The Lithography Canon $CAJPY
ASML is the single most important company in the entire semiconductor supply chain and nobody who matters has a backup plan. One Dutch company. One factory in Veldhoven. Every leading edge chip, every H100 melting through OpenAI's electricity bill goes through an ASML EUV machine. If more hanta virus ships dock in the Netherlands and hits Veldhoven the global economy enters a coma.
Enter Canon. Yes, that Canon. The camera company your dad uses to take pictures of your wife's boyfriend and your mom. They've been quietly cooking something for 10 years.
What Canon is actually doing
Canon is NOT building a better EUV machine. That race is over, ASML won, end of story.
Canon is building a completely different way to make chips called Nanoimprint Lithography. Instead of using $200M lasers to project patterns onto a wafer like ASML does, Canon literally stamps the pattern into the wafer like a wax seal. Their commercial system is the FPA-1200NZ2C and their branded version of the tech is called J-FIL.
Trillion-dollar industry. They stamp it, yes.. like your wife's boyfriend.
Current capability: 14nm linewidth (equivalent to 5nm-node logic, same league as the chip in your gaming PC)
Roadmap: 10nm linewidth = 2nm-node territory, i.e. competitive with what TSMC/Samsung are shipping in 2026-2027
Cost: a fraction of EUV. EUV photomasks cost $500K+ each. Canon's template-replica system spreads cost across 10-20 replicas per master.
Energy: no plasma source, no giant laser, way lower power draw. ESG line goes brrr.
The 2027 catalyst nobody is pricing in yet.
January 2026 Canon announced Inkjet-based Adaptive Planarization (IAP), which is a new tech for smoothing wafer surfaces between processing layers. They're commercializing it in 2027 for both logic AND memory manufacturing.
Translation: this is the missing piece that makes their whole stack viable for high volume manufacturing at advanced nodes. It's not "we'll get there someday".. there's a date on the calendar.
Texas Institute for Electronics (UT Austin consortium) already has a unit, running R&D.
Kioxia (formerly Toshiba Memory, world's #3 NAND maker) is publicly interested.
SK Hynix has been quietly evaluating NIL for 3D NAND for years.
Memory (NAND) is the beachhead. NAND is more defect tolerant than logic so it's the easier first customer. Once Canon qualifies at ONE memory fab, the floodgates argument becomes real.
Particle defects. NIL is contact-based. A single hard particle can permanently damage the template, which then prints that defect on every subsequent wafer. Canon needs particle adders below 0.001 per wafer pass to hit 1000-wafer template life. They've demoed 81 lots. Getting from demo to fab floor is the entire ball game.
Adoption timing. Even if the tech works, qualification at a real fab is 3-5 years. Meaningful revenue is 2-3 years after that.
NIL might get stuck in memory niches forever and never crack TSMC/Samsung. That's still a fine outcome at this entry price.
Canon is not a startup with negative cash flow praying for one product to work. It's a profitable diversified industrial with:
Cameras (still prints money, the camera "death" narrative was overblown)
Office printers (boring, recurring, high margin)
Medical imaging (growing)
Existing semiconductor equipment business
A dividend yield
Trades at a fraction of ASML's multiple
TL;DR
$ASML monopoly is a supply chain time bomb everyone refuses to defuse
Canon has the only credible alternative tech (NIL) and it's about to enter HVM
2027 IAP launch + memory fab qualification = catalyst window
Stock was oversold at all-time lows, I'm getting NIL as a free option
Diversified profitable business with a dividend means I literally get paid to wait
Geopolitics will tailwind this whether the market is paying attention or not
Lithography moats are 15+ year moats. This isn't a swing trade. This is a generational position.
Position
3000 shares until Kioxia announces HVM qualification or my wife's boyfriend kicks me out, whichever comes first.