Long-post re: Nikon Immersion Scanner Pricing.
This is probably the most insignificant aspect. The price is almost irrelevant to the commercial specs on these machines. Many fab demos have been done to compare ASML vs. Nikon, and as it stands, both are great if a fab is using one or the other.
The problem arises when you try to align a layer on a Nikon back to a layer patterned with an ASML. The hardware is different, and to achieve an ~1 nm scale pattern overlay, this makes all the difference. Even for a single vendor, there are subsets of tools in a fab that are matched, and product lots moving through the factory stay inside the matched set.
Nikon's immersion scanners are already priced a double-digit % below ASML's equivalent offering. And this makes no difference in Nikon's declining market share for immersion. At the top/right you can see their sales history and forecast—for 2026 their forecast is zero. Overlay is king, and even fractions of a nanometer difference equate to billions of dollars in yield loss over the lifetime of a fab part.
The success of Nikon's plan hinges on being able to match ASML's scanners that dominate the scanner fleets in today's fabs. They released this plan in their Q3 2025 report (bottom/right). They intended to develop a completely new immersion scanner platform that's designed to be matched with an ASML scanner.
Here are the two major changes to the tool design:
1) New Projection Lens: The masks used for today's Nikon scanner actually can't even be shared with ASML. The reason is the mask image is inverted differently through the lens and onto the wafer. They will redesign their lens to now match ASML mask mirroring.
2) New Wafer Stage: The overlay signature for the wafer stage hardware needs to be corrected. This is why different vendors can't be matched, and even different stages from the same vendor can't be matched well, leading to fabs using sets of scanners instead of allowing lots to use any scanner.
If they can pull this off and achieve reasonable matching to ASML, we will see many fabs again running demos to compare them head to head; if anything, to better negotiate VPAs with ASML. It's still an uphill battle; remember, even a full line of ASML immersion scanners are broken down into matched sets for yield reasons.
There are many tricks to achieve better overlay beyond scanner matching. Increasing the amount of metrology for alignment (before/inside the scanner) and overlay (after the image is patterned) is one of them. Nikon recently revealed their launch plan for the Litho Booster alignment metrology tool. This tool is part of their scanner-matching strategy and is another fab CapEx and cycle-time cost on top of the scanner price.
On China's domestic immersion scanners: We hear a lot of speculation that SMEE is shipping their immersion litho tools, and it's already "in production." Think about this tool-matching issue Nikon has and how that might play into the reality of these statements. Nikon already makes immersion scanners that are just as good as AMSL's, and yet they are shipping zero in a year when there is record demand for these machines.
In fact. SMEE is trying to make a scanner that was discontinued by ASML 15 years ago. What home does a scanner with these specs have in a fab? Every fab in China can still buy an NXT:1965Di directly from ASML—plug and play into its production line. What fab will eat the yield loss, allowing product to pass through a domestic tool? Why would a fab even install a tool that is the equivalent to a discontinued tool from another vendor?
It's easy to make comparisons to other fab tooling like etch and deposition and then speculate that the same thing applies to the scanner. But these don't have nearly the same degree of tool-matching needs the lithography bay requires.
Fun fact: Overlay is the #1 yield killer for every fab and every single chip made today. It's the only thing left in the defect Pareto after everything else is fixed.
半導体チップの保管期限は、アルミバッグの水蒸気透過率(WVTR: Water Vapor Transmission Rate)に材料定数やバリア厚さを放り込むと計算できます。期限過ぎても一度開封→ベークしてリセットもできます。この管理を怠るとリプロー密着性異常が起こり得るのでパワー半導体では致命的になります。