Dutch chip equipment maker ASML Holding NV (NASDAQ: ASML) broke ground on a new plant in Eindhoven last week, with CFO Roger Dassen noting that the AI boom has completely transformed customer sentiment. The company's existing €200 million EUV lithography machines, crucial for manufacturing advanced AI chips, are nearly sold out through 2027, while customers have already committed to adopting its next-generation High NA tools priced at around €400 million each.

"Can you give me more? Can you speed it up?" — that's how Dassen described the tone of a recent conversation with a major customer. TSMC, Samsung, and SK Hynix have all set timelines for beginning High NA use in mass production, while early adopter Intel says the machines have already met production and reliability standards. Combined with expansion plans, these developments suggest Europe's most valuable company is poised to extend its dominance in chip lithography into the 2030s.

"Honestly, I don't think we've reached the end of the AI boom yet," CEO Christophe Fouquet said by phone. From "questioning pricing" to "collective commitment," the outlook for AI chip demand remains robust.

ASML's customers span all major chipmakers, most notably TSMC, which manufactures chips for Apple and Nvidia. TSMC previously questioned whether High NA was worth the cost — around $400 million per unit — but has since committed to adopting the technology starting in 2030. For years, chipmakers evaluated whether advances in manufacturing techniques like chip packaging and 3D stacking could reduce the need for High NA. These commitments now show that the industry's biggest players still expect shrinking chip feature sizes to remain the essential path to improved performance. High NA can print features about 40% smaller than standard EUV systems, potentially reducing manufacturing steps and boosting efficiency — but at roughly twice the cost.

The trillion-dollar global semiconductor industry aligns around new tools, with investment decisions often made years ahead of mass production. Manufacturers, equipment suppliers, and materials companies must coordinate to ensure new machines, chemicals, and processes work together. Some experts previously questioned whether chipmakers would adopt High NA widely enough to recoup its development costs. "I never doubted it, because there's no alternative," said Jos Versteegh, an analyst at InsingerGilissen. Samsung says it will begin memory chip production with High NA equipment in 2028, a timeline Morningstar's Corionero notes is earlier than market expectations. Memory specialist SK Hynix has also announced High NA adoption starting in 2028; Micron has ordered equipment but hasn't set a production date. Intel, the technology's first adopter, said Tuesday (Sept. 8) it has processed more than 1 million wafers with High NA tools. Fouquet said memory makers have a lower barrier to High NA adoption due to their smaller finished chip sizes. "You'll see these players being more aggressive in many areas, including High NA," he said. "All players are advancing plans to bring High NA into volume manufacturing."

A more sweeping industry initiative was also disclosed alongside the timelines: pushing the shift from the six-inch photomask standard, in use for decades, to 12 inches. Current EUV equipment has a maximum exposure area of about 800 square millimeters per chip, and large data center chip designs from Nvidia, Google, and others are approaching this limit. High NA, using smaller masks, has even more constrained exposure areas. Expanding mask size would allow next-generation equipment to produce chips comparable to today's largest data center processors. "If we can pull this off as an industry, you'll see productivity gains of 40% on these machines," said ASML CTO Marco Pieters. The initiative plans a pilot line by 2031 and mass-production readiness by 2033; for ever-larger AI accelerators and GPUs, the transition would also avoid yield risks from stitching processes.

On the institutional side, Bank of America analyst Didier Scemama reaffirmed a "buy" rating on ASML, writing that these announcements align with his forecasts for High NA adoption (5 units this year, 6 next year, 20 by 2030). "While these orders don't change our financial forecasts, they significantly strengthen confidence that the supply chain is coordinating around a unified High NA roadmap." Barclays noted that commitments from the three major customers provide clearer adoption visibility, with "demand clearly strong." ASML shares have risen about 120% over the past year, and the company plans to expand EUV capacity by roughly 30% in 2027.

JPMorgan estimates ASML held a 94% share of the global lithography market in 2025. In the most advanced segment, its position is even stronger: since the late 2010s, ASML has held a monopoly on extreme ultraviolet (EUV) lithography machines, essential for printing the tiniest circuits in the most powerful commercial AI chips. High NA, first shipped in 2023, is the next generation of EUV technology. ASML has no commercial rivals in EUV. Japan's Nikon and Canon, along with emerging Chinese competitors, produce deep ultraviolet (DUV) lithography systems, which belong to an older technology generation. Morningstar analyst Javier Corionero dismissed concerns that these companies are about to catch up to ASML in lithography as "misinformation."

High NA EUV commercialization now has substantive data support. ASML senior vice president Greet Storms disclosed that as of mid-July 2026, global High NA EUV systems have cumulatively exposed over 1.35 million wafers. The certified EXE:5200B system achieved a throughput of 135 wafers per hour in acceptance testing, a 35% improvement over the previous-generation EXE:5000's 100 wafers per hour. Equipment stability is also improving. Global High NA fleet availability has risen from early-year levels to 84%, with ASML expecting 90% by Q4 2026 and a medium-term target of 93%. Intel is already using High NA in volume production of its Core Ultra Series 3 processors under the Panther Lake codename at the 18A process node, with its Oregon facility completing dual certification and achieving yields comparable to existing NXE platforms. Fouquet commented: "Some products you buy from Intel today are made with High NA equipment. This is a very important milestone."

Order data is equally striking. ASML disclosed in July that its required EUV equipment orders for 2027 are essentially fully booked, with a substantial number of forward orders for 2028 already secured. The company plans to increase capacity for Low-NA EUV and immersion DUV equipment by 30% annually in both 2027 and 2028. Full-year 2026 revenue guidance has been raised to €43-45 billion, a significant increase from previous expectations. InsingerGilissen's Versteeg put it bluntly: "I never doubted High NA would be adopted, because there's no alternative."

Just as markets digested the High NA long-term roadmap, an AI safety controversy erupted over the weekend. On Saturday, Sept. 12, Anthropic CEO Dario Amodei said the company would introduce more safeguards, including independent third-party assessments, and urged the industry to slow development of the most advanced models. OpenAI CEO Sam Altman voiced support, while xAI's Elon Musk responded "Dario is right." This rare consensus among the three AI leaders became a new flashpoint for chip stocks. In after-hours trading on Sept. 13 Beijing time, chip stocks collectively fell: SK Hynix and Intel dropped 4%, Micron, AMD, and SanDisk fell 3%, and Nvidia slid 2%. Market analysts noted that a slowdown in AI development could pressure chipmaker and supply chain stocks in the short term, but long-term impact may be limited given robust compute infrastructure spending. Gary Tan, a portfolio manager at Allspring Global Investments in Singapore, said: "This may create some short-term pressure, but is unlikely to derail the long-term AI trade. AI development is still in a relatively early stage, and I'm not sure other players in the AI ecosystem are willing to accept the current industry ranking and slow down while technology is still evolving so rapidly."

The safety debate forms a subtle hedge against ASML's long-term roadmap: on one side, AI leaders urge caution; on the other, chipmakers have written their capacity race commitments out to 2030 and beyond. Separately, Anthropic is in talks to bring Nvidia in as a cornerstone investor, aiming to raise up to $100 billion at a valuation possibly reaching $2 trillion, potentially the largest IPO ever — the giants' long-term bets on AI infrastructure remain undiminished by the war of words.

However, while executives talk about "slowing down" publicly, investment at the industry level has not paused. The Bank for International Settlements has warned that the world's five largest hyperscale cloud companies are expected to invest over $1 trillion in AI during 2025-2026, increasingly funded by debt. OpenAI projects its compute spending will reach approximately $750 billion by 2030, up more than 25% from an earlier forecast of around $600 billion. Oracle's Q1 capital expenditure hit $28.5 billion, exceeding its total quarterly revenue. The four major U.S. CSPs saw Q2 capital expenditure grow 86% year-over-year, and TrendForce has raised its full-year capital expenditure estimate for the nine major cloud providers to $886.7 billion. This disconnect isn't simple hypocrisy but reflects the structural dilemma facing the AI industry: no single lab is willing to bear the cost of slowing down alone. As Axios reported, more people in the AI industry are acknowledging that the pace of technological capability development may need to slow, but "no lab is willing to bear the cost of slowing down alone."