The Next Flashpoint in the AI Arms Race! Goldman Sachs: Capacitors Are the "New Memory"
```
In the infrastructure arms race of the AI supercycle, successive supply and demand bottlenecks have given rise to waves of market winners—data centers, energy infrastructure, and memory chips have become the focus of capital. Goldman Sachs’ latest research points out that the next bottleneck about to erupt is the long-overlooked multilayer ceramic capacitor (MLCC), which they characterize as the “new memory”—a passive component market segment standing at the beginning of a cycle of rising quantity and price.
Pricing signals are intensifying quickly. According to media reports in April, Murata raised prices for MLCC products used in AI servers and high-end automotive applications by 15% to 35% starting April 1; Taiyo Yuden also notified customers of price increases for certain product lines beginning in May, citing rising costs of various raw materials including precious metals. The spot price of AI-related MLCCs has already risen about 20%, and lead times for high-end products exceed 20 weeks. Trade statistics released by Japan’s Ministry of Finance on May 28 further confirm the trend: the average export price of MLCC in April rose 16% year-on-year, and export value increased by 28% year-on-year.
Goldman Sachs analyst Daiki Takayama expects the AI server MLCC market to grow more than fourfold from FY2025 to FY2030, expanding from about 215 billion yen to about 920 billion yen, with a compound annual growth rate of 34%. Meanwhile, the entire MLCC industry’s production capacity is only growing just above 10% per year, with expansion constrained by the internal production capabilities for equipment and materials. Goldman Sachs says that the current AI-driven MLCC cycle “will be the largest and longest ever, and we believe it is still in the early stages,” maintaining buy ratings for Murata, Taiyo Yuden, and TDK.
For investors, the profit elasticity brought by the MLCC supply-demand mismatch cannot be underestimated: a mere 5% increase in average price could theoretically boost Murata’s operating profit in FY2027 by about 13%, and Taiyo Yuden’s by as much as 37%. Goldman Sachs’ Asian MLCC theme stock basket has recently started to rally, but there is still significant catch-up potential compared to other popular AI themes.
The Role of Capacitors: "Instant Voltage Stabilizers" Beside AI Chips
MLCCs can be understood as extremely miniature, ultra-fast charge/discharge units. Ordinary batteries store large amounts of electrical energy and release it slowly, while MLCCs store small amounts of energy but can complete charge/discharge in a tiny fraction of a millisecond. Their core functions are twofold: first, power smoothing—absorbing instantaneous voltage spikes or filling rapid drops, providing stable current to sensitive chips; second, noise filtering—cutting off electrical interference that could cause digital data corruption.
The operating characteristics of AI servers make MLCCs indispensable. When AI models process massive computations, processors instantly draw high power at the microsecond level, then drop back to zero immediately after; the power system cannot respond fast enough to such changes. MLCCs are installed directly next to AI chips, releasing energy instantly at demand peaks to prevent server crashes. Since AI chips (like Nvidia GPUs) must handle billions of tasks simultaneously, a top-tier AI server rack may require up to 600,000 MLCCs to work in tandem for system stability.
According to Goldman Sachs analyst Nelson Armbrust, MLCCs are now the third most expensive component in AI server BOM (bill of materials), after GPUs and memory. The total MLCC market is about $15 billion, of which the server segment is around $1.3 billion and is expanding at an 80% compound annual growth rate, while demand growth in automotive, phones, and other applications is slowing. Daiki Takayama estimates the MLCC fraction in AI server BOM will rise from the current ~0.5% to around 1%.
Supply-Demand Imbalance: 10% Capacity Growth Faces Quadruple Demand Shock
Goldman Sachs analyst Allen Chang clearly identifies the core structural contradiction in the MLCC industry: production capacity is growing at just over 10% per year, and since both equipment and materials rely on internal manufacturing, production expansion is limited by internal engineering resources, making rapid acceleration difficult. Meanwhile, the demand shock from AI servers is on a completely different scale.
Goldman Sachs estimates that AI server demand for MLCCs will grow about 4.3 times from FY2025 to FY2030. At the same time, electrification in automobiles is maintaining strong demand for high-voltage, high-capacity MLCCs—the MLCC content per EV continues to rise. These two main pillars absorb already limited new capacity, making even consumer electronics customers, despite their own slowing demand, eager to secure long-term supply contracts against future shortages.
Multiple market signals now indicate tight supply and demand: lead times for high-end MLCCs (high capacity, high voltage) exceed 20 weeks; prices for low-capacity and consumer MLCCs in spot and distributor channels have risen 20% to 40% due to hoarding and double-ordering; high prices for key raw materials like nickel and silver increase cost pressure across all segments; while OEM contract prices haven’t yet risen significantly, market expectations are shifting.
In the overall AI supply chain timeline, Goldman Sachs’ analysis shows MLCC price movement started significantly later than DRAM, NAND memory, ABF substrates, and copper-clad laminates (CCL). For this reason, Goldman Sachs judges MLCCs (alongside ABF and CCL) have the longest price upside among all AI components and materials.
Price Increase Cycle Accelerates: Dual Validation of Data and Profit Elasticity
Price hikes by Murata and Taiyo Yuden mark the official start of the MLCC price cycle. Murata raised prices for MLCCs in AI server and high-end auto applications by 15% to 35% starting April 1; Taiyo Yuden notified customers of price adjustments across multiple lines, including MLCCs, inductors, RF devices, FBAR/SAW devices, and aluminum electrolytic capacitors, starting in May.
Japanese Ministry of Finance trade statistics published on May 28 provide macro confirmation: in April, the average export price of MLCC rose 3% month-on-month and 16% year-on-year; export volume grew 10% year-on-year; export value rose 28% year-on-year. Goldman Sachs notes these figures echo the performance recently announced by Japan’s MLCC manufacturers, all confirming sustained strong order momentum.
For profit elasticity, Daiki Takayama estimates a 5% price increase could theoretically boost Murata’s FY2027 operating profit by about 13%, and Taiyo Yuden’s by about 37%. Goldman Sachs’ forecasts for Murata’s FY2027 sales are 1.055 trillion yen (+13% YoY), and for Taiyo Yuden, 286 billion yen (+13% YoY). Goldman Sachs has raised its 2026 MLCC price change forecast from ~0% to 0%–5%, noting the actual increase may well exceed this.
Another important variable comes from Nvidia’s next-generation Vera Rubin server racks. According to Morgan Stanley estimates, the MLCC value per VR200 rack is about $4,300, a sharp increase from ~$1,500 in the previous GB300 generation. Morgan Stanley channel research also shows both compute and switch boards are using significantly more MLCCs, with compute boards seeing a particularly large increase. Newly introduced BlueField and ConnectX modules will further raise MLCC total per rack. This explains, in part, why high-end AI server MLCC demand is so strong, prompting ODM vendors to aggressively build inventory in preparation for mass production and delivery of Rubin racks starting in the second half of 2026.
Risk Disclosure and DisclaimerMarket risk exists; investors should be cautious. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment goals, financial situation, or needs. Users should consider whether any opinions, views, or conclusions in this article apply to their particular situation. Investing accordingly is at your own risk. ```