The Exynos 9825 in the new Galaxy Note 10 is, as far as anyone can tell, the first commercial chip built with extreme ultraviolet lithography, on Samsung 7nm EUV. That is a milestone Samsung has been chasing hard and wants the credit for, because EUV is the most expensive bet in the history of chipmaking, and Samsung made it earlier and louder than anyone except TSMC, which is racing to the same finish line from the other direction.
EUV replaces the light. For years chipmakers have patterned circuits with deep ultraviolet lasers, and below a certain size that light is simply too coarse, so the industry has been faking finer features with multiple exposures stacked on top of each other, slow and expensive and error-prone. EUV uses light at a far shorter wavelength, fine enough to draw the smallest features in a single pass, which is the whole reason it matters. The catch is the machines, made by a single Dutch company, cost well over a hundred million dollars each and have spent a decade not quite working. Samsung is betting they work now.
Going early on EUV is the same instinct that got Samsung to FinFET first. The company cannot win a war of attrition against a bigger rival, so it tries to win the next thing before the rival fully commits, and EUV is the biggest next thing on the board. The leap to FinFET at 14nm and the clean 10nm generation that followed were the warm-up; 7nm EUV is Samsung planting a flag on the technology the entire industry will have to adopt eventually. Being first to the future is the whole strategy, restated in new light.
Samsung’s 7nm EUV is shipping in Samsung’s own Exynos before it shows up in any stack of marquee customer chips, which is the tell that the process is new and the yields are still climbing. First does not yet mean dominant. EUV in volume is brutally hard, the machines run hot and break and throw fewer good wafers per hour than anyone wants, and being first to ship one chip is not the same as being first to flood the market. The milestone is real; the business that is supposed to follow from it is still mostly a promise.
The prize for getting EUV right is the next several years of leading-edge logic, the mobile chips and eventually the giant processors that need every density gain they can get. Qualcomm and Nvidia and the rest of the fabless world will all need an EUV foundry, and right now there are exactly two credible ones, with TSMC booking better than half the foundry market to Samsung’s high teens. Turn first-to-EUV into reliable-at-EUV and Samsung holds its place at the front. Let the yields stay stubborn while TSMC’s keep improving, and first becomes a footnote, with the orders going wherever the good wafers come off the line.
EUV is the right bet, made early, and that is exactly the kind of move that has worked for Samsung before. Whether it works this time depends entirely on execution, on turning a hundred-million-dollar machine that barely functions into a production line that prints money. Perso, I think Samsung gets there, because it has more reason than anyone to make EUV work and the memory profits to absorb the pain of getting it wrong for a while. But the gap between announcing the first EUV chip and owning the EUV era is wider than the press release suggests, and TSMC is sprinting through it from the other side. Samsung got to the new light first. The harder question, the one the next few years answer, is whether it can make the light pay, and being early has never guaranteed that.