely by 2023 and made ~$72B)
▫️Lost $14B on WeWork
▫️Once owned ~5% of Nvidia but sold it all for $3.6B in 2019 (that stake would now be worth $90B)
▫️In 2016, Softbank bought Arm Holdings for $32B (still owns 90% and the stake is $114B, a gain of $82B)
Based on his ownership in Softbank and other investment vehicles, his personal wealth is currently ~$15B.
Nearly 100% of intercontinental internet traffic goes through submarine cables.
It is a robust system with many redundancies.
There are 500+ subsea cables and a fleet of 60 repair shops on stand-by but Big Tech isn’t taking chances:
▫️GOOGLE invested in 25 cables (and owns 12 outright). Per The Economist, the search giant started its sea cable program in 2008.
◽META invested in 15 cables (owns 1 outright).
◽MICROSOFT partly owns 4 cable.
One of the 500+ cable gets cut every 3 days (most common reasons are shark bites, anchor drops and deep-sea fish trawlers).
Remote areas are still very at risk.
Example: In 2022, a volcano erupted near Tonga and a mudslide took out the only cable nearby. Starlink provided some free internet coverage while it took 5 weeks for the cable to be fixed (5 weeks!!).
Robert Metcalfe (inventor, ethernet cable) famously predicted internet would flame out. He thought cables couldn’t handle traffic and not enough investment in them.
Dyson created the first bagless vacuum and used the cyclone tech for related products: air purifier, hand dryer, fans and hair dryer.
It took a big swing on EVs and missed ($500m+ on R&D). There was also a washing machine flop.
But the most random product line? A strawberry farm that grows off-season so the UK can have local access to the fruit year round.
Dyson Farming was established in 2012 and — with 36,000 acres — is one of the largest farms in the country.
The semi-automated strawberry farm has 700,000 plants and will produce 750 tonnes of strawberry a year.
My wife made me buy 4 of the aforementioned Dyson products. You can probably guess which ones.
They all have cyclone technology, which James Dyson borrowed from the sawmill industry…and it is my favourite cross-industry innovation: readtrung.com/p/11-types-of-…
Dubbed “Pokémon With Guns”, the game has sold 6m+ copies in past 4 days (making its developer Pocket Pair over $100m on a ~$7m budget).
Some details on the game development process:
◻️ CEO Takuro Mizobe worked at JPMorgan Securities before launching a crypto exhange in 2014.
◻️He made some dough in crypto and used funds to launch indie game developer Pocket Pair in 2015.
◻️The original Palworld team was 4 people and started with $10,000.
◻️The main model developer is a high-schooler who the team met because he worked part-time at a convenience store they frequented.
◻️They added guns because to be a global success, it had to do well in America and the CEO said “Americans like to shoot things”.
◻️Asked about Nintendo, the CEO says they make innovative games whereas he is fine to chase trends (“I don’t [have a creative vision]. I just want to make a game people like.”)
◻️ There was no version control. The team saved files in a “bucket of USBs” and merged them to the main build when done.
◻️Switched from Unity to Unreal Engine late into the development process.
◻️ There are rumours that the game assets were mostly created by AI (but they deny it and there is little evidence).
Palworld’s early success is astounding when you consider triple AAA game titles easily break 9-figure development costs (eg. Cyberpunk 2077 has run $400m+)
Earlier today, Palworld peaked at 1.8m concurrent players. That’s the 2nd highest ever for Steam’s platform (ahead of CounterStrike 2 and behind only PUBG).
The next stat to watch: Nintendo released “Pokémon Scarlet and Violet” in 2022 and it has sold 23m+ copies.