The AI super-cycle is far from over, prompting a fresh look at safe-haven assets.
The semiconductor industry is closely watching a significant material shift. According to a recent report by The Elec, SK Hynix has successfully completed production verification for its next-generation V10 series 375-layer 3D NAND flash memory and plans to commence mass production at its Cheongju M15 plant in South Korea by the end of this year.
Originally dubbed the "400-layer" NAND internally, the final mass production layer count was adjusted down to 375 due to immense technical challenges in ultra-high-layer stacking processes, particularly the exponentially increasing difficulty of etching deep channel holes.
However, beyond the layer count adjustment, a more pivotal change has captured industry attention. This 375-layer NAND marks the first instance of introducing molybdenum (Mo) into the word-line metal gate, replacing the tungsten (W) thin film that has been the industry standard for over a decade.
SK Hynix's technological pivot is not an isolated case. Prior to this, other memory giants like Samsung Electronics and Micron have already laid out plans for products utilizing molybdenum. Leading global semiconductor equipment manufacturer Lam Research has explicitly stated that the switch from tungsten to molybdenum is the only viable path for the evolution of high-layer-count 3D NAND.
As industry leaders successively shift from tungsten to molybdenum, a clear signal is being sent: the tungsten material system, which has dominated the memory chip industry for over ten years, is reaching a replacement inflection point. Molybdenum has surged to become the core critical material enabling the realization of NAND flash memory with over 300 layers of ultra-high stacking.
In this semiconductor material revolution, why are global memory giants collectively turning to molybdenum? What irreplaceable advantages does it hold over the established conductive metal tungsten? How will this material substitution reshape the semiconductor materials supply chain and redraw the competitive landscape of the global industry?
Understanding the Shift to Molybdenum
To grasp the rationale behind "replacing tungsten with molybdenum," one must first understand the logic of 3D NAND technology evolution.
It is well known that 3D NAND flash increases capacity by vertically stacking memory cells. As the number of layers climbs, the number of word lines running through them surges, and the line width of these word lines is continuously compressed to nanometer-scale limits. Word lines are the core circuits connecting the control gates of memory cells, responsible for selecting and operating specific rows of memory cells. Their material properties directly determine the chip's signal transmission efficiency and storage density.
Reviewing the history of word line materials: the early solution was polysilicon. Due to its high resistance, the mainstream solution shifted to tungsten, which has lower resistivity, starting from the 64-layer and 96-layer generations. At that time, tungsten represented a victory at the material level, supporting the golden era of 3D NAND's leap from double-digit to triple-digit layers.
However, when layer counts break through the 300+ barrier, the inherent structural flaws of traditional tungsten materials—such as high resistivity, the encroachment of barrier layers on valuable space, and long-term reliability concerns—are laid bare.
Thus, in the current 300+ layer era, tungsten has completely hit its physical and process ceiling in high-layer-count NAND. The material红利 of this generation has been fully exhausted.
Tungsten's Limits and Molybdenum's Ascent Spark New Material Race
Meanwhile, molybdenum, which has long existed in the semiconductor field merely as an auxiliary material for sputtering targets, photomask blanks, and the like, was a niche metal with extremely low industry attention. Now, leveraging its unique physical and chemical properties, molybdenum is staging a comeback from a peripheral auxiliary material to a core functional material for high-layer-count memory chips.
Molybdenum is a refractory metal with a density about half that of tungsten, a melting point as high as approximately 2623°C, low thermal expansion coefficient, and excellent thermal conductivity. These characteristics make it naturally suited for high-density, high-heat, high-reliability chip manufacturing environments and have led to its widespread use in metallurgy, special alloys, and photovoltaics. In the semiconductor industry, it has undergone a complete transformation from a peripheral auxiliary to a core functional material.
From basic physical parameters, both molybdenum and tungsten are high-conductivity, high-melting-point metals. Their bulk resistivity differs very little—tungsten is about 5.28 μΩ·cm, molybdenum about 5.34 μΩ·cm—making their macroscopic conductive capabilities nearly equal. However, at the nanoscale, within microstructures like 3D NAND gates and contact holes, the performance gap between the two is dramatically magnified. This is the core reason high-layer-count flash memory is choosing molybdenum.
Within shrinking chip structures, tungsten's resistivity plummets as line width decreases and aspect ratios increase, leading to signal delay, increased chip power consumption, and aggravated heat generation. Molybdenum has a shorter electron mean free path; at the nanoscale, its resistivity increase is only about 60% of tungsten's, allowing it to maintain stable conductive performance over the long term.
Furthermore, when tungsten is used as a gate material, it must be paired with a TiN (titanium nitride) barrier layer to prevent metal diffusion and leakage. This auxiliary layer continuously occupies stacking space. In high-stack architectures like 375 or 400 layers, the additional barrier layer on each layer cumulatively occupies 30%-40% of the effective structural thickness, directly capping the potential for increased storage density. Molybdenum, with its excellent interface stability, requires no additional barrier layer. This means that under the same line width conditions, the effective conductive cross-section of a molybdenum word line is significantly larger than that of a tungsten word line. The resulting improvement in equivalent conductive performance far exceeds the impact suggested by simple resistivity comparison data. In multi-layer stacked structures, this directly saves substantial vertical physical space, creating room for increased storage density.
Additionally, significant differences exist in their compatibility with manufacturing processes. Traditional tungsten metal primarily relies on CVD (Chemical Vapor Deposition) for film formation. Facing the high aspect ratio channel structures of 3D NAND, often exceeding 40:1, CVD filling is prone to defects like voids and uneven films, directly lowering product yield. Molybdenum, in contrast, perfectly adapts to the mainstream ALD (Atomic Layer Deposition) technology of advanced processes. It offers superior filling uniformity, higher film flatness, and better conformity, perfectly matching the manufacturing requirements of ultra-high stacking architectures. Moreover, molybdenum adheres more strongly to insulating dielectrics like silicon dioxide and has better electromigration resistance, effectively reducing failure risks during long-term chip use and significantly improving product reliability.
Looking at the application history of molybdenum materials in the semiconductor industry, its development can roughly be divided into three stages.
In the early stage, molybdenum existed only as an auxiliary material, primarily used in non-core areas like semiconductor sputtering targets, photomask substrate materials, and packaging heat dissipation components. The market volume was limited, and industry attention was low.
As ALD deposition processes and high-purity metal purification technologies matured, and molybdenum precursors achieved commercial mass production, molybdenum began to make small inroads into scenarios like contact holes in logic chips and TSVs (Through-Silicon Vias) in advanced packaging, completing its transition from auxiliary to functional material.
The true breakout point is precisely the era where 3D NAND is moving towards 300+ layers of ultra-high stacking. With traditional tungsten materials hitting physical limits, molybdenum has taken the baton, becoming the preferred solution for word-line metal gates and officially entering the ranks of core semiconductor materials.
A semiconductor material iteration wave led by molybdenum has already begun. It will not only reconstruct the technological evolution path of 3D NAND but also has the potential to reshape the global semiconductor materials supply chain landscape in the future.
Beyond NAND: Molybdenum Opens Incremental Opportunities Across Semiconductor Scenarios
NAND: A Market Primed for Explosive Growth
As mentioned, NAND is currently the largest and most certain application market for molybdenum materials. With memory giants successively adopting it, the demand volume for molybdenum is rapidly increasing.
Industry calculations indicate that Samsung's molybdenum material procurement last year was about 4 tons, expected to increase to 10 tons this year. Based on the continued advancement of its technology roadmap, it is projected to reach 80 tons by 2030. SK Hynix will begin large-scale adoption of molybdenum processes next year, with an initial annual demand of about 4 tons. It's important to note that these procurement figures represent only the direct usage for word line processes. Considering broader applications like targets, the actual demand is even higher.
DRAM: The Next Growth Market Takes Shape
The application prospects for molybdenum materials in the DRAM field are also highly noteworthy. In fact, molybdenum precursor suppliers in the NAND field have already begun related layouts in mass production equipment. It is highly likely that DRAM will follow closely in introducing molybdenum materials.
The application of molybdenum in the HBM (High Bandwidth Memory) field is particularly noteworthy. HBM increases bandwidth by vertically stacking DRAM layers, with stacks already reaching 8 to 12 layers, and HBM4 specifications being even higher. In such high-density stacking scenarios, the shortcomings of tungsten—high resistance, fluorine residue, and filling difficulties—are magnified to the extreme.
In comparison, molybdenum's resistivity is 30% to 40% lower than tungsten's, requires no TiN barrier layer, reduces contact resistance by about 56%, and offers higher yield. Market information suggests that the molybdenum target usage per HBM chip is about 3 to 5 times that of ordinary DRAM, and the penetration rate of molybdenum in HBM4 is close to 100%. As Samsung, SK Hynix, and Micron fully transition to molybdenum word lines in their HBM3e/HBM4 products, demand for molybdenum in the DRAM field is rapidly catching up to NAND.
Long-Term Potential in Logic Chips
From NAND to DRAM to logic chips, a clear transmission path for molybdenum's application in the semiconductor field is forming.
In the logic chip field, molybdenum is being actively explored as a replacement material for copper interconnects. Copper interconnects face an exponential rise in resistivity at advanced nodes below 10nm due to surface scattering and grain boundary scattering. Molybdenum's electron mean free path is much shorter than copper's, making it less negatively affected by size effects at the nanoscale. Other research indicates that molybdenum and ruthenium (Ru) outperform traditional solutions in specific structures.
Industry expectations are that logic chips will begin gradually adopting molybdenum interconnect solutions within the next two to three years. This would propel molybdenum's market space from a niche application towards a global transformation in semiconductor materials.
From an investment perspective, the NAND sector represents the most certain opportunity window currently—memory giants' technology roadmaps are clear, molybdenum demand is showing exponential growth, and the process of domestic Chinese molybdenum target companies entering the supply chains of major memory manufacturers is accelerating, offering vast space for import substitution. In the medium term, molybdenum penetration in the DRAM and HBM fields is rapidly increasing and will become the next significant demand driver. In the long run, the transformation of interconnect solutions in logic chips will open even greater imaginative space for molybdenum.
Global Players Stake Their Claims, Supply Chain Value Reassessed
As "replacing tungsten with molybdenum" becomes an industry trend, the technology roadmaps and product iteration pace of global memory manufacturers are beginning to diverge. The upstream supporting supply chain—materials, equipment, and consumables—is also facing new market increments and a reshaped competitive landscape.
Looking first at memory manufacturers, Samsung's technology roadmap is quite clear: it began introducing molybdenum into the metal wiring process starting with its ninth-generation 286-layer 3D NAND, mass-produced in April 2024. Its tenth-generation 400+ layer product will be launched in the second half of this year, with the application scope of molybdenum materials continuing to expand. SK Hynix is following closely, with its 375-layer product set for mass production by year-end, to be followed by 480-layer and 604-layer products, indicating that molybdenum material penetration in the NAND field will continue to rise.
Micron is pursuing a dual-track strategy, applying molybdenum materials in both NAND and DRAM fields, exploring composite metal technology routes to differentially capture the advanced process market. In contrast, Kioxia and Western Digital are relatively conservative, currently still in the technology verification stage with no clear mass production plans yet.
Extending to the upstream supply chain, this material transformation is driving a reassessment of value across the entire semiconductor supply chain.
Within SK Hynix's supply chain system, France's Air Liquide, America's Entegris, and Germany's Merck have been identified as primary suppliers. South Korean domestic company SK Specialty is also actively entering the fray; discussions are underway regarding a plan for it to build supply capabilities by utilizing Air Liquide's distribution infrastructure.
Regarding equipment, according to disclosures, after evaluating equipment from Lam Research and Tokyo Electron (TEL), SK Hynix ultimately chose the latter's equipment. Lam Research's equipment uses a single-wafer processing method, handling wafers one by one. Tokyo Electron's furnace-type equipment can complete deposition for about 100 wafers in one batch, offering better cost-effectiveness in terms of equipment procurement cost, floor space usage, and molybdenum material consumption. Samsung has chosen Lam Research's deposition equipment to handle molybdenum materials.
Simultaneously, in the target field, demand for high-purity molybdenum raw materials and semiconductor-grade molybdenum targets is exploding. As 3D NAND layer counts continue to increase and application scenarios expand, the global market size for semiconductor-grade molybdenum materials is expected to expand by over 4 times between 2026 and 2028. Data shows that the global electronic-grade high-purity molybdenum target market reached sales of 7.752 billion yuan in 2025 and is projected to reach 13.20 billion yuan by 2032, with a compound annual growth rate of 7.9%, indicating massive incremental space. Domestic Chinese companies are accelerating their catch-up efforts and have achieved certain breakthroughs.
Secondly, molybdenum precursors, as core consumables, currently rely heavily on imports, making this a key sector for domestic material companies to tackle. Furthermore, demand for ALD equipment adapted to molybdenum processes continues to climb. Domestic equipment manufacturers are accelerating technology R&D and customer verification, hoping to achieve a technological leapfrog during this material iteration wave. Additionally, electronic chemicals supporting molybdenum processes, such as CMP slurries and specialized cleaning solutions, will also encounter entirely new incremental markets.
At the terminal application level, the performance improvements brought by molybdenum materials will also transmit downstream across all scenarios. For example, 3D NAND flash memory equipped with molybdenum gates can see read/write speeds increase by 20% to 30%, power consumption decrease by 15% to 20%, and single-chip storage density increase by over 30%. For AI servers and data centers, higher-density, lower-latency storage products can effectively alleviate storage bandwidth bottlenecks in high-computing scenarios. For consumer electronics like smartphones and tablets, they can support thinner and lighter terminal designs while significantly optimizing battery life, aiding in terminal product iteration and upgrades.
Overall, this round of material iteration presents a rare golden window for import substitution within China's semiconductor industry. Unlike the generational gap barriers in traditional process catch-up, molybdenum materials belong to a completely new technological track. The pace of industrial R&D and mass production is largely synchronized globally, with no absolute technological generation gap. Simultaneously, China possesses globally leading molybdenum resource reserves and a mature basic molybdenum industry cluster, providing a natural supply chain advantage.
Upstream, the industry can leverage local resources to tackle "choke-point" technologies like high-purity molybdenum purification and high-end precursors. Midstream, domestic ALD equipment can use this mass production wave to complete customer verification and rapidly achieve import substitution. Downstream, domestic memory manufacturers can simultaneously follow the molybdenum material technology roadmap. Therefore, there is potential to break free from the困境 of following development patterns and achieve a technological leapfrog.
Concerns and Challenges for Molybdenum's Mass Production
Although molybdenum's technical advantages comprehensively surpass traditional tungsten materials, transitioning from lab technology to scaled mass production still faces multiple industrialization barriers. This is a core reason why many industry players remain in the verification stage and have not yet embarked on large-scale mass production.
Industry experts point out that current core difficulties are concentrated in several dimensions: material purification, precursor preparation, process control, and production line adaptation.
Ultra-high purity purification presents a high barrier: Molybdenum materials used in core semiconductor processes require purity levels of 6N to 7N (99.9999% to 99.99999%). Trace impurities can cause chip leakage, performance degradation, and shortened lifespan. The global market for high-end, high-purity molybdenum raw materials and high-purity molybdenum precursors has long been monopolized by overseas giants like Merck and Air Liquide. Domestic traditional molybdenum enterprises mostly focus on industrial-grade products; the stability and consistency of high-end products still require continuous refinement.
Precursor delivery and control are difficult: Unlike gaseous tungsten hexafluoride, mainstream molybdenum precursors are solid at room temperature and cannot directly adapt to traditional gaseous delivery production lines. Production requires specialized equipment for high-temperature heating while precisely controlling material supply quantity and delivery rate. This places extremely high demands on production line hardware modification and精细化 process parameter control, resulting in high initial equipment investment costs.
Solid precursors inherently have disadvantages in thermal stability and feeding uniformity compared to gaseous or liquid precursors. Stable deposition of large-grain molybdenum films is crucial for integration success. The resistivity of small-grain molybdenum is as dependent on thickness as tungsten's, leading to significantly compromised performance.
Research institutions like imec have repeatedly warned: a significant gap exists between material bulk properties and actual device performance. The final electrical, thermal, and electromigration characteristics exhibited by molybdenum depend entirely on the grain size and grain boundary structure of the deposited film. Not just any "molybdenum" can achieve low resistance—the quality of the process solution determines the upper limit of performance.
Retrofitting existing production lines is costly: Existing memory production lines designed for tungsten CVD processes cannot directly adapt to molybdenum ALD deposition processes. Companies need to add new equipment and重构 process flows, facing significant upfront capital expenditure pressure.
Thin film process yield control is stringent: The thickness, uniformity, and adhesion of molybdenum ALD films are highly sensitive to parameters like chamber temperature, pressure, and gas flow. Slight parameter deviations can lead to batch product quality fluctuations, requiring long-term process accumulation and mass production refinement by companies.
Molybdenum ore supply and price volatility risks: As molybdenum usage in the semiconductor field rapidly climbs, bottleneck issues in upstream ore resource supply are becoming increasingly prominent. Molybdenum powder prices have already seen significant increases, and a supply-demand gap for semiconductor-grade target molybdenum is expected to persist. If demand rapidly expands while ore-end production expansion lags, sharp fluctuations in molybdenum prices could impact the cost structures of midstream target manufacturers and downstream chip makers.
From a global supply-demand perspective, molybdenum resource distribution is highly concentrated. If major producing regions face geopolitical or policy interference, supply chain security will be tested. This is both a challenge and a factor that further strengthens the investment logic for molybdenum material import substitution.
To address these barriers, the entire industry chain is progressively exploring solutions to circumvent technical risks and retrofit cost pressures, accelerating the industrialization and落地 of molybdenum materials.
It is also worth noting that "replacing tungsten with molybdenum" itself is not the终点 of technological evolution.
In the race for semiconductor materials, ruthenium (Ru) is another highly regarded direction. Ruthenium's resistivity is even lower than molybdenum's, but its cost and process waste issues severely limit the feasibility of large-scale commercial application.
If cost and process waste problems can be solved, ruthenium material remains a competitive challenger in high-end scenarios. imec fellow Tőkei has pointed out: molybdenum has better resistivity than tungsten and requires no barrier layer; compared to ruthenium, it is lower cost and has better adhesion.
More importantly, new material directions like topological semimetals are also快步 entering the research视野. Domestic research teams are already exploring chip manufacturing possibilities using two-dimensional materials like molybdenum disulfide. Topological semimetals like molybdenum phosphide exhibit resistivity even lower than copper in extremely thin nanowires, showing remarkable potential.
This means that while molybdenum has gained the initiative in this round of material revolution, the赛道 of semiconductor material competition is still extending. For industry participants, the current key is to translate molybdenum processes into product advantages as quickly as possible. For investors, while closely watching the molybdenum赛道, it is necessary to maintain forward-looking observation of potential future alternatives.
Final Thoughts
As semiconductor manufacturing approaches the edge of physical limits, the locus of innovation is gradually shifting from architectural design and process微缩 to fundamental breakthroughs in materials and processes.
Molybdenum's journey from the lab to the mass production line, its扩散 from a single Samsung production line to a whole-factory改造 at SK Hynix, and its progression from NAND word lines to DRAM HBM stacking and further exploration in logic chip interconnects, all signify that the strategic value of metallic materials is being reassessed across the entire semiconductor industry.
Traditionally, the industry has习惯 attributed chip performance improvements to transistor微缩 driven by Moore's Law. However, in today's era where 3D stacking has become mainstream and two-dimensional微缩 is逼近 limits, material revolution is becoming the key variable in延续 the semiconductor performance improvement curve.
Looking ahead, "replacing tungsten with molybdenum" is no longer a question of if it will happen, but rather a question of how quickly it will happen. Once this material transformation fully unfolds, which semiconductor critical material will be the next to take center stage?
Comments