Authors: Block Analytics Ltd X Merkle 3s Capital

Introduction: After GPUs, who is quietly raising prices?

A recent report from Huaqiangbei (a major electronics market in Shenzhen) has caused quite a stir: MLCC (Multi-Layer Ceramic Capacitors) prices are set to rise across the board, with increases ranging from 10% to 70%, officially taking effect on July 1st . This isn't an isolated move by a single manufacturer, but a collective price adjustment across the entire industry chain. Murata's ferrite beads, surface-mount capacitors, and surface-mount inductors are seeing price increases concentrated between 50% and 70%; Yageo's high-capacitance MLCCs are experiencing even more dramatic increases, with prices jumping from 5% to 275%. Leading traders have put it bluntly: now it's not as simple as just wanting to buy; whoever has stock on hand is the boss .

The phrase "supply falling short of demand" hasn't been heard in this industry for a long time. For the past decade, MLCCs have been perceived as "standard components at rock-bottom prices," with prices often measured in cents per unit. Prices plummeted without warning, and surged without much concern. Every few years, the industry experiences a cycle of "price increases—capacity expansion—overcapacity—price collapse," leaving veteran players wary and their first reaction to price increases is often caution rather than excitement. But this time is different. When a low-profile sector with an annual output value of $15 billion begins to speak with "spot market dominance," there must be a much larger force driving it from behind .

Moreover, the structure of this price increase is quite unique. The most dramatic price hikes weren't in the readily available standard components, but rather in high-capacity, small-size, automotive-grade, and server-grade high-end models—meaning the higher up the pyramid, the less available and more expensive they become. This is completely different from the past scenario of industry-wide price increases followed by a general decline. It indicates that this round of price increases is not simply a game of inventory, but rather a structural drive from genuine demand in the highest-end applications .

That force is AI .

A recent research report has yielded a surprising finding: in the cost structure of AI servers, MLCCs have quietly climbed to become the third largest cost item, second only to GPUs and memory. The fact that a tiny capacitor costing just a few cents can be listed on the same cost list as GPUs costing tens of thousands of dollars indicates that the rules of the game are being rewritten. It's worth noting that on this cost list, GPUs and memory, which rank ahead of MLCCs, are universally recognized as high-value commodities and have been repeatedly hyped by the capital market over the past two years. MLCCs' rise to the top three isn't due to their high price per unit, but rather the sheer volume of components used – the total cost of hundreds of thousands of these small parts far surpasses that of countless other, more expensive components.

When the name of a component starts appearing on the cost sheet of computing power, it is no longer just a component, but a strategic resource.

This article aims to tell the story of how AI is completely reshaping a seemingly insignificant and overlooked sector: electronic components. Demand is expanding exponentially, while supply is struggling to keep up, creating a gap that could become a supercycle lasting until 2030. The three companies at the top of this sector are being revalued.

Let's take them apart one by one.

Demand side: from 48,000 units to 600,000 units

To understand how dramatic this change is, let’s look at some usage figures.

A traditional general-purpose server typically uses around 2,000 MLCCs. This is a fairly common number, similar to a high-end mobile phone. However, once we enter the AI ​​era, the numbers start to spiral out of control. A training server with 8 GPUs can use 25,000 to 28,000 MLCCs, more than ten times that of a traditional server .

And the numbers don't stop there. NVIDIA's GB300 NVL72 rackmount uses 440,000 chips per unit. Moving to the next generation, the Vera Rubin platform's VR200 is expected to use 600,000 chips per unit. The top-of-the-line Vera Rubin Ultra NVL576 will use between 3 million and 3.5 million chips. From 2,000 to 3.5 million chips, that's a leap of over a thousand times.

Why did it surge to this extent? The reason is actually not complicated; the key lies in "electricity" .

The power density of next-generation GPUs is increasing, while the voltage is decreasing. Take Rubin, for example; it operates on a power rail of less than 1 volt, yet consumes a staggering 1,800 watts. Power equals voltage multiplied by current; a voltage below 1 volt means a current exceeding 1,800 amps. What does this mean? It's like cramming the electricity of a small factory into a palm-sized chip . With such a massive current, even the slightest fluctuation can cause the chip to malfunction.

The job of an MLCC is to act as a "voltage stabilizer" for this surging current. When the current fluctuates, it is responsible for instantly replenishing or absorbing charge to stabilize the voltage; this process is called decoupling. The larger the current, the lower the voltage, and the faster the fluctuations, the more and denser the "voltage stabilizer" is needed. Therefore, the more powerful the GPU, the greater the demand for MLCCs, and the increase is non-linear.

Beyond the explosive growth in quantity, a structural replacement is also underway. Aluminum polymer capacitors (APCs) were previously widely used in servers, but they are now being replaced by multi-layer capacitors (MLCCs). This replacement is leading to a 1.5 to 2-fold increase in usage. Because MLCCs are smaller, more stable, and have a longer lifespan, their advantages are overwhelming on high-density computing boards where space is extremely limited. Space on computing boards is fixed, but the current that needs to be stabilized is constantly increasing. Engineers can only make individual components smaller and use them more densely, making small and stable devices like MLCCs the natural choice. This replacement is not a one-off event but will continue with each new platform iteration, adding another layer of structural growth to the already explosive increase in quantity.

There's another easily overlooked point here: MLCCs aren't better off being as far away from the GPU as possible; quite the opposite, they should be placed as close to the GPU as possible. This is because current fluctuations are measured in nanoseconds, and the closer the current source, the more timely the replenishment. Therefore, high-end solutions will have a large number of MLCCs densely packed around and directly below the GPU. This layout itself dictates that the number used will only be greater, not less.

As the quantity increases, the value per unit also rises. In a GB300 rack, each MLCC is worth approximately $1,530. With Vera Rubin, this figure jumps to $4,320, an increase of 182%. That means that the value of each rack increases by nearly $3,000 just from MLCCs. The fiercer the computing power arms race, the bigger this pie becomes.

The end of computing power is electricity, and the one that controls electricity is this cheapest component.

Beyond AI, there's a second driving force: new energy vehicles. A single pure electric vehicle uses 18,000 MLCCs, six times that of a gasoline-powered vehicle. If we add Level 3 or higher advanced autonomous driving capabilities, the number increases to 15,000 to 20,000 MLCCs. Electrification combined with intelligent technology has created a massive new market for MLCCs, and the unit price and profit margin of automotive-grade products are significantly higher than those of consumer-grade products.

The significance of automotive-grade products lies not only in their large quantity but also in their quality. MLCCs used in vehicles must withstand repeated exposure to high temperatures, vibrations, and humidity, requiring reliability standards several orders of magnitude higher than consumer-grade products, and with much longer certification cycles. This means there are inherently fewer manufacturers capable of producing automotive-grade products, resulting in a cleaner competitive landscape and more stable prices. For leading manufacturers, AI servers and new energy vehicles represent two sectors that are both highly reliable, high-value, and have high barriers to entry. Furthermore, the peak demand periods for these two sectors overlap, allowing them to fully utilize their production capacity .

Looking at these together, the trend becomes clear. The market size of MLCCs used in AI servers was approximately $1.4 billion in fiscal year 2025, and is projected to reach $6.1 billion by fiscal year 2030, representing a compound annual growth rate of 34% over five years. It's important to note that currently, MLCCs used in AI servers only account for about 5% of the global MLCC market. This 5% segment is the fastest-growing of all segments, meaning its marginal impact on the entire industry far exceeds its current size.

The demand story is now complete, showing a steep upward curve. But the key issue is never just demand. What truly determines how far and how rapidly this cycle can go is whether the supply side can keep up.

The answer is: It's very difficult.

Supply side: Why is it so difficult to expand production?

Let me first explain in plain language how MLCCs are made, and you'll understand where the barriers to entry are in this business.

The first step is powder preparation. The core dielectric of MLCCs is barium titanate, but not just any barium titanate; it must be ultrafine powder with a particle size controlled between 50 and 300 nanometers. How small is this particle size? A few hundred such particles can fit into the diameter of a human hair. The quality of the powder preparation directly determines the upper limit of the final product's performance.

The second step is casting, which involves mixing the powder into a slurry and then spreading it into an ultra-thin film, much like making a pancake. High-end products have a single layer thickness of only 0.4 to 0.5 micrometers, which is dozens of times thinner than plastic wrap, and it is required to be of uniform thickness and without any defects .

The third step is to print internal electrodes on the membrane. The fourth step is to stack the membrane with printed electrodes layer by layer; high-end products can have more than 1,000 layers. After stacking, the membrane is sintered in a high-temperature, reducing atmosphere at 1,200 to 1,300 degrees Celsius to fuse these thousands of layers into a dense whole. Finally, the membrane is sealed, electroplated, and tested.

The entire process sounds simple, but each step is incredibly difficult. Murata achieved the world's first mass production of a 0402-sized, 47 microfarad capacitor in 2025. What level of technology is that? It's like packing a capacitance that previously required much larger components into a volume the size of a sesame seed. Only a handful of manufacturers worldwide can achieve this level of advanced technology .

Why is it so difficult? Ultimately, it's because of six layers of barriers that, when stacked together, form an almost insurmountable moat.

The first barrier is the technological barrier . The material formulation of MLCCs is the result of nearly 80 years of accumulation by Japanese manufacturers. The subtle differences in the formulation are simply incomprehensible and impossible for outsiders to copy. Even more critical is the core equipment—high-precision casting machines, laminating machines, and special kilns. These are all manufactured by leading manufacturers themselves and cannot be bought on the market. Having money is useless because the key machines are not for sale.

The second barrier is customer barriers . The certification cycle for MLCCs used in AI servers takes 12 to 18 months; automotive-grade certification is even more stringent, taking 2 to 3 years. Once a manufacturer enters the supply chain of a major customer, the customer will not easily switch, because switching to another manufacturer requires recertification, and the time and risk costs are frighteningly high. This stickiness makes the leading manufacturers' position exceptionally stable.

The third barrier is the capital barrier . An investment of $300 million to $500 million is required for a high-end production line, and it takes four to five years from completion to full production. This means that the money you invest today will only see a full return five years later, during which time you also bear the risks of technological iteration and demand fluctuations. Without substantial capital and long-term patience, it's simply impossible to participate.

The fourth barrier is the patent barrier . Murata holds the most patents in the industry and even received the IEEE Milestone Award in 2024. It's extremely difficult for newcomers to bypass these patents and develop high-end products. The fifth barrier is the talent barrier. It takes 5 to 10 years to train a core engineer to work independently, and the lifetime employment system of Japanese companies keeps these valuable talents firmly within the system, making them impossible to poach. The sixth barrier is the scale barrier. Leading manufacturers produce trillions of units annually; the cost advantages and accumulated process data from this scale are unattainable for new entrants.

The real moat is never a single technology, but something built up over decades that cannot be bought or copied.

Because of these six barriers, MLCC capacity expansion is extremely slow, with the industry's overall capacity growth at only around 10% annually. Behind this are eight intertwined reasons: the delivery time for key equipment alone is 12 to 18 months; process debugging for new production lines takes 6 to 12 months; yield ramp-up is a slow process that cannot be rushed; there is a chronic shortage of high-end talent; upstream raw material bottlenecks exist; manufacturers remember the painful lessons of blindly expanding capacity in the past and are hesitant to make significant investments; technology iterates too quickly, and production lines invested in today may be outdated tomorrow; coupled with structural mismatch in capacity, what can be produced is not what the market demands. With these eight factors combined, capacity expansion cannot be accelerated even if desired.

The most intriguing reason here is the sixth one—the lessons of the past. In the last cycle, many manufacturers frantically expanded production at the peak, only to see demand decline and the concentrated release of new capacity cause prices to plummet, taking years to recover . This memory has made today's leading manufacturers exceptionally restrained in expanding production. They would rather earn less from capacity expansion than destroy the hard-won high-price cycle. This collective "restraint" is essentially a supply discipline, and it is precisely this discipline that makes this round of supply-demand gap more difficult to fill than ever before. In other words, the slow pace of expansion is half due to objective limitations and half due to subjective unwillingness.

So here's the question: China's electronics industry has made rapid progress in recent years, so why can't it still produce high-end MLCCs ?

The gap is real. In terms of dielectric layer thickness, high-end products need to achieve 0.4 micrometers, while mainland China's current level is 1 to 2 micrometers, a difference of nearly two generations. Regarding the number of layers, high-end products can stack over 1,000 layers, while the mainstream in mainland China is still at 300 to 500 layers. Even more of a bottleneck is the upstream high-end powder, which is heavily reliant on Japan's Sakai Chemical, which alone holds 28% of the global market share. This triple dominance in formulation, equipment, and materials makes it difficult for mainland manufacturers to break into the high-end market in the short term; they can mainly compete in the mid-to-low-end segment.

So the current situation is this: demand is surging at a rate of 34% annually, while supply is only crawling along at 10% per year. This disparity forms the most solid foundation of this supercycle. The supply-demand gap won't disappear immediately; instead, it will continue to widen. This leads to the most crucial question—who will get the biggest slice of this feast ?

The Big Three: Who is the biggest winner?

The global high-end MLCC market is essentially a game between three companies. Each has its own characteristics and its own strategy.

Murata – the absolute leader

Murata is the undisputed king of this industry. Its stock price is approximately 8,711 yen, with a market capitalization of 17.65 trillion yen, equivalent to approximately US$114.5 billion. It holds a 40% market share in the global MLCC market, and in the most valuable AI server MLCC segment, its share reaches 45% to 70%. In other words, at least one out of every two AI servers uses high-end capacitors from Murata .

Murata's profitability is equally impressive. With a gross profit margin of 42.1% and an operating profit margin of 15.4%, it ranks among the top tier in the manufacturing industry. In fiscal year 2026, its capacitor business revenue is projected to reach 936.4 billion yen, accounting for 51.1% of total revenue—a veritable half of its business. Murata is also willing to invest heavily in capacity expansion, with a capital expenditure plan of 250 billion yen for fiscal year 2027. However, even with this, the annual growth rate of MLCC capacity will only reach 10%—even industry leaders cannot achieve such rapid growth, which precisely reflects the rigidity of the supply side. Its new factory in Izumo has been a 10-story building, with an investment of 47 billion yen, demonstrating its long-term strategic commitment.

In terms of valuation, Murata's TTM P/E ratio is 68.7x, with a projected P/E ratio of 40 to 55x, expected to drop to 30 to 40x by fiscal year 2028. It has received positive ratings from multiple institutions. More notably, in May 2026, Murata announced a ¥150 billion share buyback program. A leading company's willingness to use real money to buy back its own shares is itself a strong endorsement of its future prospects.

Murata's role is clear: it is the most stable player in this field and the first choice for those who want certainty .

Samsung Electro-Mechanics (SEMCO) – The King of Growth Flexibility

If Murata represents stability, then Samsung Electro-Mechanics represents resilience. With a stock price of approximately 1,664,000 won and a market capitalization of 125.7 trillion won (about US$96 billion), it holds a 20% to 25% market share in the global MLCC market and a 39% to 40% share in AI server MLCCs, firmly establishing itself as a strong second-place player.

Its most attractive feature is its growth potential. In the first quarter of 2026, revenue reached 3.21 trillion won, a year-on-year increase of 17%; operating profit reached 280.6 billion won, a significant year-on-year increase of 40%. Profit growth far outpaced revenue growth, indicating a shift in product mix towards higher-end products and improved profitability. Even more aggressive is its expansion plan—capital expenditure in 2026 will more than double, jumping from 1.15 trillion won to over 2 trillion won. It has also secured a 1.5 trillion won order for silicon capacitors and AI, with delivery scheduled for 2027-2028, locking in future growth potential.

Structurally, MLCCs account for about 45% of Samsung Electro-Mechanics' revenue, but contribute more than half of its operating profit—making it an absolute cash cow. Furthermore, backed by the entire Samsung Group ecosystem, it has a natural advantage in customer resources and upstream/downstream collaboration.

What's most appealing is its valuation elasticity. The TTM P/E ratio is over 150, which seems alarming, but looking ahead, it will compress to 59 in fiscal year 2027 and further to 41 in fiscal year 2028—the fastest P/E compression among the three companies. The underlying logic is explosive earnings growth: earnings per share are projected to increase 4.6 times over three years, from 9,361 won to 43,348 won. When profits grow at this rate, what seems high today will appear cheap tomorrow.

The so-called flexibility is about whose sails are hoisted the highest when the winds of change blow through the industry.

Samsung Electro-Mechanics' role is: those who want to maximize upside potential will be eyeing it.

Taiyo Yuden – Highest Purity MLCC

The third company is Taiyo Yuden. With a stock price of approximately 15,000 yen and a market capitalization of 2 trillion yen (about $12.4 billion), it is the smallest of the three. It holds 8% to 10% of the global MLCC market share, smaller than the other two, but it has a unique characteristic—the highest purity. MLCCs account for 70.9% of its revenue, the highest in the entire industry. This means it is almost the purest target in the MLCC sector; every slight change in the industry is amplified and reflected in it.

Taiyo Yuden is at a clear turning point. Its operating profit margin rebounded from a low of 2.8% in fiscal year 2024 to 5.6% in fiscal year 2026, with a target of 7.8% in fiscal year 2027 and a goal of reaching 15% by 2030. This is a clear profit recovery curve. The driving force is clear: its AI server MLCC sales are projected to grow by 80% in fiscal year 2027. Its medium-term plan is also ambitious, aiming for a cumulative capital investment of ¥270 billion over five years by 2030.

In terms of valuation, Taiyo Yuden's TTM P/E ratio is between 134 and 147, with a forward P/E ratio of 46 to 81, falling back to 30 to 40 by fiscal year 2028. Because it has the smallest market capitalization and the highest purity, its Beta is also the highest among the three. Simply put, it rises the most sharply when the industry rises, and it falls the most sharply when the industry falls.

Its role is: people who want the purest MLCC exposure will choose it .

Valuation Comparison and Investment Framework

Comparing the three companies side-by-side makes the picture much clearer.

At first glance, the TTM P/E ratios of these three companies are not low: Murata at 68, Taiyo Yuden at over 134, and Samsung Electro-Mechanics at a staggering 161. Does this mean they are already too expensive and chasing the price is risky?

This judgment needs to be analyzed more carefully. A high price-to-earnings (P/E) ratio has completely different meanings at different stages of the cycle. If a company's profits have peaked, a high P/E ratio is a warning sign; however, if profits are on the verge of a surge, then today's high P/E ratio is precisely because the denominator (profit) hasn't yet risen. The forward P/E ratios of the three companies are all rapidly compressing downwards—Murata from 68 times to over 30 times, and Samsung Electro-Mechanics from 161 times to 41 times— this compression is not achieved through stock price declines, but through profit increases . This is a typical characteristic of the early stages of a cycle: the market has already priced in some AI expectations, but it is far from fully reflecting the upcoming price increase dividends.

The market has given this cycle a very heavy definition: the largest and longest MLCC supercycle in history, which will continue until 2030. And we are currently only in the early stages of the upward cycle, which is roughly equivalent to the second half of 2017 – the show has just begun.

Why is a price increase so crucial? Because MLCCs are a business highly dependent on capacity utilization, with fixed costs making up the majority of the revenue. Once prices rise, the extra money can almost directly become profit. It's estimated that for every 5% increase in average price, Taiyo Yuden's operating profit can increase by 37%. This is the power of operating leverage—small price changes can be amplified several times over by profit.

In an industry where supply is locked up, almost every penny of price increase will be converted into profit.

The potential for price increases in this round is considerable. The potential price increase for high-end MLCCs could reach 100% to 150%, and even for standard products, there's room for a 30% to 50% increase. Adding this price elasticity to the aforementioned supply-demand gap—with capacity increasing by 10% annually and demand by 34%—the gap will widen all the way to 2028—and you can understand why this is called a supercycle. The supply ceiling is firmly in place, while the demand floor is constantly rising; the space in between represents the potential for profit and stock price growth.

ETFs and Purchase Channels

At this point, many people will ask: How do I participate?

First, a slightly regrettable fact: there are no purely MLCC-themed ETFs on the market . This sector is too niche, and there are no dedicated index products covering it yet. However, it's still possible to indirectly invest in MLCCs through some sophisticated tools.

In the South Korean market, the SOL AI Semiconductor TOP2 Plus ETF is the most noteworthy, with Samsung Electro-Mechanics accounting for 27.3% and boasting a net asset value of approximately 5 trillion won. It's a good option for participating in Samsung Electro-Mechanics' growth potential. In the Japanese market, NEXT FUNDS' 1625.T fund is worth considering, as Murata, TDK, and Taiyo Yuden together account for about 8% to 12%, essentially a basket of Japanese giants. In the US stock market, EWJ's MLCC-related stocks total about 3.5%, and MKOR's Samsung Electro-Mechanics accounts for 4.85%, both relatively low concentrations, making them more suitable as part of an allocation rather than a primary focus.

If you want more direct exposure, you can consider ADRs. Murata has MRAAY, and Taiyo Yuden has TYOYY, both of which can be bought on the US stock market, saving you the trouble of trading Japanese stocks directly .

Risks and Conclusion

In any investment, understanding the risks is just as important as understanding the opportunities. There are five risk factors to keep in mind in this sector.

First, there's the reduction in AI capital expenditure , which is a high-risk factor. The entire demand-side narrative is built on the continued investment of cloud vendors and computing power players. Once the pace of industry investment slows down, the demand curve will flatten, directly impacting the logic of the supercycle.

Secondly, the valuation is too high , which also carries high risk. As mentioned earlier, the current price-to-earnings ratio has already reflected some expectations. If subsequent earnings performance falls short of expectations, there will be downward pressure on the valuation.

Third is the expansion of production capacity in mainland China , which is a medium-risk factor. The capacity expansion of mainland manufacturers in the low-to-mid-end market may cause price fluctuations, but they are unlikely to break into the high-end market in the short term, so the impact on the core business of the three giants is limited.

Fourth is the appreciation of the yen , which carries a medium risk. Murata and Taiyo Yuden are both Japanese companies. If the yen appreciates significantly, it will erode their overseas revenue and profits, putting pressure on their yen-denominated stock prices.

Fifth, the weakness in consumer electronics also poses a medium risk. Consumer electronics remains the traditional major driver of MLCC production, but this market is experiencing a K-shaped divergence, with high-end markets remaining stable while low-end markets are weak, and the overall drag cannot be ignored.

Presenting these risks is not intended to scare anyone away, but rather to make it clear that while the logic behind this supercycle is solid, it is not a one-sided market without variables. The sustainability of demand, the digestion of valuations, and exchange rate fluctuations all require continuous monitoring.

Returning to the initial question: After GPUs, what's quietly driving up prices? The answer is clear. It's MLCCs, those small capacitors that were once the least valued. They are undergoing a transformation—from a commodity whose price fluctuates and anyone can manufacture, to a strategic resource locked in by certification, constrained by production capacity, and repriced by AI.

When computing power becomes the oil of this era, MLCCs, which control every drop of current, are like the unnoticed yet indispensable pipelines that no one can live without.