On October 21, 2026, MLCC-related stocks staged a volatile intraday rebound. Huaci Co. hit the daily limit-up for the third time in six trading days, while Sanhuan Group, Jiemi Technology, Guoci Materials, Yunzhong Technology, and Fenghua Advanced Technology also gained. The catalyst was a Goldman Sachs research report, which said that MLCCs had become the third-largest cost item in AI server bills of materials, behind GPUs and memory chips.
This shift in ranking is more worth unpacking than the “rebound” itself. It means MLCCs are being revalued within the AI computing ecosystem—not merely as an inventory variable in the passive-components cycle, but as an unavoidable link in the chain of hard constraints on computing capacity. At the same time, the market action on October 21 points to another reality: the thesis is sound, but its impact is being felt unevenly.
AI servers are pushing MLCCs from “thousands of units” to “tens of thousands of units” and “tens of thousands of dollars”
A traditional server uses roughly 1,800 to 2,500 MLCCs, while an eight-GPU AI server has pushed that figure to around 20,000. At the rack level, the number rises further: estimates put MLCC usage in a GB300/Rubin rack at about 450,000 to 650,000 units, representing a value of roughly RMB 374,000 to RMB 540,000 per rack. The MLCC bill of materials (BOM) value per rack is also shifting from the $5,000 range to the $8,000 range, an increase of about 60% year over year.
The surge in usage is accompanied by a shift toward higher specifications. Rising GPU power consumption and demands for power stability and high-frequency delivery are pushing requirements to the limits of high capacitance, miniaturization, and reliability. GPU/HBM package decoupling, high-voltage DC power delivery, and high-frequency filtering in optical modules are the three key demand drivers. Standard MLCCs cannot meet these requirements, structurally narrowing effective supply. In the MLCC industry, “excess overall capacity” and “scarcity of high-end capacity” can coexist.
The China Business Industry Research Institute’s forecast provides an overall market framework: the global MLCC market is expected to expand from $23.44 billion in 2025 to $47.88 billion in 2030, a compound annual growth rate of about 15%. But structural shifts are the real marginal driver: the value growth of high-end MLCC demand related to AI servers is far outpacing the overall market. Estimates put the CAGR for AI-server MLCC value at over 85% from 2026 to 2030. Moderate growth in the overall market masks a sharp revaluation of the high-end segment within it.
Industry fundamentals are already coming through—not just “trading on expectations”
One key difference between this MLCC rally and pure concept-driven speculation is that performance and pricing signals from upstream manufacturers and materials suppliers are already corroborating one another.
In the first half of 2026, Sanhuan Group reported net profit attributable to shareholders of RMB 1.932 billion, up 56.18% year over year; Fenghua Advanced Technology reported RMB 290 million, up 74.01%; and Jiemei Technology’s net profit attributable to shareholders rose 39.87%. On the pricing front, Samsung Electro-Mechanics was first to adjust its Q4 2026 quotes: average prices for consumer-grade X5R products rose 25%–30%, while high-end X6S products for AI servers rose 10%–20%. Amid spillover demand for mid- to high-capacitance consumer-grade orders, other manufacturers are expected to follow with Q4 price increases of 10%–20%. Orders and shipments are also supportive: in June 2026, the BB ratios for Murata, Samsung Electro-Mechanics, and Taiyo Yuden reached 1.30, 1.31, and 1.25, respectively, with each recording its highest monthly shipments in nearly five years.
Capacity expansion is also progressing, but on a longer timeline. Sanhuan Group invested an additional RMB 560 million in its high-capacitance MLCC expansion project in the first half of 2026. Cumulative investment reached RMB 1.861 billion at the end of June, or 49.91% of the planned investment, and the project is expected to be ready for use by the end of May 2027. At Fenghua Advanced Technology’s Xianghe Industrial Park high-end capacitor base, cumulative investment has reached 98.56% of the budget. Capacity does not come online at the flip of a switch: the expansion cycle for high-end MLCCs and barriers to customer certification mean that tight supply will not ease within a single quarter.
Huaci’s three limit-up sessions in six days and Sanhuan Group’s gains are not the same thing
The market action on October 21 can easily be summed up as a “broad rebound in MLCC concept stocks,” but a closer look shows that the drivers differ.
The gains in Sanhuan Group, Fenghua Advanced Technology, Guoci Materials, and Jiemei Technology reflect industry players with revenue, profits, capacity expansion, and price pass-through. Their share-price upside comes from improving conditions gradually showing up in financial results, as well as the market-share opportunity from replacing imports in high-end products.
Huaci’s three limit-up sessions in six days, by contrast, have a much stronger thematic-speculation element. On September 15, the company had already issued a clarification and risk warning: MLCC-related powders sold to Chaozhou Sanhuan were still at the small-scale validation stage, with no volume shipments yet. In the first half of 2026, the entire advanced ceramic materials segment generated revenue of only RMB 6.64 million, or 1.15% of the company’s total revenue. The company’s traditional revenue remains heavily concentrated in daily-use ceramics, and factors including a sharp increase in EU anti-dumping duties and the removal of export tax rebates are expected to drive a further year-over-year decline in third-quarter results. Equating Huaci’s consecutive limit-up sessions with confirmation of an “MLCC industry trend” blurs the line between a thematic spike and a fundamental trend.
This does not mean the theme has no basis in short-term pricing, but it is a reminder: once “the third-largest cost item in AI servers” becomes a widely accepted narrative, capital spreads outward along the theme’s orbit, and stocks at the outer edge are often furthest from the core industry logic. The broader the narrative, the more important it is to examine companies’ revenue mix.
South Korea’s market on the same day: cheap is no reason to buy; the real problem is disappearing buyers
On October 20, the Korea Composite Stock Price Index (KOSPI) closed down 3.73%, its lowest close since April 7, after briefly falling below 5,500. JPMorgan’s assessment points to a fact about capital flows: South Korean stocks look quite cheap based on share prices and trading indicators, but buybacks—the main force supporting the market since late August—are fading. Samsung Electronics has completed its buyback program, while SK Hynix is nearing the end of its own.
This does not contradict the strength of the MLCC industry. South Korean memory-chip stocks are the “canary in the coal mine” for the AI trade. Their declines mainly reflect crowded AI positioning, rising U.S. long-term yields, and investor caution over excessive investment—not a systemic deterioration in companies’ earnings power. In fact, earnings forecasts for Samsung and SK Hynix are still being revised upward. When the buying that supports the market disappears, cheapness alone is not enough to trigger a rebound.
Looking at South Korean equities and A-share MLCC stocks on the same day offers a shared takeaway: the AI computing-power trend has not reversed, but capital is shifting from “broad-based spillover” to “concentrating on areas with the ability to deliver results.” This is also why the MLCC rebound has been choppy rather than a one-way rally: the overall market narrative provides direction, but the staying power of individual stocks ultimately depends on where they actually sit on the high-end MLCC revenue curve.
Key divergences to watch
The view that MLCCs are moving from “commodity components” to a “hard constraint on AI computing capacity” is supported by Goldman Sachs’ analysis of cost structures, the China Business Industry Research Institute’s market-size forecasts, and industry pricing and volume data. It is not a vague conceptual upgrade. But “the right sector” does not necessarily mean “the right stock.” There is a clear dividing line between the profit growth and capacity expansion progress of manufacturers such as Sanhuan Group and Fenghua Advanced Technology, and the validation stage and revenue contribution of upstream materials suppliers.
For stocks such as Huaci, which have been pushed higher at the outer edge of the sector’s thematic orbit, three limit-up sessions in six days reflect the spread of the narrative, not the weight of the MLCC business in the company’s financial statements. As Samsung Electro-Mechanics’ price hikes spread from consumer-grade products to higher-end models, and as Sanhuan Group’s expansion projects gradually reach full production in 2027, the companies most likely to be repriced are those whose high-end MLCC revenue has already entered a rapid-growth phase and whose capacity ramp-up is aligned with customer certification progress. A theme can drive consecutive limit-up sessions in a week, but industry trends are realized over quarters and years.
Disclaimer: The article’s content and data are for reference only and do not constitute investment advice. Investors who act on this information do so at their own risk.