"The next wave of stocking up emerges! AI consumes MLCC production capacity: delivery time exceeds 20 weeks, urgent shortage of 'electronic industry rice'"

When the market is still focused on the GPU, HBM, and advanced packaging capacity battle, a wave of construction of artificial intelligence infrastructure is pushing another key electronic component into the supply chain storm Multilayer Ceramic Capacitors (MLCC). Industry insiders point out that after GPU, HBM memory, and advanced packaging, MLCC is becoming another potential "bottleneck" in the expansion process of AI infrastructure. Multilayer ceramic capacitors, known in the industry as the "rice of the electronic industry," are experiencing an unprecedented shortage crisis. According to the latest data from several industry organizations, the spot price of high-end MLCC has risen approximately 20%, with the delivery period extending from the usual 8 to 12 weeks to over 20 weeks, and some tight specifications may take up to 24 weeks to be delivered. Bill Tang, the chairman of Murata, even stated that demand for high-end MLCC has reached the highest level in more than twenty years, with a delivery period of over 20 weeks, and the supply situation is expected to further tighten in the next 18 months. This crisis is not a simple fluctuation of the traditional supply and demand cycle, but rather a structural change where a "supporting" component is being pushed to the center of the stage under the frenzy of AI computing infrastructure expansion. MLCC has climbed to the third place in the materials cost of AI servers, just behind GPU and memory. A top-tier AI server rack may require up to approximately 60,000 MLCCs working together. The consumption level of MLCC in AI servers is expanding at an unprecedented rate. Morgan Stanley's BOM breakdown shows that the value of MLCC content in a single rack of Nvidia's next-generation VR200 NVL72 is approximately $4,320, an increase of about 182% from the previous GB300's approximately $1,530. A brokerage analyst who has long been following the semiconductor industry told Economic Observer that for an example of a MLCC product with the largest quantity used in AI servers, the number of MLCCs around each GPU increased from over 200 in the H100 era to close to 500 in the B200 era, reaching nearly 1,500 in the GB200 era, and the latest Vera Rubin platform is close to 5,000. For four generations of products, the quantity of the same MLCC used in a single GPU has increased more than twenty times. Behind the surge in demand, the supply side is facing difficulties in responding quickly due to technological and capacity constraints. High-end MLCCs used in AI servers need to increase the number of ceramic layers from 100 layers to over 1,300 layers, with an extremely complicated production process, doubling the product completion cycle, with the current product qualification rate for AI MLCCs only around 40%. "To make a high-end product for AI use, the capacity occupied is equivalent to four ordinary products, and no matter how many production lines are added, they cannot keep up," described a market manager from a South China capacitor manufacturer. Therefore, despite AI servers accounting for only 2% to 3% of global MLCC shipments, they consume approximately 10% of the entire industry's capacity. More importantly, this structural imbalance is moving towards long-term development. Tsuneo Nakashima, President of Murata Manufacturing, stated at an analyst meeting in May that the global trend of constructing AI data centers and servers will continue to heat up, and the strong demand for high-end MLCCs can at least be sustained until 2030. The wave of price increases has spread from the high-end to the general market Faced with continuing tightness in the supply chain, global MLCC industry leaders have initiated a relay-style price increase. The earliest action was taken by the industry leader Murata Manufacturing in Japan....