Will MLCC become the next memory?

Will MLCC become the next memory?

```

Since Nvidia’s Rubin architecture was released in March this year, the MLCC (multilayer ceramic capacitor) sector has rapidly become the focus of capital markets. Last week, the weighted weekly increase of the market value for 10 passive component companies from China, Japan, and Korea reached 35%. The market has begun to discuss whether "MLCC will become the next memory."

According to the latest report from Huatai Securities, the Rubin architecture has driven the MLCC value per cabinet from $3,000 in the H100 era to $22,000, and it is expected to reach $40,000 in 2027, with both volume and price rising clearly. On the supply side, expansion of high-end MLCC production is squeezing out ordinary product capacity, and price increase signals have already landed in succession. The logic is highly consistent with HBM crowding out DRAM.

However, while high-capacity MLCCs for AI servers show a dual oligopoly pattern similar to HBM, the capital barriers are far lower than memory, and mid- to long-term supply risks should not be ignored. In terms of valuation, the sector’s expected median PE for 2026 has reached 43.3 times, significantly higher than the memory sector’s 8.7 times. Subsequent momentum must shift from “valuation correction” to “profit realization.”

Architecture upgrade drives both volume and price, MLCC for AI servers enters a high boom cycle

The core driving force of this round of MLCC market activity comes from the continuous upgrade of Nvidia GPU architecture, greatly increasing the per-unit value. According to SemiAnalysis estimates, MLCC value per cabinet has risen from about $3,000 in the H100 era to $12,000 in the GB200 era, and the latest Rubin VR200 platform has further increased this to $22,000. With the Rubin Ultra entering mass production in 2027, MLCC cabinet value is expected to climb further to $40,000.

Not only has value risen, but demand is also rapidly expanding. MLCC usage per cabinet has increased from about 15,000 pieces to over 90,000 pieces. Demand for high-voltage, high-capacity, miniaturized products has grown significantly. The widespread adoption of vertical power architecture and the introduction of new technologies such as embedded components and silicon capacitors at the substrate level have further intensified the demand for high-end MLCCs.

Faced with greater-than-expected order growth, market sentiment toward industry prosperity has turned markedly optimistic. Murata's president stated at an investor meeting on May 27 that the peak in AI data center capital expenditures, previously expected in 2028, has been delayed to 2030, meaning the high boom period for demand will last longer.

To meet future orders, Murata announced in May it will add 80 billion yen in capital expenditures over the next two years, focusing on expanding domestic Japanese production capacity for high-end, high-capacity MLCCs. Overall, the continued upgrading of AI server demand for high-end MLCCs is pushing the industry into a new boom cycle characterized by both volume and price increases.

High-end capacity squeezing drives price hike signals intensively landing

This round of price hikes is fundamentally different from previous inventory replenishment cycles. Each high-capacity MLCC for AI consumes dozens of times the sintering capacity of general-purpose low-capacity products—this is highly similar to how HBM squeezes out ordinary DRAM capacity: high-end demand forms structural erosion on general capacity, rather than simply growing overall demand.

Several price hike signals have landed:

  • Murata: In April, increased prices for high-capacity MLCCs for AI servers by 15%–35%;
  • Taiyo Yuden: Followed up and announced price increases for mid-to-low-capacity consumer and automotive MLCCs by 6%–13% starting May;
  • SEMCO: According to Digitimes, also plans to raise MLCC prices again by 5%–10%;
  • TDK: Its president listed AI data centers as a core investment direction at an earnings meeting on April 28, targeting 25%–30% CAGR for AI-related business by fiscal 2031.

In addition, Japanese Ministry of Finance trade data from April shows trading firms are stockpiling MLCCs in Asian markets, a clear sign of demand moving forward.

Dual oligopoly dominance, depth of moat questionable

High-capacity MLCCs for AI servers are currently dominated by Murata and SEMCO, with a combined market share of about 90%, similar to the HBM market (SK Hynix and others totaling about 80%). SEMCO’s president stated at an April Busan event that capital expenditures are focused on AI server and automotive high-capacity segments, mirroring Murata’s direction.

However, competitive barriers between MLCC and memory are fundamentally different. A new MLCC production line costs less than $1 billion, while a DRAM advanced Fab single plant costs $15–$20 billion. Despite 12–18 months of product validation forming a short-term moat, the capital threshold is much lower than for memory chips. This is the main mid–long term risk for supply and demand balance.

Additionally, there are hidden bottlenecks outside the production line—the supply of upstream high-end ceramic materials (nano-level barium titanate dielectric powders). Domestic manufacturers are accelerating breakthroughs, and some are entering validation. The supply rhythm of high-end ceramic powders warrants close tracking.

Premium fully reflected, focus shifts to profit realization

The MLCC sector traditionally uses PE for valuation. According to Factset consensus, the sector’s expected median PE for 2026 is 43.3, while the memory sector is only 8.7—the market has already fully priced in MLCC's AI asset value.

This means the sector’s subsequent upside will shift from “valuation correction” to “profit realization.” The evolution of ASP (average selling price) will become the core observation indicator.

Huatai Securities’ report gives the following directional investment layout suggestions:

  • AI high-capacity direction: The leading Japanese and Korean companies in Nvidia’s supply chain are expected to continue benefiting from rising ASP for high-capacity MLCCs in AI servers;
  • General product price increase direction: Under trends of capacity contraction and price hikes for general products, firms with higher share of general-purpose products merit attention;
  • A-share localization direction: The localization process in high-capacity MLCC and inductor fields continues to advance, and related domestic firms have mid–long term tracking value.

Risk Warning and DisclaimerThe market involves risks and investment requires caution. This article does not constitute individual investment advice, nor does it take into account the specific investment targets, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article suit their specific circumstances. Investment based on this article is at your own risk. ```