Two of the world’s largest chipmakers put their signatures on a machine that costs roughly $400 million each, and in doing so settled a debate that has divided the industry for years. ASML said this week that Samsung will bring High NA extreme-ultraviolet lithography into DRAM mass production starting in 2028, and TSMC will adopt it for advanced logic from 2030. Intel has already taken delivery of the equipment.
Samsung’s commitment covers two machines reported at about 1.1 trillion won, around $773 million, a figure that hints at the economics of the technology. High NA is the next step in EUV, the printing technique that carves the finest features onto silicon, and it has been adopted cautiously because of its cost and its limitations.
The most consequential part of the announcement was not the orders but a change to a standard that has held for decades. ASML, Samsung, and TSMC said they will jointly push the industry’s photomask standard from six inches to twelve, with the new size expected to enter advanced mass production around 2033. Marco Pieters, ASML’s chief technology officer, said the larger mask raises High NA throughput by 40 percent and removes stitching constraints.
The mask change attacks a physical problem directly. High NA’s projection field is half the size of the previous generation, which means large chips must be printed in pieces and stitched together, a process that slows production and invites defects. A twelve-inch mask lets more of a chip be printed in a single exposure, effectively removing a layer of complexity from the production line.
TSMC chief executive C.C. Wei framed the collaboration in his own terms, saying the work solves technical barriers ahead of economic ones. The statement is a nod to the sequence in which these problems actually bite: first the physics has to work, then the cost has to come down. The mask initiative attacks the first, and the volume commitments attack the second.
SK Hynix is watching from the sidelines, according to the announcement, still evaluating whether to join the mask effort. Its absence is notable. The Korean memory maker is Samsung’s direct rival in the DRAM market where High NA will first be used at scale, and its decision will help determine how quickly the technology becomes the industry standard rather than one company’s bet.
The timeline reveals the pacing of the transition. Samsung will use High NA in DRAM from 2028, TSMC in logic from 2030, and the twelve-inch mask arrives in volume around 2033. The staggered schedule suggests the industry is treating High NA as a controlled migration rather than a rush, with each step gated on the previous one proving out.
ASML stands to benefit twice. It sells the machines, and it sells the ecosystem that surrounds them, from masks to services. The mask standard it is helping to rewrite will shape the equipment market for a decade, and ASML’s central role in setting that standard is a competitive advantage no rival lithography supplier currently holds.
The cooperation is unusual in an industry defined by rivalry. Samsung and TSMC compete for the same customers, and Intel is a third pole. That all three agreed to coordinate on a mask standard, with ASML as the convener, signals that the cost of going it alone now exceeds the benefit of keeping secrets.
For TSMC, the logic commitment comes as its AI-driven demand strains capacity. High NA is a longer-term bet on the nodes after the current generation, the printing technology that will define the 2-nanometer era and beyond. Wei’s framing that technical barriers come first reflects a company that wants the physics settled before it commits billions.
The price of the machines explains the caution. At roughly $400 million each, High NA systems are the most expensive tools in the fab, and their throughput has to be high enough to justify the capital. The 40 percent throughput gain from larger masks is precisely the kind of number that makes a skeptical finance committee approve a tool this expensive.
Intel’s early position gives it a head start that its rivals are now matching on their own schedules. The company took delivery of High NA systems first, part of its attempt to leapfrog in manufacturing. Whether that advantage survives Samsung’s 2028 entry and TSMC’s 2030 entry depends on how much Intel can produce in the interim.
The commitments give ASML a booked future that stretches years out, smoothing the boom-and-bust cycles that have defined lithography demand in past cycles.
The week’s announcements amount to an industry voting with its checkbook. High NA is no longer a question of whether but when, and the answer now has dates attached. The mask standard is the quieter change, but it may matter more: a twelve-inch photomask reshapes how chips are printed for the next decade, long after this year’s orders are delivered and forgotten.
