Original title: Goldman, Morgan Stanley Sound Alarm: MLCCs Become AI Supply Chain's Scarcest Puzzle Piece, Nvidia Next-Gen Rack Usage Surges 182%

Original author: BitGo Finance

Original compilation: Peggy

Editor's note: Supply bottlenecks in AI infrastructure are continuing to spread from GPUs, memory, data centers and power systems to lower-level hardware components. Goldman Sachs and Morgan Stanley have recently set their sights on MLCCs, multilayer ceramic capacitors that have long been considered common passive components.

In AI servers, MLCC is responsible for stabilizing current and filtering noise, and is a key component to ensure high-speed operation of the chip. As Nvidia's next-generation rack architecture drives up the usage of single-rack MLCCs, their value is rising rapidly. Goldman Sachs predicts that the AI ​​server MLCC market will grow more than fourfold between 2025 and 2030, while the annual growth rate of industry production capacity is only slightly more than 10%. The mismatch between supply and demand is becoming the core variable of this round of market conditions.

More importantly, the price cycle has begun. Japanese leaders such as Murata and Taiyo Yuden took the lead in raising prices, and Japanese export data also began to verify the strength of demand. For the capital market, the logic of MLCC is not complicated: demand comes from AI servers and high-end cars, supply expansion is limited, and price increases can significantly amplify profit elasticity.

From chips to capacitors, the pricing power of the AI supply chain is being transmitted to more subdivided and hidden links. Whether MLCC will become the "next memory chip" still depends on whether the demand for AI servers can continue to be realized; but what is certain is that this once inconspicuous basic component has reached the starting point of a new round of rising volume and price cycles.

The following is the original text:

Supply bottlenecks in the artificial intelligence (AI) arms race are igniting opportunities in various hardware sectors. After data centers, energy infrastructure, and memory chips became the focus of capital attention, Wall Street giants Goldman Sachs and Morgan Stanley also pointed to a long-term undervalued basic component in their latest reports: multilayer ceramic capacitors (MLCC). The two institutions predict that MLCC will become the next key battlefield for "both volume and price increase", and this AI-driven growth cycle may be the largest in history.

Goldman Sachs analyst Daiki Takayama pointed out in a report that the AI server MLCC market size is expected to soar from approximately 215 billion yen (approximately US$1.4 billion) in fiscal 2025 to approximately 920 billion yen (approximately US$5.8 billion) in fiscal 2030, an increase of more than four times, with a compound annual growth rate of 34%. Goldman Sachs bluntly stated 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."

MLCC: The "invisible heart" that keeps the AI server running

MLCC (Multi-layer Ceramic Capacitor) can be understood as an extremely tiny, extremely fast-response charge and discharge unit. Unlike ordinary batteries that store large amounts of energy and release it slowly, MLCCs store very little energy but can charge and discharge in extremely short milliseconds or even less. Its core function is to smooth power supply fluctuations and filter noise: absorbing instantaneous voltage spikes, or quickly replenishing current when voltage sag, providing stable current to sensitive chips, and blocking electrical interference that may destroy digital signals.

The operational characteristics of the AI server make MLCC indispensable. When an AI model performs large-scale calculations, the power consumption requirements of the processor can soar in microseconds and then quickly drop to near zero after the calculation is completed. The power system itself is difficult to respond to such severe fluctuations in a timely manner. MLCC is usually installed directly near the AI ​​chip to instantly release energy when power consumption peaks occur to prevent server downtime. Since AI chips such as Nvidia GPUs need to handle billions of tasks simultaneously, a top-level AI server rack may require up to 600,000 MLCCs working together to maintain system stability.

Goldman Sachs analyst Nelson Armbrust further pointed out that MLCC has become the third most expensive component in the AI server bill of materials (BOM), second only to GPU and memory. The current overall MLCC market size is approximately US$15 billion, of which the server-related market is approximately US$1.3 billion and is expanding at a compound annual growth rate of 80%. In contrast, demand growth in other application areas such as automobiles and smartphones has slowed down significantly. Daiki Takayama predicts that the cost proportion of MLCC in the BOM of AI servers will gradually increase from the current approximately 0.5% to approximately 1%.

Structural contradiction between supply and demand: annual production capacity growth is only 10%, making it difficult to withstand a fourfold demand shock

The core factor that ignites market concern is that the MLCC industry is facing a serious structural imbalance between supply and demand. Goldman Sachs analyst Allen Chang made clear that annual capacity growth for the entire MLCC industry is just over 10%. In addition, since equipment and materials rely heavily on internal production of manufacturers, the production expansion schedule is limited by internal engineering resources, making it difficult to significantly accelerate. However, the demand impact from AI servers is not at the same level. Goldman Sachs expects MLCC demand from AI servers to grow approximately 4.3 times between fiscal 2025 and fiscal 2030.

What worries the market even more is that the demand for high-voltage and high-capacity MLCCs driven by vehicle electrification is still strong, and the usage of MLCCs in bicycles continues to increase. The two pillars of demand, AI servers and electric vehicles, are jointly consuming the already limited new production capacity. This has also led to even though demand for consumer electronics has declined, relevant customers are still actively seeking long-term supply agreements to prevent the risk of future shortages.

Signs of current market tension have appeared at multiple levels: the delivery cycle of high-end MLCCs (high-capacity, high-voltage specifications) has exceeded 20 weeks; low-capacity and consumer-grade MLCCs have been affected by hoarding and repeated orders, with spot and distribution channel prices rising by 20% to 40%; prices of key raw materials such as nickel and silver are still at high levels, putting pressure on the costs of various products.

The price increase cycle has officially begun: Japan's duo took the lead in raising prices, and official data confirms the trend

Price signals are strengthening rapidly. The price increases of Japan's two leading companies, Murata Manufacturing and Taiyo Yuden, mark the official start of the MLCC price increase cycle. Murata will increase the prices of MLCC products in AI servers and high-end automotive applications by 15% to 35% starting from April 1 this year. Taiyo Yuden has also notified customers that it will make price adjustments on multiple product lines starting in May, involving MLCCs, inductors, radio frequency devices, FBAR/SAW devices and aluminum electrolytic capacitors, etc., citing the continued rise in the cost of various raw materials such as precious metals.

Trade statistics released by the Ministry of Finance of Japan on May 28 verified this price increase trend from a macro level. Data show that in April, the average export price of MLCC increased by 3% month-on-month and 16% year-on-year; the export volume increased by 10% year-on-year; and the export value increased by 28% year-on-year. Goldman Sachs believes that this data confirms the signal released in the recent financial reports of Japanese MLCC manufacturers: all companies confirmed that order momentum remains strong.

Looking at the timeline of the entire AI supply chain, Goldman Sachs’ analysis framework shows that the price increase of MLCC clearly lags behind that of core AI components such as DRAM, NAND storage, ABF carrier boards, and copper clad laminates (CCL). Therefore, Goldman Sachs judges that among all AI components and materials, MLCC has the longest and most sustainable price increase space. Goldman Sachs has raised its 2026 year-over-year MLCC price change forecast to 0% to +5%, from around 0% previously, and stressed that actual future increases are likely to be much higher than this level.

Amazing profit elasticity: if the price increases by 5%, the operating profit will be the highest or increase by 37%

For investors, the profit elasticity brought about by the mismatch of MLCC supply and demand cannot be underestimated. Daiki Takayama estimates that a product price increase of only 5% can theoretically boost Murata's operating profit in fiscal year 2027 by about 13%, and Taiyo Yuden's operating profit by up to 37%.

Goldman Sachs predicts that Murata’s sales will reach 1.05 trillion yen (approximately US$6.6 billion) in fiscal 2027, a year-on-year increase of 13%; Taiyo Yuden’s sales will reach 286 billion yen (approximately US$1.8 billion), also a year-on-year increase of 13%. Goldman Sachs maintains "buy" ratings on Murata, Taiyo Yuden and TDK. The Asian MLCC theme stock portfolio it constructed has begun to strengthen recently, but compared with other popular AI themes, there is still significant room for compensatory growth.

Morgan Stanley dismantles Nvidia’s new rack: the importance of peripheral components has increased, and MLCC usage has surged by 182%

Another major catalyst comes from Nvidia’s next-generation Vera Rubin AI rack. In a teardown of Nvidia's latest VR200 chassis, Morgan Stanley found that peripheral components are rapidly rising in importance in the latest BOMs.

The value of MLCC in a single rack has risen from about $1,530 in the previous generation GB300 era to about $4,320, an increase of 182%. Although the absolute amount of MLCC is still lower than that of GPU, memory and PCB, its growth rate is extremely prominent among peripheral components.

Morgan Stanley's channel research further shows that the usage of MLCCs on both computing boards and switching boards has increased significantly, with the increase in computing boards being more obvious. In addition, the newly introduced BlueField and ConnectX modules will further increase the total usage of single-rack MLCC. This partly explains why the current demand for high-end AI server MLCCs is so strong, and has prompted many ODM manufacturers to actively stock up in preparation for the mass production and delivery of Rubin racks in the second half of 2026.

Morgan Stanley's teardown of the Nvidia Vera Rubin rack shows the following changes in the value of key components:

Market intelligence shows that in the infrastructure arms race of the AI super cycle, the rotation of supply bottlenecks has spawned waves of market winners. Goldman Sachs' latest assessment describes MLCC as a "new memory chip" - a passive component industry segment that is standing at the starting point of a cycle of rising volume and price.

With the exponential impact of demand for AI servers and Nvidia Rubin racks, high-end MLCC delivery cycles have exceeded 20 weeks, Japanese industry leaders have started to raise prices, and official export data continues to be strong. All signals point to the same conclusion: This AI-driven MLCC super cycle has just begun.

Original link