What happened?

According to a plan announced by ASML, Samsung and TSMC are preparing to move to High NA EUV lithography systems for advanced chip manufacturing. Samsung will use the technology in DRAM production from 2028, while TSMC will bring it into advanced process nodes starting in 2030.

That means that after Intel, which already uses High NA EUV, the world's two largest contract and semiconductor manufacturers will also make the jump to next-generation lithography.

What does High NA EUV mean?

High NA EUV is an upgraded form of existing extreme ultraviolet lithography systems with a higher numerical aperture. In practical terms: smaller and denser circuit patterns can be drawn on the silicon wafer. That makes it one of the methods manufacturers will reach for in advanced process nodes.

The companies are also running a separate joint effort: replacing today's 6-inch photomasks with larger 12-inch ones. That change could raise manufacturing efficiency while reducing the cost and complexity of advanced chips.

Why are the timelines so far out?

The dates 2028 and 2030 look distant at first. The reason: moving to new lithography systems does not end with installing machines in a fab. Production lines, photomasks and other processes all have to be made compatible with the new systems.

The difficulties Intel encountered bringing this technology into production show that the timetables Samsung and TSMC announce also depend on factors like production capacity and yield rates. These dates are a target rather than a commitment; they can slip according to the technical problems that emerge in the move to high-volume production.

Where does Intel stand now?

  • Intel is the company using ASML's High NA EUV technology in commercial production.
  • It draws on the systems in its 18A process, used to manufacture certain Core Ultra Series 3 (Panther Lake) processors.
  • According to ASML, Intel has produced more than a million wafers on 18A so far.
  • The company is also working on an improved process called 18A-P; Apple is reported to be evaluating it for its next-generation A-series processors.

The connection to AI is direct: it is no accident that Samsung will use High NA EUV in DRAM production first. The bottleneck for AI accelerators has long sat not in the processor but in the high-bandwidth memory feeding it. Smaller, denser circuit patterns mean capacity on the memory side directly.

A change of lithography generation is, in other words, the bottom layer of the compute debate. How many tokens a model can produce ends up depending on which machine can draw how fine a pattern onto silicon.

The real story here

The ordering is more striking than the numbers. In lithography the largest manufacturers usually move first; here it went the other way, with Intel carrying the technology into commercial production first and its two big rivals pointing to two to four years later.

That could mean a window of advantage for Intel — but it is also a risk. The company that moves to a new lithography generation first is the one that hits the yield problems first; Samsung's and TSMC's later timetables can also be read as a strategy of waiting for someone else to solve them.

That is why the report that Apple is evaluating 18A-P matters: if a customer like Apple moves onto Intel's process, the cost of going first will have paid off. If it stays at the evaluation stage, the price of the early move stays with Intel.