en.Wedoany.com Reported - Japan's Kyocera has announced an expansion timeline, planning to invest approximately 100 billion yen (about 4.2 billion yuan) over the seven fiscal years ending March 2031, specifically to boost production capacity for high-end multilayer ceramic capacitors (MLCCs) required for AI servers. This decision stems from the rising power consumption of chips driven by the scaling of generative AI models, making MLCCs a critical component for the stable operation of AI servers.
Kyocera President Shiro Soda stated that the vast majority of funds will be allocated to the Kokubu site of the Kirishima Factory in Kagoshima, gradually increasing the supply capacity of high-spec MLCCs through new plant construction and phased equipment installation. A single AI server can require 5 to 10 times more MLCCs than a traditional server, typically ranging from 15,000 to 25,000 units, with strict demands for small size, high capacitance, and high-temperature resistance.
Data from market research firm TrendForce shows that MLCCs are accelerating their replacement of traditional aluminum electrolytic capacitors and tantalum capacitors. For example, on the AMD MI450 platform, the demand for 47µF 0402-spec MLCCs per board surged from 1,440 units to 10,544 units during the verification period, an increase of 632%; on NVIDIA's Vera Rubin platform, the usage of 100µF 0805-spec MLCCs per board was raised from 320 units to 500 units. This trend of rising volumes and prices is putting pressure on the MLCC supply chain.
Supply-side expansion faces bottlenecks. Due to technical barriers such as material formulations, lamination processes, and sintering yields, effective capacity growth for high-end MLCCs is slow. Although Murata Manufacturing has announced an 80 billion yen expansion plan, its new Izumo plant is not expected to reach full production until 2027. The book-to-bill ratio of major Japanese and South Korean manufacturers has been rising month by month since April this year, with delivery lead times for some high-capacitance products extending from the usual 8 weeks to 20 weeks. The industry expects that from the second half of 2026, as self-developed chip platforms like Google TPU, Amazon Trainium, and Meta MTIA ramp up production, high-end MLCCs may face a structural shortage, with this boom cycle potentially lasting until 2028 or even 2030.
The supply-demand imbalance is driving price increases. According to industry sources, companies such as Murata and Samsung Electro-Mechanics have raised prices for high-capacitance MLCCs used in AI servers, with increases ranging from 15% to 35%.
Facing AI-driven demand for passive components, the three major Japanese manufacturers are accelerating expansion. Kyocera is making a long-term bet on high-spec MLCCs for AI with its seven-year, 100 billion yen plan; Murata Manufacturing began mass production of new 0402-size products like 47µF MLCCs by the end of 2025 and has added an additional 80 billion yen in investment, with current capacity utilization near 95%; Taiyo Yuden has raised its annual MLCC capacity growth rate from 10% to 15% to meet urgent order demands from cloud service providers.
