Intel's High NA "First" Is Real—and It Pays Two Different Stocks for Two Different Reasons

Intel is now the first chipmaker in the world shipping high-volume processors made on the transition to ASML's most advanced lithography machine. In a joint announcement on July 15, the two companies said Intel Foundry had entered high-volume manufacturing on selected Intel 18A layers of its Panther Lake laptop chips using ASML's 0.55-numerical-aperture High NA EUV scanner, with yields matched to the incumbent 0.33 NA platform. The language of the release was collaborative—ecosystem partners "accelerating industry readiness" for a machine that is supposed to replace EUV for the rest of the decade. What the milestone actually is depends entirely on which of the two stocks you are holding.

Why the machine is the story, not the chip

The conventional reading treats this as an Intel technology win. It is worth slowing down on the machinery first, because the machine is where the economics bifurcate. High NA—0.55 aperture versus the 0.33 aperture of the NXE EUV tools ASML has sold for years—resolves finer features in a single exposure, cutting reliance on multi-patterning. It is also, by analyst estimate, roughly 2.5x the cost per exposure of the existing platform. That premium is the reason the industry does not agree about it, and it is the reason the collaboration is a two-market story rather than a unanimous roadmap.

The two markets are now visible. Intel has led from the start: it installed ASML's first-generation High NA tool (the EXE:5000) in Oregon in 2024 and accepted the second-generation EXE:5200B, and it has now converted production learning into a volume run. TSMC, by contrast, still patterns leading-edge chips with 0.33 NA multi-patterning, installed only evaluation tools in early 2026, and is expected to defer High NA until after its A14 node. Intel's bet is that two years of production learning—field stitching, defect management, slower dosing—compress its ramp risk on 14A, the first node designed around High NA, and let it beat TSMC on the economics of single-exposure patterning even at 2.5x the per-exposure cost.

That bet is what the July milestone is really proof of. It is not proof of unit demand. Intel dual-qualified High NA on a subset of 18A layers, meaning those wafers are interchangeable with its existing NXE fleet—a deliberate hedge that keeps the technology at the leading edge of the run rather than across all of it. High NA is the production learning curve, not yet the cost story.

A milestone that validates ASML's order book, delivered by one customer

The two-market split changes how the milestone should be read on each side of the collaboration.

For ASML, this is a supply-side validation event. ASML's growth plan rests on High NA: it has reaffirmed a 2030 revenue target near €60 billion ($71 billion), a level that requires the premium-priced scanner to be adopted by a broad customer base, not one. Intel's volume run is the proof point that makes that book credible—the first demonstrated instance that High NA works at production yields and that the 2.5x premium does not wreck the economics. But the near-term revenue is not coming from High NA. ASML has guided to roughly 30% capacity increases for low NA EUV and DUV immersion, and its raised full-year guidance to €43–45 billion rests on those mature lines. The Intel milestone underwrites the 2028–2030 revenue, not the current year, and it rests heavily on a single, unusually committed customer.

For Intel, the milestone is not demand but migration. The value converts only at 14A, and Intel has been explicit about the gate: it intends to build 14A capacity only against firm external foundry commitments in the 2026–2027 window. The external PDK for 14A reaches customers in October 2026, with risk production in 2028 and high-volume manufacturing in 2029. Everything the July announcement validates—Intel's ownership of the High NA learning curve—becomes worth money only if someone else signs up to pay for the node.

The valuation is doing the predicting

Both stocks are asking the market to pay today for that 2028–2029 conversion. The difference is how much they already demand.

ASML trades at a trailing P/E near 54x and around 42x EV/EBITDA—a premium explicitly underwritten by the 2030 High NA target. The Intel proof point supports that multiple; the concentration of High NA demand in one customer is the risk against it. Intel is the harder case. It remains unprofitable on an operating basis, spends on the order of $12 billion a year in capex against a negative free-cash-flow margin, and yet trades near $96, roughly 8.9x sales and a $500 billion market capitalization, after rising about 160% year to date from a 52-week low near $24. Substantially all of that re-rating is a bet on the foundry turn—on 14A credibility, external customers, and a profitable return on the capital Intel has been pouring into the transition.

The key issue is not whether the High NA milestone is real. It is. The question is who captures the value, and the data says the two companies are not equally positioned. ASML already owns the machine and the proof point; its exposure to High NA volume is real but spread across a decade and currently concentrated in one buyer. Intel owns the learning curve, but its payoff is deferred to 2028–2029 and gated by external commitments it has not yet demonstrated. The condition that resolves the trade is the same for both names: whether Intel's 14A learning advantage converts into firm foundry customers at the scale that pays for a node, and whether that pulls a second and third High NA buyer into ASML's book. Until then, the collaboration to "accelerate industry readiness" is supply-side machinery doing what supply-side machinery does—building the capacity and the proof point ahead of the demand that has yet to be signed.