scmp.com web signal

MLCCs Surge Up to 4x in Shenzhen as AI Server Boom Strains Supply

chips china ai ai infrastructure ai-infrastructure

TL;DR

  • High-capacity MLCCs at Shenzhen's Huaqiangbei market have jumped two to four times in price since Chinese New Year.
  • A Murata Manufacturing distributor says inquiry volume is high but actual transactions are slow, as buyers resist elevated prices.
  • Murata reportedly holds around 40% of the global MLCC market and an even larger share in the AI server-grade segment.

The smallest components in an AI server rack are now among the hardest to source at a reasonable price. At Shenzhen's Huaqiangbei electronics market, spot prices for high-capacity multilayer ceramic capacitors (MLCCs) have jumped two to four times since Chinese New Year, with some models climbing from 10 yuan to 40 yuan per 1,000 units, according to the South China Morning Post. MLCCs are compact power-regulation components, no larger than a postage stamp, used in virtually every circuit board. The global AI server boom and electric vehicle manufacturing are both driving demand at the same time, straining a supply chain that was not built for this pace.

A distributor identified as Wu, who mainly sells components from Japanese industry leader Murata Manufacturing, told the Post: "I don't see any signs of prices cooling down." Wu noted that inquiry volume is unusually high but actual purchase transactions are slow, a sign that many buyers are resisting the elevated prices rather than absorbing them. The reporting notes that customer inquiries for high-end MLCCs are running at double existing manufacturing capacity and described by one source as "completely unsatisfiable." Murata reportedly holds around 40% of the global MLCC market and a larger share (between 45% and 70%) in the AI server segment specifically.

The honest caveat is that Huaqiangbei spot prices are a volatile leading indicator, not a settled benchmark. Whether those rates translate into the contract pricing that hyperscalers and large OEMs actually pay is a different question, and the reporting does not address it. Missing from the account is any distinction between genuine supply scarcity and speculative forward-buying by distributors anticipating further increases, a dynamic that has amplified price moves in electronics spot markets before and can reverse sharply.

For anyone building or procuring AI infrastructure, the MLCC story echoes the GPU supply crunch in one important way: it is not only the headline silicon that creates bottlenecks. Cheap, unglamorous components can become the binding constraint, and by the time the spot price is moving fourfold at Huaqiangbei, the lead times are typically already long.