Author: Block Analytics Ltd X Merkle 3s Capital

Introduction: After GPUs, who is quietly raising prices?

Recently, there was an alarming report from Huaqiangbei: MLCC prices are set to rise across the board, with increases ranging from 10% to 70%, effective July 1. This is not a standalone action by a single manufacturer, but a collective price adjustment across the entire industry chain. Murata's magnetic beads, chip capacitors, and chip inductors, are seeing increases concentrated between 50% to 70%; Yageo's high-capacity MLCC models are even more dramatic, with increases ranging from 5% all the way up to 275%. Trade insiders are blunt: It’s not just about wanting to buy, those who have stock are in charge.

The phrase "supply and demand" has not been seen in this industry for a long time. For the past decade, MLCCs have been viewed as "standard parts at a low price," often costing mere cents per unit, with prices dropping without a bottom and rising without anyone taking notice. Every few years, the industry cycles through "price increases—capacity expansion—overcapacity—price collapse," leaving old players haunted by this cycle; the first reaction to price hikes is often vigilance rather than excitement. But this time is different. When a low-key market valued at $15 billion starts speaking in terms of "spot market reigns supreme," there must be a larger force at work behind it.

Moreover, the structure of this price increase is very special. The sharpest increases are not seen in ubiquitous standard materials but rather in high-capacity, small-sized, automotive-grade, and server-use high-end models—which means the further up the pyramid one goes, the harder it is to obtain and the more expensive it becomes. This is completely different from past scenarios of universal industry-wide price hikes followed by collective drops. This indicates that the driver behind this round of price increases is not simply inventory speculation, but is driven by structural, genuine demand from high-end applications.

That force is AI.

The latest research report provides an unexpected judgment: within the cost structure of AI servers, MLCC has quietly climbed to become the third largest cost item, following GPUs and storage. A capacitor that costs a few cents can surprisingly rank alongside GPUs that cost tens of thousands of dollars on the same cost sheet, indicating that the rules of the game are being rewritten. It is important to note that on this cost sheet, both GPUs and storage, which are above MLCC in rank, are recognized hard currencies and have been inflated by the capital market in the past two years. MLCC has secured a top-three position not primarily through high prices per unit, but rather through horrific volume stacking—tens of thousands of small components combined together results in totals that surpass many other components with higher unit prices.

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

This article aims to clarify this kind of story: a humble, overlooked electronic component sector is being thoroughly reshaped by AI. Demand is expanding at an exponential rate while supply is lagging like an old bull pulling a cart; the resulting gap is shaping into a super cycle that could last until 2030. The three companies at the top of this sector are being repriced.

Let’s break them down one by one.

Demand Side: From 48,000 to 600,000

To understand how drastic this change is, let’s look at some usage numbers.

A traditional general-purpose server typically uses about 2,000 MLCCs. This volume is quite ordinary, comparable to a high-end smartphone. But once we enter the AI era, the numbers start to spiral out of control. An 8-card training server directly jumps to using 25,000 to 28,000 MLCCs, which is several times that of traditional servers.

Even more staggering is what comes next. NVIDIA's GB300 NVL72 rack uses as many as 440,000 MLCCs per unit. For the next generation, the Vera Rubin platform's VR200 is expected to reach 600,000 MLCCs per unit. The highest configuration, the Vera Rubin Ultra NVL576, will use between 3 million to 3.5 million MLCCs. From 2,000 to 3.5 million, that's a leap of over a thousand times.

Why has it surged to such an extent? The principle is quite simple, primarily revolving around "electricity."

The power density of new-generation GPUs is increasing, yet voltage levels are decreasing. Take Rubin as an example; it operates on a power rail of less than 1 volt with a power draw of up to 1,800 watts. Power equals voltage times current, so with voltage below 1 volt, it means current must exceed 1,800 amperes. What does this mean? It’s equivalent to pumping the electricity of a small factory into a chip the size of a palm. With such large currents, even slight fluctuations can cause issues.

MLCC handles the job of acting as a "voltage reservoir" for this surge of electricity. When curren

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。