ASML (NASDAQ:ASML | ASML Price Prediction) has positioned itself in the enviable position of being the only equipment supplier capable of manufacturing lithography systems that can delineate patterns on semiconductor wafers at the 5nm node and below. ASML has a 100% share of the EUV lithography market, an 85% share of the ArF DUV immersion market, a 70% share of the ArF DUV dry market, an 80% share of the KrF DUV market, and a 25% share of the i-line lithography market, according to The Information Network’s report entitled “Sub-100nm Lithography: Market Analysis and Strategic Issues.”
EUV, or Extreme Ultraviolet lithography, uses extremely short wavelength light to print extremely small circuit patterns onto semiconductor wafers. Without EUV, it becomes increasingly difficult to manufacture advanced AI processors and advanced memory chips at smaller process nodes with acceptable yields and power consumption.
The ASP, or average selling price, of a Low-NA (numerical aperture) EUV system in 2026 is estimated at approximately €215 million, while an advanced High-NA EUV system can cost roughly €450 million. This article focuses on Low-NA EUV systems because they are becoming increasingly important in advanced DRAM manufacturing and High Bandwidth Memory (HBM) production.
HBM is a specialized form of DRAM memory designed for artificial intelligence accelerators such as Nvidia (NASDAQ:NVDA) GPUs. Unlike conventional DRAM, HBM stacks multiple memory dies vertically, dramatically increasing memory bandwidth while reducing power consumption. As AI models become larger and more complex, HBM is becoming one of the most critical components in AI servers because GPUs increasingly depend on moving enormous amounts of data between processors and memory at very high speed.
The future growth of ASML increasingly depends on memory manufacturing rather than logic foundry spending as Samsung Electronics (NASDAQ:SSNLF)), SK hynix, and Micron Technology (NASDAQ:MU) compete aggressively for HBM leadership in the AI infrastructure market.
HBM Growth Is Driving A New DRAM Technology Race
I show in Table 1 that the HBM market is projected to expand from approximately $17.4 billion in 2024 to nearly $60 billion in 2026. That growth is forcing Samsung, SK hynix, and Micron into an increasingly aggressive technology race centered on advanced DRAM scaling.
Unlike prior DRAM cycles that relied primarily on wafer capacity expansion, the current HBM cycle increasingly depends on tighter pitches, higher layer complexity, lower power consumption, and improved yields. As a result, EUV lithography is becoming one of the most important competitive tools in advanced memory manufacturing.
HBM differs significantly from conventional DRAM used in PCs and servers. Traditional DRAM places memory chips side-by-side on a module and communicates with the processor through relatively narrow data channels. HBM, by contrast, vertically stacks multiple DRAM dies using through-silicon vias (TSVs) and places the memory extremely close to the GPU or AI accelerator using advanced packaging technologies.
This design dramatically increases memory bandwidth while simultaneously lowering power consumption and reducing latency. Those characteristics are becoming increasingly important in AI servers because modern AI workloads require processors to move enormous amounts of data continuously between memory and GPUs (Graphical Processor Units). As a result, HBM is becoming one of the primary performance bottlenecks — and competitive differentiators — in advanced AI systems.
SK hynix Established An Early HBM Lead
I show in Table 2 that SK hynix established a dominant early lead in HBM production during the same period in which the company accelerated EUV adoption more aggressively than Micron. SK hynix expanded HBM output from 0.06 billion GB units in 2022 to a projected 3.32 billion GB units by 2027, while Samsung increases from 0.03 billion GB units to 2.54 billion over the same period.
GB refers to gigabytes of memory shipped, which is one measure of total HBM output volume. The rapid increase in HBM gigabyte shipments reflects the enormous amount of memory required by modern AI accelerators and AI servers.
Micron, grouped within “Others” during the earlier years of the HBM cycle, remained materially behind both competitors as it adopted EUV lithography more conservatively.
Micron’s Catch-Up Phase Could Benefit ASML
I show in Table 3 that Samsung and SK hynix adopted Low-NA EUV lithography significantly earlier than Micron, allowing both companies to scale advanced DRAM manufacturing more aggressively during the early stages of the HBM expansion cycle.
SK hynix established an early lead in HBM after producing the industry’s first HBM memory in 2013, while the first commercial devices using HBM were Advanced Micro Devices (NASDAQ:AMD) Fiji GPUs in 2015. Micron, by contrast, initially pursued an alternative memory architecture known as Hybrid Memory Cube (HMC) before later shifting toward HBM and more aggressive EUV insertion.
Micron’s slower EUV adoption coincided with a later transition into large-scale HBM production. While Samsung and SK hynix expanded EUV insertion earlier to accelerate advanced DRAM scaling and HBM output, Micron remained more conservative on EUV adoption and used essentially no Low-NA EUV systems in 2021 and 2022.
Importantly, I show in Table 3 that Micron’s projected Low-NA EUV demand accelerates sharply after 2024, increasing from just one system in 2023 to a projected eight systems by 2027 as the company expands advanced HBM production and narrows portions of the technology gap with Samsung and SK hynix.
Investor Takeaway
For years, investors associated ASML primarily with logic foundry transitions at companies such as Taiwan Semiconductor Manufacturing (NYSE:TSM) (TSMC). That relationship remains important, but the AI infrastructure cycle is now creating a second major growth engine centered on advanced memory manufacturing.
HBM is rapidly becoming one of the most critical components in AI servers because GPUs increasingly require enormous amounts of very high-speed memory to process and move AI data efficiently. As AI models become larger and more complex, memory bandwidth is becoming just as important as raw GPU performance. That trend is forcing Samsung, SK hynix, and Micron into an increasingly aggressive competition involving tighter DRAM pitches, greater layer complexity, improved yields, and higher EUV intensity.
Importantly, ASML benefits regardless of which DRAM manufacturer ultimately wins the HBM race because all three companies are being forced toward greater use of Low-NA EUV systems. At approximately €215 million per Low-NA EUV system, even relatively small increases in tool demand can materially impact ASML’s revenue growth. As Micron accelerates its catch-up phase against Samsung and SK hynix, the AI memory war may become one of the semiconductor equipment industry’s most important long-term growth drivers.
Top Gaining Stocks
Top Losing Stocks