In the infrastructure arms race of the AI supercycle, the shifting bottlenecks of supply and demand have created successive waves of market winners. Data centers, energy infrastructure, and memory chips have all taken turns as focal points for capital. Goldman Sachs' latest analysis points to the next imminent bottleneck: the long-underappreciated Multi-Layer Ceramic Capacitor (MLCC), which it characterizes as the "new memory"—a passive component sub-sector standing at the starting point of a cycle marked by rising volume and prices.
Price signals are rapidly intensifying. According to media reports in April, Murata increased prices for MLCC products used in AI servers and high-end automotive applications by 15% to 35% effective April 1st. Taiyo Yuden also notified customers of price increases across several product lines starting in May, citing sustained cost increases for various raw materials, including precious metals. Spot prices for AI-related MLCCs have risen approximately 20%, with lead times for high-end products exceeding 20 weeks. Japan's Ministry of Finance trade statistics released on May 28th further corroborate this trend: the average export price of MLCCs in April rose 16% year-over-year, with export value increasing 28% year-over-year.
Goldman Sachs analyst Daiki Takayama forecasts that the AI server MLCC market will grow more than fourfold from FY2025 to FY2030, expanding from approximately ¥215 billion to about ¥920 billion, representing a compound annual growth rate of 34%. Concurrently, the overall MLCC industry's annual capacity growth rate is only slightly above 10%, with expansion constrained by the internal production capabilities for equipment and materials. Goldman Sachs states that the current AI-driven MLCC cycle "will be the largest and longest in history, and we believe it is still in its early stages," maintaining Buy ratings on Murata, Taiyo Yuden, and TDK.
For investors, the profit elasticity stemming from the MLCC supply-demand imbalance is significant. A mere 5% increase in average selling prices could theoretically boost Murata's FY2027 operating profit by approximately 13% and Taiyo Yuden's by as much as 37%. Goldman Sachs' constructed basket of Asian MLCC-themed stocks has recently begun to strengthen but still shows significant room for catch-up relative to other popular AI themes.
**The Capacitor's Role: The "Instantaneous Voltage Stabilizer" Beside AI Chips**
An MLCC can be understood as an extremely miniature, ultra-fast charge/discharge cell. While a regular battery stores a large amount of energy and releases it slowly, an MLCC stores a small amount of energy but can charge and discharge in fractions of a millisecond. Its core functions are twofold: first, power smoothing—absorbing instantaneous voltage spikes or filling sudden drops to provide stable current to sensitive chips; second, noise filtering—blocking electrical interference that could corrupt digital data.
The operational characteristics of AI servers make MLCCs indispensable. When AI models process large-scale computations, processors can spike power demand at the microsecond level and then drop it to zero immediately after computation. The power supply system simply cannot respond quickly enough to such changes. MLCCs are installed directly adjacent to AI chips, releasing energy instantaneously during demand peaks to prevent server crashes. As AI chips (like NVIDIA GPUs) need to handle billions of tasks simultaneously, a single top-tier AI server rack can require up to 600,000 MLCCs working in concert to maintain system stability.
In the view of Goldman Sachs analyst Nelson Armbrust, MLCCs have already become the third most expensive component in the Bill of Materials (BOM) for AI servers, trailing only GPUs and memory. The entire MLCC market is valued at approximately $15 billion, with the server segment at about $1.3 billion, expanding at an 80% compound annual growth rate. Meanwhile, demand growth in other applications like automotive and mobile phones has slowed. Daiki Takayama expects the MLCC share of the AI server BOM to rise from the current ~0.5% to approximately 1%.
**Supply-Demand Imbalance: 10% Capacity Growth Faces a 4x Demand Shock**
Goldman Sachs analyst Allen Chang clearly points out the core structural conflict in the MLCC industry: annual industry capacity growth is only slightly above 10%. Furthermore, as both equipment and materials rely on in-house production, the pace of capacity expansion is limited by internal engineering resources, making significant acceleration difficult. The demand shock from AI servers, however, operates on a completely different scale.
Goldman Sachs estimates AI server demand for MLCCs will grow about 4.3 times from FY2025 to FY2030. Simultaneously, demand for high-voltage, high-capacity MLCCs from automotive electrification remains robust—with usage per electric vehicle continuing to increase. These two major demand pillars are absorbing the already limited new capacity, prompting customers in the consumer electronics sector, despite facing their own demand slowdown, to actively seek long-term supply contracts to hedge against future shortage risks.
Signals of current market tightness are evident across multiple levels: lead times for high-end MLCCs (high-capacity, high-voltage specifications) exceed 20 weeks; spot and distributor channel prices for low-capacity and consumer-grade MLCCs have risen 20% to 40% due to hoarding and double-ordering; prices for key raw materials like nickel and silver remain high, creating cost pressure across all segments; and while OEM contract prices have not yet increased significantly, market expectations are shifting.
Within the overall AI supply chain timeline, Goldman Sachs' analytical framework shows that the MLCC pricing cycle has started notably later than for other core AI components like DRAM, NAND memory, ABF substrates, and Copper Clad Laminate (CCL). Precisely for this reason, Goldman Sachs judges that MLCCs (alongside ABF and CCL) have the longest runway for price appreciation among all AI components and materials.
**Price Cycle Accelerates: Dual Validation from Data and Profit Elasticity**
The price increases by Murata and Taiyo Yuden mark the formal start of the MLCC price cycle. Murata raised prices by 15% to 35% in AI server and high-end automotive applications effective April 1st. Taiyo Yuden notified customers of price adjustments across several product lines starting in May, including MLCCs, inductors, RF devices, FBAR/SAW devices, and aluminum electrolytic capacitors.
The trade statistics from Japan's Ministry of Finance on May 28th provide macro-level validation: the average MLCC export price in April rose 3% month-over-month and 16% year-over-year; export volume grew 10% year-over-year; and export value increased 28% year-over-year. Goldman Sachs views this data as consistent with the recent earnings reports from Japanese MLCC manufacturers, where companies have confirmed the continuation of strong order trends.
Regarding profit elasticity, Daiki Takayama estimates that a 5% product price increase could theoretically boost Murata's FY2027 operating profit by approximately 13% and Taiyo Yuden's by about 37%. Goldman Sachs forecasts FY2027 sales of ¥10.55 trillion (up 13% year-over-year) for Murata and ¥286 billion (up 13% year-over-year) for Taiyo Yuden. The firm has raised its 2026 year-over-year MLCC price change forecast from approximately 0% to a range of 0% to +5%, noting that actual increases could ultimately be significantly higher.
Another important variable comes from NVIDIA's next-generation Vera Rubin server rack. According to Morgan Stanley estimates, the MLCC content value in the VR200 rack is approximately $4,300, a significant jump from the roughly $1,500 in the previous-generation GB300. Morgan Stanley's channel checks also indicate that MLCC usage is increasing significantly on both compute boards and switch boards, with a more pronounced increase on compute boards. Newly introduced BlueField and ConnectX modules will further elevate the total MLCC usage per rack. This partly explains the current robust demand for high-end AI server MLCCs and is driving ODM manufacturers to actively build inventory in preparation for the volume production and delivery of the Rubin rack starting in the second half of 2026.