--- title: "Goldman Sachs has found the \"next storage\": MLCC" type: "News" locale: "en" url: "https://longbridge.com/en/news/288228768.md" description: "Goldman Sachs research report points out that multilayer ceramic capacitors (MLCC) may become the \"next supply bottleneck\" in the AI industry chain. In the cost of AI servers, MLCC has risen to the third largest item. Its market size is experiencing explosive growth at a compound annual growth rate of 80%, and it is seen as a potential target with the ability to replicate the market conditions of memory chips" datetime: "2026-06-01T13:31:57.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/288228768.md) - [en](https://longbridge.com/en/news/288228768.md) - [zh-HK](https://longbridge.com/zh-HK/news/288228768.md) --- # Goldman Sachs has found the "next storage": MLCC **Financial Associated Press, June 1 (Editor: Xiao Xiang)** Although NVIDIA, as a pure AI enabler, and various hyperscale cloud vendors have been among the first beneficiaries of this new AI supercycle in recent years, it has become increasingly noticeable that the "most explosive" winners may not be those purely AI concept stocks, but rather the "shovel sellers" or infrastructure suppliers that directly address the core bottlenecks of AI implementation. **The evolution of market trends can confirm this:** Initially, attention was focused on the data center sector, which then spread to infrastructure stocks, energy, and nuclear power; since September of last year, the storage chip industry—originally characterized by strong cycles and commoditization, prone to extreme volatility—has exhibited an astonishingly favorable trend in this cycle. The AI infrastructure hierarchy chart provided by Goldman Sachs' thematic investment team clearly indicates that the phased mismatch in the supply chain is continuously transforming into corporate pricing power and stock price ammunition. Regarding the various rack components that NVIDIA is set to adopt in the next-generation Vera Rubin architecture, Financial Associated Press has recently cited Morgan Stanley's research report for a more in-depth analysis in past articles. **It is evident that the core question the market is most urgently seeking to answer is: Who will be the "next supply bottleneck" in the AI industry chain?** In other words, which sub-sector is likely to replicate or even surpass the previous "Davis Double Play" trend of storage chips? **In this regard, Goldman Sachs believes the answer may lie in the unassuming and mundane capacitors—more specifically, multilayer ceramic capacitors (MLCCs).** **Why are MLCCs becoming a "must-have" in the AI wave?** **Goldman Sachs analyst Nelson Armbrust pointed out in a recent research report that MLCCs have now risen to become the third largest cost item in the current AI server bill of materials, following GPUs and storage chips.** Currently, the overall market size for MLCCs is approximately $15 billion, with the market size in the AI server sector being $1.3 billion, and it is experiencing a robust explosion at an 80% compound annual growth rate, while growth in other key industries such as automotive and mobile phones has slowed. From a technical perspective, MLCCs function as a miniaturized, ultra-fast energy buffering library. Compared to traditional batteries characterized by "high energy density and slow release," MLCCs exhibit extreme performance with "micro energy and millisecond-level ultra-fast discharge." **They primarily serve two core functions in circuits:** > Voltage stabilization and power smoothing: Instantaneously absorbing voltage pulse surges and filling in drop gaps, ensuring a pure and stable current supply for highly sensitive high-performance chipsNoise filtering: Blocking unexpected electrical "noise" (interference) that may disrupt digital transmission data. In the actual operation of AI infrastructure, the data processing of AI servers is far from a smooth linear flow; instead, it exhibits extreme pulsing characteristics. When AI models perform massive concurrent computations, processors instantaneously trigger peak current demands within microseconds; and when the computations wind down, the load demand drops to zero in an instant. Traditional power management modules are simply unable to respond to such high-frequency and drastic load variations. MLCCs are positioned close to AI chips and can immediately release power to the chips when needed, thus preventing server crashes. Given that core computing chips like those from NVIDIA have a vast parallel computing system, their demand for capacitor networks increases exponentially. It is estimated that a single standard high-end AI server rack requires as many as 600,000 MLCCs, each performing its role and working in coordination, to ensure the overall system stability of the computing base. **Who are the leading players in this field currently?** **In terms of competitive landscape, for the "low-voltage high-capacity" MLCCs adjacent to GPUs/ASICs in AI servers, technological iterations are evolving towards the dual dimensions of "achieving extreme miniaturization and ultra-high capacitance within extremely limited PCB space," significantly raising the entry barriers.** Currently, **Japan's Murata, South Korea's Samsung Electro-Mechanics (SEMCO), and Japan's Taiyo Yuden** are the three major global suppliers in this field, and they are expected to continue benefiting from the demand dividends brought about by the expansion of computing power, thereby gaining pricing power to further drive up prices. On the other hand, **TDK Corporation** currently lacks the technology to enter the GPU/ASIC "low-voltage high-capacity" MLCC market (hoping to achieve breakthroughs in materials through joint R&D with Japanese chemical industries). However, the orders for its "high-voltage high-capacity" MLCCs used around power circuits are very strong, as they can utilize product technologies almost identical to those used in automotive applications like electric vehicles, which is expected to help improve factory utilization rates. **How severe is the "bottleneck" in this field?** **Goldman Sachs analyst Allen Chang recently pointed out that the real bottleneck in this field lies in the production equipment and core raw materials of the MLCC industry, which largely depend on in-house R&D. Limited by internal engineering expert resources, the capacity expansion elasticity across the industry is extremely rigid, with annual growth rates locked at a very low level of around 10%.** Once the newly added capacity is consumed by the surging AI demand, this cycle may evolve into a protracted structural supply-demand shortage. Although facing a sudden increase in demand for AI servers, major MLCC manufacturers have successively initiated price adjustment mechanisms. However, in fact, as Allen Chang illustrated in his sequence diagram regarding AI supply-demand bottlenecks and the prices of components and materials, unlike memory (DRAM and NAND), ABF substrates, and CCL (copper-clad laminates), MLCCs have seen almost no significant price increases to date And **this also means that among all upstream hardware and materials for AI, MLCC may actually have the largest price increase potential.** Just as memory prices surged when the market's focus was on specific bottlenecks in the deployment of the AI supercycle by the end of 2025; today, driven by the huge demand opportunities brought by AI servers—Goldman Sachs expects its demand to grow approximately 4.3 times from the fiscal year 2025 to the fiscal year 2030—the entire MLCC industry is increasingly feeling the tension between supply and demand. It is noteworthy that traditional consumer electronics customers such as smartphones and PCs, under the pressure of traditional terminal shipments and supply shortages, have recently begun to unusually seek to sign MLCC "long-term contracts." The deeper logic behind this is that as the core production capacity of the entire MLCC industry is prioritized and shifted to the more profitable AI server supply chain, those traditional long-tail customers lacking high-priority premium capabilities have developed a strong anxiety about future component procurement due to fears of supply shortages. **According to industry chain research, the marginal changes driving this round of MLCC price reassessment also include:** > High prices of key upstream metals—nickel and silver—are affecting all sub-markets; > There is a supply-demand imbalance in the high-voltage, high-capacity MLCC sub-market (used in automotive and servers); > The delivery cycle for high-voltage, high-capacity MLCC has exceeded 20 weeks; > Driven by channel stockpiling and downstream companies placing repeat orders (especially in the Chinese market), the spot/distributor prices for low-capacity and consumer device MLCC have risen by 20%-40%; > OEM contract prices have not yet seen significant increases (this also means that there is still great momentum for price transmission to the contract market in the future). **What investment opportunities does Goldman Sachs see?** **Goldman Sachs believes that while investors should position themselves in the three leading companies in this field—Murata Manufacturing, Taiyo Yuden, and Samsung Electro-Mechanics, companies listed on the Taiwan Stock Exchange such as Yageo and A-share companies like Fenghua Advanced Technology and Sanhua Group are gradually becoming industry leaders.** In addition, investors can also buy Goldman Sachs' Asian MLCC stock portfolio GSXAMLCS. This portfolio covers all major participants in the Japanese, Korean, Taiwanese, and A-share markets, providing differentiated investment opportunities in the MLCC value chain, including pure MLCC manufacturers, diversified passive component manufacturers, and upstream material suppliers. **As shown in the figure below, this portfolio has recently begun to significantly widen the gap in catch-up (orange line), but still lags behind other popular AI themes (blue line and purple line).** ### Related Stocks - [MRAAY.US](https://longbridge.com/en/quote/MRAAY.US.md) - [6981.JP](https://longbridge.com/en/quote/6981.JP.md) - [TTDKY.US](https://longbridge.com/en/quote/TTDKY.US.md) - [6762.JP](https://longbridge.com/en/quote/6762.JP.md) - [6976.JP](https://longbridge.com/en/quote/6976.JP.md) - [GS.US](https://longbridge.com/en/quote/GS.US.md) - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [MS.US](https://longbridge.com/en/quote/MS.US.md) - [4092.JP](https://longbridge.com/en/quote/4092.JP.md) - [000636.CN](https://longbridge.com/en/quote/000636.CN.md) - [300408.CN](https://longbridge.com/en/quote/300408.CN.md) - [W4VR.SG](https://longbridge.com/en/quote/W4VR.SG.md) - [NVD.DE](https://longbridge.com/en/quote/NVD.DE.md) - [MS-O.US](https://longbridge.com/en/quote/MS-O.US.md) - [MS-Q.US](https://longbridge.com/en/quote/MS-Q.US.md) - [MS-E.US](https://longbridge.com/en/quote/MS-E.US.md) - [MS-I.US](https://longbridge.com/en/quote/MS-I.US.md) - [MS-L.US](https://longbridge.com/en/quote/MS-L.US.md) - [MS-P.US](https://longbridge.com/en/quote/MS-P.US.md) - [MS-A.US](https://longbridge.com/en/quote/MS-A.US.md) - [MS-F.US](https://longbridge.com/en/quote/MS-F.US.md) - [MS-K.US](https://longbridge.com/en/quote/MS-K.US.md) - [PJX.SG](https://longbridge.com/en/quote/PJX.SG.md) ## Related News & Research - [Billionaire investor Dan Loeb dismisses AI bubble talk: 'We're barely scratching the surface'](https://longbridge.com/en/news/288060733.md) - [CertiK Launches AI Skill Scanner, An Antivirus Software for the AI Age](https://longbridge.com/en/news/287832378.md) - [Omdia: AI Factory Market Enters Industrialization Era as Five Dynamics Redefine AI Infrastructure in 2026 | TTGT Stock News](https://longbridge.com/en/news/287879230.md) - [Redwood AI Announces Letter of Intent for Potential Acquisition of Quantum.IQ | RDWCF Stock News](https://longbridge.com/en/news/287977147.md) - [Nvidia shares dropped after stellar earnings. Is this a sign of what's coming for artificial intelligence (AI) stocks?](https://longbridge.com/en/news/288177566.md)