How Japan and Korea's AI Pivot Is Handing China the Electronics Supply Chain

The Rice Shortage Nobody Was Watching

There is a tiny ceramic component inside nearly every electronic device on earth. It stores electricity and releases it cleanly when circuits demand it. The industry calls it MLCC—multilayer ceramic capacitor—and the nickname for it across Asia is the rice of electronics. You do not think about rice until you run out.

What is happening right now is not a crisis yet. But the supply chain is shifting underneath it in ways that will echo through AI servers, automobiles, consumer appliances, and every factory that builds them.

Murata Manufacturing, the Japanese firm that has long sat atop the global MLCC market, recently informed customers that it is beginning a phase-out of several general-purpose capacitor lines, including some used in IT equipment and automotive applications. The company will not stop selling those parts entirely—customers say the timeline stretches toward 2029—but the direction is unmistakable. Murata is moving its capacity toward capacitors built for AI infrastructure, where margins are richer and demand is surging.

Samsung Electro-Mechanics, South Korea’s largest capacitor maker, is doing the same thing. Its AI server MLCC division is operating near full tilt, with utilization rates in the high 90s percent range. Company filings show average selling prices for MLCCs rose nearly 14 percent in the first half of the year compared to the same period last year. The firm is not raising prices on a whim. It is choosing customers and orders in a market where demand vastly outstrips the handful of suppliers who can actually make the product.

Together, Samsung and Murata are estimated to hold close to 90 percent of the AI server MLCC market. That concentration is unusual for a component market, and it gives these two firms pricing power that few other suppliers in the electronics chain currently enjoy.

Why AI Changes Everything About a Tiny Capacitor

MLCCs for AI work differently than MLCCs for your television or a basic power supply. An AI server runs graphics processing units and high-bandwidth memory chips that draw enormous amounts of power in short, sharp bursts. The capacitors sitting next to those chips must handle high voltage, high temperature, and rapid cycling without degrading. They also need to store far more charge in a small footprint.

That last point is what matters most for the supply crunch. To store more charge, a manufacturer must stack many more thin layers of ceramic dielectric material inside a single capacitor. Industry estimates put the layer count for AI-grade MLCCs at roughly ten times that of general-purpose parts. More layers mean longer production time, more careful handling, and fewer good parts coming off the line per wafer.

When Samsung and Murata shift their best production capacity toward AI-grade components, they are not simply swapping one product for another on the same assembly line. They are reallocating a scarce resource: factory floor space, skilled labor, and the specialized kilns and deposition equipment that take months to qualify for new products. Every extra AI capacitor produced is a general-purpose capacitor that is not.

The Overflow Is Going to China and Taiwan

This is where the second-order effects appear.

Customers who relied on Murata or Samsung for standard X5R-class capacitors—the workhorse part used in everything from industrial controllers to EV charging systems—now need a supplier. Those orders are moving. Taiwan and China are absorbing them.

Yageo, the Taiwanese capacitor maker, reported that its August revenue jumped 51.8 percent year over year to 16.33 billion New Taiwan dollars, a monthly record for the company. The firm said the growth came across its entire product range, not just from one category. Its Q3 utilization is expected to clear 90 percent.

Valuelink Technology, another Taiwanese firm, said its order book has swelled to the point where it cannot ship fast enough. The company’s leadership described the situation in interviews as a direct consequence of major suppliers reducing general-purpose MLCC output and redirecting it elsewhere.

Chinese maker Viiyong is positioned similarly. It does not compete with Samsung or Murata on AI server components. It competes on the parts that go into appliances, consumer electronics, and automotive systems—the very volumes that are now being displaced.

The switch is feasible because the technical barrier to changing MLCC suppliers for general-purpose applications is relatively low. DB Securities noted that IT-grade MLCC specifications tend to be standardized enough that a buyer can qualify a new supplier without redesigning its circuit board. That makes the transition faster than it would be for a custom component.

Merrill Lynch Securities is projecting that the shortage will broaden beyond AI-specific parts into the general MLCC market as a whole, which could push prices higher across the board.

Who Wins, Who Loses

Samsung and Murata win on margin. They are selling a smaller volume of parts at significantly higher prices, and the AI infrastructure buildout shows no sign of slowing. Their utilization numbers prove there is demand to back the strategy.

Yageo, Valuelink, Viiyong, and similar firms win on volume. They are picking up orders that larger competitors no longer want, and they are doing so with factories that already exist. The capital expenditure required is modest compared to building AI-grade lines from scratch.

Automakers and appliance manufacturers lose in the near term. They face longer lead times and potentially higher prices for components that have been cheap and reliable for decades. The auto industry, already strained by semiconductor shortages, now has a parallel problem in the passive-component chain.

The Chinese government and its domestic supply-chain players win in the long term. Every order that flows to a Chinese MLCC maker is experience, yield data, and customer relationships that cannot be easily unwound. The company may start by making simple capacitors, but it will learn how to make better ones. That is how supply chains mature.

What Happens Next

The AI capex cycle is still early. Data-center construction, GPU shipments, and HBM demand are all tracking upward. Samsung and Murata will keep prioritizing AI-grade MLCCs as long as the price premium holds. That premium has room to run further before it hits a ceiling, because there are genuinely few other companies capable of making these parts at scale.

But the overflow effect is self-reinforcing. As Taiwanese and Chinese makers fill the general-purpose gap, they generate revenue that can eventually be funneled into higher-end product development. Yageo and Valuelink are already cash-flow positive enough to invest in capacity expansion. Viiyong and other Chinese firms have state-backed financing and domestic demand as a safety net.

Murata’s formal phase-out timeline through 2029 suggests the shift is deliberate, not accidental. The company is giving its customers a multi-year window to find alternative suppliers before cutting them loose. That is a long time to be uncertain about your component supply.

For the electronics industry, the lesson is simple. The AI boom is not just consuming advanced semiconductors. It is quietly restructuring the entire downstream supply chain, one tiny ceramic capacitor at a time.