The AI server construction boom has turned "electronic rice" MLCCs (Multi-Layer Ceramic Capacitors) into a hot commodity, now even jokingly called "electronic gold." Spot prices for high-end products have surged by as much as 3 to 5 times this year, with quotes in Shenzhen's Huaqiangbei market changing as fast as every 30 minutes. A structural supply shortage, driven primarily by AI demand, is sweeping the global MLCC market.
According to reports, this round of price increases is concentrated on high-capacity, high-temperature-resistant specifications like 22μF and 47μF. For some AI server-specific products, the price per thousand pieces has skyrocketed from around 80 RMB at the start of the year to over 350 RMB. In markets like Shenzhen and Shanghai, prices for certain products are even fluctuating daily. Since the rally began in late February, overall MLCC spot prices have risen by 15% to 20%, with high-capacity components for AI servers seeing even steeper increases of 50% to 60%.
Supply Side Under Intense Pressure
Major manufacturers like Murata are tightening shipment controls. For orders in the millions to tens of millions of pieces, actual delivery rates are only about 10% to 20%. Some overseas suppliers have even reinstated bundled sales strategies. Murata, Taiyo Yuden, and Samsung Electro-Mechanics have collectively raised prices. Capacity utilization for high-end production lines has exceeded 90%, and lead times for tight-supply orders have extended to approximately four months.
Global Cloud Service Providers (CSPs) are accelerating the deployment of their own ASIC accelerators, fundamentally reshaping the MLCC count on each board from the demand side. Taking AMD's MI450 platform as an example, a design change caused the number of 47μF specification MLCCs per board to surge by 632%. Given that the construction cycle for high-end production lines is 18 to 24 months long, analysts believe the risk of a structural shortage emerging in the second half of 2026 is already difficult to ignore. Goldman Sachs predicts that MLCCs are the third-largest cost component in AI servers after GPUs and memory, with related demand expected to grow 4.3 times from 2025 to 2030.
Huaqiangbei Spot Market: Price Hikes Across the Board
A trader in the Huaqiangbei market revealed that spot prices for some MLCC models have surged from about 10 RMB per thousand pieces to around 40 RMB. Another supplier, representing both Samsung Electro-Mechanics and domestic manufacturer Chaozhou Three-Circle Group, stated that popular specifications like 0805 22μF have seen increases exceeding 20 times.
The price hikes are not limited to high-capacitance products. A trader specializing in Samsung MLCCs noted that since May, prices for all models in their portfolio have risen significantly, with the price per reel for some specifications climbing from 100 RMB to 400-500 RMB. Supply tightness has intensified concurrently. In an inquiry for Murata's 1206 47μF (476), 10V, X5R specification product, four out of five traders reported no spot availability.
Domestic manufacturers are also feeling the supply-demand pressure. It's reported that some products from Fenghua Advanced Technology are now quoted at levels comparable to imported brands. Lead times for high-end specifications domestically have stretched from the usual 3-6 weeks to over 20 weeks, while ordinary specifications now exceed 12 weeks.
AI Platform Design Iteration: Soaring Per-Board Usage Drives Qualitative Demand Shift
Latest industry research indicates that global CSPs, in the AI arms race, are accelerating the advancement of their own ASIC accelerators, continuously driving up demand for small-form-factor, high-capacitance, high-temperature-resistant MLCCs. Next-generation AI accelerator platforms frequently change designs during final certification stages, directly causing a significant jump in high-end MLCC usage per board.
According to data, during the Bill of Materials (BOM) verification for AMD's MI450 platform, the replacement of all aluminum electrolytic and tantalum capacitors with MLCCs caused the per-board usage of 47μF 2.5V X6S 0402 specification MLCCs to surge from 1,440 pieces to 10,544 pieces, an increase of 632%. On Nvidia's Vera Rubin platform, the per-board demand for 100μF 4V X6S 0805 specification MLCCs also increased from 320 pieces to 500 pieces.
AI servers use approximately 8 to 12 times more MLCCs than ordinary servers. This order-of-magnitude difference, combined with the long-term growth trend in AI compute demand, is reshaping the entire MLCC industry's supply-demand dynamics from a structural perspective.
Supply Bottleneck: High-End Capacity Expansion Takes 18-24 Months
High-end MLCC production lines face high technical barriers and stringent yield requirements, making it difficult for new capacity to come online in large quantities in the short term. The typical timeline from construction to mass production for a high-end line is 18 to 24 months, meaning the tight supply situation is highly likely to persist until the second half of 2026 or even beyond 2027.
Some major manufacturers have shifted capacity from consumer electronics applications like smartphones and PCs towards higher-margin AI servers and automotive applications. While this structural shift follows profit incentives, it further exacerbates the supply-demand imbalance for high-end products. Simultaneously, demand for low-capacitance MLCCs used in consumer electronics remains weak, creating a starkly polarized market.
Supply tightness has also spurred bundled sales practices. Some overseas suppliers have resumed requiring customers purchasing high-capacitance MLCCs to simultaneously buy medium and low-capacitance products. Analysts point out that unlike the across-the-board shortage in 2018, the core contradiction this time lies in the structural imbalance between surging AI demand and limited high-end capacity, not a general supply-demand mismatch.
As leading overseas manufacturers drastically compress shipment rates, both Chaozhou Three-Circle Group and Fenghua Advanced Technology are expected to further expand their market share by capturing transferred orders. Observers note that, given the rigid constraints of the high-end capacity expansion cycle, this supply gap will be difficult to close in the short term, providing domestic manufacturers with a sustained window to capture overflow orders.
Comments