The multilayer ceramic capacitor (MLCC) sector has rapidly gained prominence in capital markets since NVIDIA unveiled its Rubin architecture in March. Last week, a market-cap-weighted index of ten passive component firms from China, Japan, and South Korea surged 35%, sparking widespread debate on whether "MLCCs could become the next memory market."

A recent report from Huatai Securities highlights that the Rubin architecture drives the MLCC value per server rack to jump from $3,000 in the H100 era to $22,000, with a potential rise to $40,000 by 2027, indicating a clear volume-and-price growth narrative. On the supply side, capacity expansion for high-end MLCCs is squeezing out production for general-purpose components, with price hike signals already emerging—a dynamic highly consistent with how HBM production crowds out standard DRAM capacity.

However, while high-capacitance MLCCs for AI servers exhibit a duopoly structure similar to HBM, the capital barriers are significantly lower than in the memory sector, presenting non-negligible medium- to long-term supply risks. Valuation-wise, the sector's median forward P/E ratio for 2026 has reached 43.3x, substantially higher than the memory sector's 8.7x. Future upward momentum must therefore shift from "valuation recovery" to "earnings realization."

Architecture Upgrades Fuel Volume and Price Growth, AI Server MLCCs Enter High-Growth Cycle

The core driver of the current MLCC rally is the substantial increase in value per server, fueled by continuous upgrades in NVIDIA's GPU architecture. According to SemiAnalysis, MLCC value per rack has risen from approximately $3,000 in the H100 era to $12,000 in the GB200 era, and further to $22,000 with the latest Rubin VR200 platform. As the Rubin Ultra platform enters volume production by 2027, the value per rack is projected to climb to $40,000.

Beyond rising value, the scale of demand is also expanding rapidly. The number of MLCCs per rack has increased from about 15,000 to over 90,000, with significant growth in demand for high-voltage, high-capacitance, and miniaturized products. The adoption of vertical power delivery architectures and the integration of new technologies like embedded components in substrates and silicon capacitors are further intensifying demand for high-end MLCCs.

Faced with stronger-than-expected order growth, market participants are turning notably more optimistic about the industry's growth cycle. The President of Murata stated in an investor meeting on May 27 that the peak in AI data center capital expenditure, previously expected around 2028, is now projected for 2030, suggesting a prolonged period of robust demand.

To meet future order requirements, Murata announced in May an additional ¥80 billion in capital expenditure over the next two years, focusing on expanding high-end, high-capacitance MLCC capacity in Japan. Overall, the escalating demand for high-end MLCCs from AI servers is propelling the industry into a new cycle of simultaneous volume and price growth.

High-End Capacity Squeeze Drives Widespread Price Increase Signals

The current wave of price hikes differs fundamentally from previous inventory restocking cycles. A single high-capacitance MLCC for AI consumes sintering capacity dozens of times greater than a general-purpose, low-capacitance component—a dynamic highly consistent with how HBM production crowds out standard DRAM capacity. High-end demand structurally erodes capacity for general-purpose products, rather than simply reflecting an overall demand increase.

Price increase signals are emerging across the board:

Murata: Raised prices for high-capacitance MLCCs used in AI servers by 15%–35% in April.

Taiyo Yuden: Followed suit, announcing price increases of 6%–13% for mid-to-low capacitance consumer and automotive MLCCs effective May.

SEMCO: According to Digitimes, also plans another price increase of 5%–10% for MLCCs.

TDK: The CEO highlighted AI data centers as a core investment direction in the April 28 earnings call, targeting a 25%–30% CAGR for AI-related business by fiscal year 2031.

Furthermore, Japan's Ministry of Finance April trade data shows trading companies accumulating MLCC inventories in Asian markets, a clear signal of forward demand.

Duopoly Structure, But Moat Depth Questionable

The market for high-capacitance MLCCs used in AI servers is currently dominated by Murata and SEMCO, which together hold approximately 90% market share—a concentration level similar to the HBM market (where SK Hynix and others hold about 80%). The President of SEMCO explicitly stated during an April event in Busan that capital expenditure is focused on AI server and automotive high-capacitance segments, aligning closely with Murata's strategy.

However, a key difference exists in competitive barriers. Building a new MLCC production line requires less than $1 billion, whereas constructing an advanced DRAM fab costs $15–20 billion. Although the 12- to 18-month product qualification cycle creates a short-term moat, the far lower capital barrier compared to memory chips represents the core risk for medium- to long-term supply-demand dynamics.

Additionally, beyond production lines, a hidden bottleneck exists: the supply of upstream high-end ceramic materials (nanoscale barium titanate dielectric powder). Domestic manufacturers are accelerating breakthroughs, with some already in the qualification phase, making the supply timeline for high-end ceramic powder a critical factor to monitor.

Premium Already Priced In, Focus Shifts to Earnings Delivery

The MLCC industry traditionally uses P/E valuation. According to FactSet consensus estimates, the sector's median forward P/E for 2026 is 43.3x, while the memory sector's is only 8.7x—indicating the market has already priced in a significant premium for MLCCs' AI-related value.

This implies that future upward momentum for the sector must shift from "valuation recovery" to "earnings realization," with the actual evolution of Average Selling Prices (ASP) becoming the key metric to watch.

The Huatai Securities report provides the following directional investment suggestions:

High-Capacitance AI Direction: Japanese and Korean leaders with higher shares in the NVIDIA supply chain are poised to benefit from rising ASPs for high-capacitance MLCCs in AI servers.

General-Purpose Price Increase Direction: Companies with significant exposure to general-purpose components warrant attention amid capacity constraints and the trend of rising prices for these parts.

A-Shares Localization Direction: The localization push in high-capacitance MLCCs and inductors continues, giving related domestic manufacturers long-term tracking value.