Author: Ada, Deep Tide TechFlow

According to industry research institution TrendForce, SK Hynix has completed the design verification of 375-layer NAND, with mass production planned to start by the end of 2026, as the company will convert existing production capacity. The completion of verification for SK Hynix's 375-layer NAND has pushed a long-awaited industry inflection point to the forefront, where tungsten, used in chips for nearly a quarter of a century, is being replaced by molybdenum. The real winners of this material substitution are not the storage manufacturers, but rather the upstream selling equipment and consumables.

Tungsten supported for nearly 25 years, scaling is forcing it to its physical limits

It is worth mentioning that the first to introduce molybdenum in metal wiring was Samsung, not SK Hynix. Samsung adopted molybdenum in its 286-layer ninth-generation NAND as early as 2024, which will enter mass production in April 2024, and is currently expanding the use of molybdenum to more process steps. This is the first time SK Hynix has used molybdenum in its product line, positioning it as a follower rather than an innovator in the industry context.

The 375-layer product itself also has a history of being downgraded. According to TheElec, SK Hynix initially aimed for a 400-layer level, but due to the complexity of manufacturing high-layer stacks, it was eventually downgraded to 375 layers. Even so, it still represents a key leap in SK Hynix's NAND roadmap, with the more advanced 480-layer and 604-layer products expected to rely even more on molybdenum.

Although the replacement of tungsten with molybdenum is not unique to storage, it has brought about an industry-level inflection point. Reports indicate that tungsten has been used as an interconnect metal in NAND, DRAM, and logic/foundry mid-process for nearly 25 years, but scaling requirements are now breaking through tungsten's limits, making molybdenum the most promising alternative candidate.

The advantages of molybdenum go beyond low resistance. Unlike tungsten and copper, molybdenum can prevent diffusion without a barrier layer, saving processes and improving yields; its high melting point and resistance to oxidation support direct deposition, making it more suitable for high aspect ratio structures like 3D NAND and GAA (gate-all-around) at the logic end. In other words, as the number of layers increases and nodes decrease, tungsten struggles more while molybdenum has greater penetration potential. This is fundamentally the basis for the "selling shovel" logic: once substitution is underway, the beneficiaries are those providing tools and materials.

Lam Research: The only shovel seller with mass-produced ALD molybdenum tools

The most direct and hard-hitting narrative in this line is Lam Research (LRCX). Its ALTUS Halo, launched in February 2025, is the industry’s first atomic layer deposition (ALD) tool to use molybdenum for mass production, with resistance improvements exceeding 50% compared to traditional tungsten metallization in most cases. Lam has disclosed that early adopters have been implemented in high-capacity 3D NAND factories and advanced logic factories in South Korea and Singapore, corresponding to Samsung and SK Hynix in South Korea and Micron in Singapore.

Business flexibility lies in the complexity of processes. According to Zacks Research, Lam is currently the only supplier with ALD molybdenum tools in mass production, serving foundry and NAND customers; although molybdenum deposition is slower and more complex, it allows Lam to expand its serviceable market (SAM) for single-wafer metal deposition at these advanced nodes to three times its previous size. Increased process difficulty translates to growth for equipment suppliers.

Entegris sells consumables, Micron is the only pure storage entity in the US stock market

Applied Materials (AMAT) is taking a different path. In February of this year, it launched the Spectral ALD system, replacing tungsten with molybdenum in today's transistor contacts, reducing resistance at this critical connection between transistors and copper wiring networks, and it has already been adopted by several leading logic foundries. It is important to distinguish that Lam's molybdenum story leans towards NAND/DRAM wordlines, while AMAT's molybdenum story focuses on 2nm GAA logic contacts; the downstream applications differ and should not be considered under the same benefit logic.

Representing the materials side is Entegris (ENTG). The company supplies molybdenum solid-state precursor, dichlorodioxomolybdenum (MoO₂Cl₂), customized for DRAM and 3D NAND, along with its ProE-Vap delivery system. Its logic is based on the chain effect of material changes. According to Entegris, the shift from copper and tungsten to molybdenum will affect multiple processes including precursor selection, polishing pad design, slurry formulations, etching materials, and filtering, leading to a more dispersed process that spans multiple steps.

As for the storage manufacturers themselves, neither Samsung nor SK Hynix are listed on the US main board, while Micron (MU) is the only genuine storage entity in the US stock market and has gained a position in molybdenum. Lam cites Micron's NAND development VP Mark Kiehlbauch as saying that molybdenum metallization enables Micron to achieve industry-leading I/O bandwidth and storage capacity in its latest generation of NAND products. However, Micron is a "user of molybdenum" rather than a "beneficiary from selling molybdenum," with its stock price primarily driven by the storage cycle and HBM, while molybdenum is just a performance-enhancing factor.

Will molybdenum miners benefit? Semiconductor demand is just a fraction

Theoretically, molybdenum metal miners could also be considered marginal beneficiaries of this narrative, with Freeport-McMoRan (FCX) in the US stock market corresponding to molybdenum as a byproduct of copper mining. However, the volume of molybdenum used in semiconductors is indeed very small. According to TheElec and industry estimates, Samsung procured about 4 tons of molybdenum last year, around 10 tons this year, and SK Hynix started with about 4 tons; the entire industry is expected to only reach 80 tons by 2030. Compared to the global molybdenum market, which mainly consists of steel alloys and consumes several hundred thousand tons annually, semiconductor demand is just a drop in the bucket. Connecting miner stock prices to the NAND narrative does not hold water.

This also defines the true focal point of the "tungsten out, molybdenum in" story: SK Hynix's 375 layers are just the starting point, with the real range spanning across NAND, DRAM, and materials across three major categories. In this metal transition, the segments selling tools and consumables have higher certainty, as storage manufacturers using molybdenum are the performance beneficiaries, not the valuation beneficiaries, while the metal side basically does not gain any advantage.

Statement: This article does not constitute any investment advice