Author: Ada, TechFlow by Shenchao
According to industry research firm TrendForce, SK Hynix has completed the design verification of its 375-layer NAND, with mass production scheduled to begin by the end of 2026, utilizing existing production capacity. The completion of this verification brings a long-anticipated industry inflection point into focus: tungsten, used in chips for nearly a quarter-century, is being replaced by molybdenum. The true beneficiaries of this materials substitution are not the memory manufacturers, but rather the upstream suppliers of equipment and consumables.
Tungsten has held for nearly 25 years, and scaling is now pushing against its physical limits.
It is worth noting that Samsung, not SK Hynix, was the first to introduce molybdenum in metal wiring. Samsung adopted molybdenum as early as its 286-layer 9th-generation NAND, which began mass production in April 2024, and is now expanding the use of molybdenum to additional process steps. This marks SK Hynix’s first use of molybdenum in its own product line, placing it as a follower rather than an innovator in the industry.
This 375-layer product itself has a history of downward revision. According to TheElec, SK Hynix originally targeted 400 layers, but due to the manufacturing complexity of high-stack structures, it was ultimately revised down to 375 layers. Even so, it represents a significant leap on SK Hynix’s NAND roadmap, with the more distant 480-layer and 604-layer products expected to rely even more heavily on molybdenum.
Although molybdenum replacing tungsten is not unique to memory, it has triggered an industry-wide inflection point. Tungsten has been used as an interconnect metal in NAND, DRAM, and logic/foundry middle-of-line processes for nearly 25 years, but scaling demands are now pushing beyond tungsten’s limits, making molybdenum the most promising alternative.
Molybdenum’s advantages extend beyond its low resistivity. Unlike tungsten and copper, molybdenum eliminates the need for a barrier layer to prevent diffusion, simplifying processes and improving yield. Its high melting point and oxidation resistance enable direct deposition, making it better suited for high-aspect-ratio structures such as GAA (gate-all-around) in 3D NAND and logic devices. In other words, as layer counts increase and nodes shrink further, tungsten becomes increasingly strained, while molybdenum’s adoption potential grows. This is precisely the foundation of the “selling shovels” logic: once substitution gains momentum, the primary beneficiaries are the providers of tools and materials.
Lam Research: The only supplier with a commercially deployed ALD molybdenum tool
The most direct and compelling story in this line is Lam Research (LRCX). Its ALTUS Halo, launched in February 2025, is described by the company as the industry’s first atomic layer deposition (ALD) tool to use molybdenum in high-volume manufacturing, reducing resistance by more than 50% compared to traditional tungsten metallization in most cases. Lam has disclosed that early adoption is already underway in high-capacity 3D NAND and advanced logic fabs in South Korea and Singapore—with Samsung and SK Hynix in South Korea, and Micron in Singapore.
Business resilience lies in the complexity of processes. According to Zacks Research, Lam is currently the only supplier with an ALD molybdenum tool already in mass production, serving foundry and NAND customers; although molybdenum deposition is slower and more complex, it enables Lam to triple its serviceable addressable market (SAM) for single-wafer metal deposition at these advanced nodes. More complex processes translate into growth opportunities for equipment suppliers.
Entegris sells consumables, while Micron is the only pure-play storage stock on U.S. exchanges.
Applied Materials (AMAT) is pursuing a different path. In February this year, it launched the Spectral ALD system, replacing tungsten with molybdenum in transistor contacts—the critical connection between transistors and copper interconnects—to reduce resistance. This solution has already been adopted by several leading logic foundries. It is important to distinguish that Lam Research’s molybdenum story focuses on NAND/DRAM word lines, while AMAT’s molybdenum story centers on 2nm GAA logic contacts; their downstream applications differ, and they cannot be treated as benefiting from the same logic.
A representative on the materials side is Entegris (ENTG). The company supplies molybdenum solid-state precursors, specifically molybdenum dioxide dichloride (MoO₂Cl₂), customized for DRAM and 3D NAND, alongside the ProE-Vap delivery system. The rationale lies in the cascading effects of material switching. According to Entegris, the shift from copper and tungsten to molybdenum impacts multiple stages—including precursor selection, polishing pad design, slurry formulation, etch materials, and filtration—resulting in a more dispersed but end-to-end process.
As for the memory manufacturers themselves, Samsung and SK Hynix are not listed on the U.S. primary stock exchanges, while Micron (MU) is the only pure-play memory stock on U.S. markets and has gained an advantage in molybdenum adoption. Citing Mark Kiehlbauch, Vice President of NAND Development at Micron, Lam stated that molybdenum metallization has enabled Micron to achieve industry-leading I/O bandwidth and storage capacity in its latest-generation NAND products. However, Micron is a “user of molybdenum,” not a “beneficiary selling molybdenum,” and its stock performance remains primarily driven by memory cycles and HBM, with molybdenum serving only as a performance enhancement.
Molybdenum miners benefit? Semiconductor demand is just a fraction.
Theoretically, molybdenum mining companies could be marginal beneficiaries of this narrative, with U.S. stocks like Freeport-McMoRan (FCX), which produces molybdenum as a byproduct of copper mining, being relevant. However, the volume of molybdenum used in semiconductors is extremely small. According to TheElec and industry estimates, Samsung purchased approximately 4 tons of molybdenum last year and around 10 tons this year; SK Hynix started at about 4 tons, and total industry demand is projected to reach only about 80 tons by 2030. In comparison to the global molybdenum market, which consumes tens of millions of tons annually—primarily for steel alloys—semiconductor demand is negligible. Linking mining company stock prices to the NAND narrative is an unjustified causal assumption.
This also defines the true focal point of the “tungsten exit, molybdenum enter” narrative: Hynix’s 375-layer stack is merely the starting point, with the real scope spanning NAND, DRAM, and materials. In this transition of metals, the segments selling equipment and consumables offer higher certainty; memory manufacturers using molybdenum benefit in performance, not valuation, while the metal suppliers themselves gain little.
Disclaimer: This article does not constitute any investment advice.