How tiny capacitors became the latest AI-driven investor darling


Demand for multilayer ceramic capacitors (MLCCs), tiny electronic components found in everything from smartphones to electric vehicles, is surging as the artificial intelligence boom triggers new production.

MLCCs, which act as electrical buffers in circuit boards, are increasingly deployed in massive volumes in high-performance servers, emerging as the latest AI-driven investor darling following memory chips and optical modules.

The AI hardware boom was squeezing production capacity as MLCC manufacturers prioritised high-end supply for data centres over consumer-grade capacitors, prompting suppliers to raise prices, TrendForce, a market research firm based in Taipei, said in a report released on Monday.

The sector was seeing growing demand because next-generation AI hardware required significantly larger volumes of MLCCs, TrendForce said. Nvidia’s next-generation Rubin architecture, which will power advanced AI computing platform but require more complex power management, used 12,000 MLCC units on a single board, compared with 6,500 on the current GB200 platform, the report said.

An AI server, densely packed with graphics processing units, consumes up to 10 times more power than a traditional server, and requires up to 28,000 MLCCs per unit, a 13-fold increase from a standard set-up, China Securities said in a research note issued last week.

The MLCC industry was poised for explosive growth, China Securities said – “potentially repeating the ‘growth miracle’ previously seen in the optical module sector”.

The global MLCC market is dominated by Japanese and South Korean manufacturers, led by Murata Manufacturing and Samsung Electro-Mechanics (SEMCO), which have a tight grip on the high-end capacitors used in AI servers.

The current tight supply of high-end MLCCs ... is unlikely to ease in the short term
TrendForce

Mainland Chinese players, while still small in scale, have been growing rapidly. They include Chaozhou Three-Circle and Guangdong Fenghua Advanced Technology, which have been working to expand into high-end capacitors for cars and, increasingly, data centres.

Guangdong Fenghua saw its share price rise 5 per cent on Wednesday, after a 10 per cent surge on Tuesday and a 9 per cent jump on Monday. Chaozhou Three-Circle’s stock price slipped less than 1 per cent on Wednesday after a 10 per cent gain on Tuesday. Both Chinese companies have seen their share prices more than double in the past year.

Global leader Murata has seen its stock price surge nearly 90 per cent this year. Last month, the Kyoto-based firm projected an up to 90 per cent jump in server-related sales for its capacitor business this year, with revenue predicted to rise 13 per cent.

SEMCO said demand for high-end MLCCs for AI servers and data centres remained strong and it had been discussing price increases with customers. Its share price has risen 265 per cent since the start of this year.

MLCC prices, which had been trending downwards for the past three years, were projected to rebound amid surging AI demand, the TrendForce report said.

“The current tight supply of high-end MLCCs ... is unlikely to ease in the short term,” it said.

Follow us on our official WhatsApp channel for breaking news alerts and key updates!

Next In Aseanplus News

Japan man with 'grudge against society' stabs bookshop worker: police
Strong earthquake shakes central Japan, shattering store windows and injuring at least 22
Delhi building collapse: Death toll climbs to seven
Se-young continues winning ways with 11th title
Continuous rain expected in four states until Nov 11
House arrest: DAP ministers staying on to complete reforms after Loke resigns, says Chan
Crew member goes missing as Indonesian water-bombing helicopter crashes into river
North Korea rolls out more than 60 varieties of cupbap-style ready-to-eat meals
House arrest: ‘Highly unreasonable’ for Dusuki to dither, says former AG
Busan grandmothers over 70 debut as trot girl group

Air Pollutant Index

Note: API is calculated based on 24hr running average of the most dominant parameter.

Highest API Readings

    Select State and Location to view the latest API reading

    Others Also Read