Samsung deepened its partnership with ASML on next-generation chipmaking machinery, committing to put ultra-precise High-NA EUV tools into memory production by 2028 — first in the industry — while joining the push toward bigger photomasks.
The most expensive precision machine on earth just got its most important new customer.
Samsung Electronics has deepened its strategic partnership with Dutch lithography giant ASML around High-NA extreme ultraviolet lithography — the chip-printing technology that draws the finest circuit features in the history of manufacturing. The announcement, carried September 25 by The Indian Eye, is a declaration of intent: Samsung wants to lead the memory industry into the High-NA age.
The centerpiece commitment is a date: 2028. That is when Samsung plans to introduce High-NA EUV into high-volume DRAM production — the first in the industry, by its account. DRAM is the memory inside every data center, laptop and phone, and making it is a game of printing ever-smaller features ever more economically.
Why it matters is a matter of geometry. Higher numerical aperture — the "NA" in High-NA — means finer resolution. Finer resolution means fewer manufacturing steps to achieve the same density, and steps are money. Fewer steps, lower cost, a roadmap that stretches further into the future.
But machines this precise need new infrastructure around them. Samsung has also joined the industry initiative — originally launched with Taiwan's TSMC — to graduate photomasks from the 6-inch format that has served the industry for decades to 12-inch masks built for High-NA EUV. The timeline runs long: a pilot line by 2031, full production readiness by 2033.
A mask is the stencil through which the light shines; doubling its size is a fundamental retooling of the supply chain. The 2031–2033 horizon tells you how seriously the industry takes this transition: it is planning in half-decades, not quarters.
Meanwhile the race is no longer theoretical. Intel already runs High-NA EUV on select layers of its 18A process. Samsung's memory rival SK Hynix is also targeting 2028 for DRAM. And TSMC — long the technology's most prominent skeptic — has now scheduled volume use from 2030. The skeptics are aboard; the race is on.
Demand is running ahead of supply. ASML's existing EUV capacity was reported nearly sold out through the end of 2027, with NeoTeo reporting on September 15 that AI-driven chip complexity is accelerating every customer's expansion plans. The bottleneck of the AI era is not data, not talent, not even GPUs — it is the machines that print the machines.
The logic is cumulative. Every layer of AI sophistication demands more memory bandwidth; every memory generation demands finer lithography; every lithography generation demands rarer machines. Samsung's bet is that owning the transition early is worth the enormous capital outlay.
There is also a geopolitical undertow, and everyone in the industry knows it. High-NA EUV sits at the center of the world's most consequential supply chain — Dutch machines, Japanese and American components, Korean and Taiwanese fabs. Who masters the tooling masters the roadmap, and everyone else queues.
The next milestone is 2028. If Samsung delivers High-NA DRAM on schedule, it will have done something none of its memory rivals have: turned the most exotic machine in semiconductor manufacturing into a memory workhorse. If it slips, the race resets — and SK Hynix, Intel and TSMC will be waiting.
Western coverage — Zacks, TechTimes — frames the partnership as the commercial validation of ASML's biggest bet.
The emphasis: ASML's existing EUV tools are nearly sold out through 2027, and High-NA is the next growth engine. Samsung's 2028 DRAM target gives the technology a memory-industry beachhead, not just a logic one — expanding the addressable market.
The caution, rarely stated aloud: these machines cost a fortune, shipments are measured in units per year, and the economics only work if the memory market cooperates. High-NA is a masterpiece; masterpieces are expensive.
The Western read of the mask initiative: 2031 pilot, 2033 volume — a sign the ecosystem is maturing exactly on schedule for a technology this young.
Eastern coverage — Samsung Newsroom, Korean and Japanese trade press — treats the deal as industrial strategy as much as technology.
The emphasis: Samsung is not merely buying machines; it is co-developing the future of memory manufacturing with the one company that can deliver it. "Industry first" is both a claim and a warning to rivals.
For Korea, the partnership is sovereignty-adjacent: in an era when chip supply chains are geopolitical assets, owning the lithography relationship is owning the memory roadmap. SK Hynix's parallel 2028 target sharpens the domestic rivalry.
MarkLines and the Japanese supply-chain press note the secondary effect: a 12-inch mask transition means new tooling orders for the entire lithography ecosystem — a multi-year equipment super-cycle.
The South — The Indian Eye, regional tech press — reads the announcement as a map of who builds the AI era and who buys it.
The emphasis: the value chain of artificial intelligence runs through Veldhoven, Suwon and Hsinchu. The countries racing to build AI data centers — India prominent among them — are downstream of these partnerships, buyers of the memory the machines print.
The question is whether any of the value escapes upstream. India's semiconductor ambitions are precisely an attempt to stop being a permanent customer; Samsung's announcement is a reminder of how steep the ladder is.
The South's quiet hope: every memory generation that gets cheaper also makes AI cheaper to run everywhere else. The machines print in Korea; the benefits, eventually, print everywhere.
