上调至买入：AI服务器拉动MLCC盈利增长加速

Murata (6981.T) Upgrading to Buy - AI servers fuel acceleration in MLCC profit growth CITI'S TAKE We upgrade Murata to Buy from Neutral and raise our target price to¥15,000from¥3,900,reflecting revisions to our earnings forecasts,changes to our valuation assumptions (risk premium down from 4% to 3%),and a shift in our base year (from FY3/28 to FY3/29). We raise our earningsforecasts significantly, reflecting our increased AI server unit outlook andfavorable MLCC pricing in the AI server space. While we do not currentlyanticipate company-led price revisions for general-purpose products, wenote that such revisions would create additional upside. MLCC outlook— Murata is now in a phase of accelerated MLCC sales/profit growthin the AI server space. Rising capacitance has led to unit price hikes, with theresulting improvement in product mix boosting profit growth significantly. Given thehigh entry barriers for high value-added products, we think it is relatively easy forthe firm to raise unit prices for server applications, meaning these hikes are likely tobe sustained. We do not currently assume price hikes driven by supply-demand. OP forecasts— We raise our OP forecasts from ¥380bn to ¥440bn for FY3/27(+56.1% YoY) and from ¥426bn to ¥630bn for FY3/28 (+43.2%), and newly forecast¥800bn for FY3/29 (+27.0%). The big increase in our forecasts reflects our increasedAI server unit outlook and a higher MLCC ASP (in yen) thanks to an improved productmix (+14% YoY for FY3/27). Share price view— Our target price (¥15,000) is calculated using RoIC-WACC. Itequates to an FY3/29E PER of c43x. We consider this level acceptable to someextent given the widening of the RoIC-WACC spread from accelerating profit growth.Upside risks include the potential for broad-based MLCC price increases. While wecurrently see the likelihood as low given the company's long-term strategy, weestimate that a 10%+ price increase across all products would yield a fair value ofc¥18,000. For April-June results, we expect a rising BB ratio to act as a tailwind forthe stock price. Takayuki NaitoAC+81-3-6776-4644takayuki.naito@citi.com Fundamentals: Favorable AI server-relatedprices and volumes to drive profit growth We raise our OP forecasts for Murata from ¥380bn to ¥440bn (+56.1% YoY) forFY3/27 and from ¥426bn to ¥630bn (+43.2%) for FY3/28, and we add a forecast of¥800bn (+27%) for FY3/29 (Figure 1). Our outlook for stronger profit growth factorsin 1) continued growth in demand for AI-server MLCCs supported by a sustainedincrease in AI server volume, and 2) revisions to our unit price assumptions for AI-server MLCCs, incorporating a smaller decline in MLCC unit prices and a resultantsustained increase in the ASP (Figure 2). We assume an exchange rate of¥160/$ from July-September 2026. AI server-related demand to accelerateprofit growth; we make large upwardrevisions to forecasts Growth in the number of AI servers is accelerating year by year (Figure 3). We hadexpected volume growth to slow from CY27, but we now forecast it will expand to+27% in CY27 from +15% in CY26. While we expect this to be positive for allelectronic components, we think it will have a particularly large impact on MLCCvolumes (Figure 4). Changes in sales and profit growth drivenby an increase in the number of AI servers In AI servers, the increase in power consumption and sharp fluctuations in electriccurrent caused by accelerators place a heavy load on passive components. As aresult, the number of MLCCs per board is expected reach around 11,000 for GPUsin CY26 compared with around 2,000 for general-purpose servers. The increase in required capacitance is raising the added value of MLCCs andcreating a high barrier to entry. While regular servers use MLCCs with a capacitanceof around 10μF, most current AI-server MLCCs have a capacitance of 47μF.Considering the high capacitance required to deal with sudden voltage drops andalso mounting area constraints, we forecast 100μF MLCCs will become standardgoing forward, resulting in a continued increase in MLCC capacitance and unitprices. Raising capacitance leads to an increase in the number of thin film layers. This, inturn, increases the production load and reduces production yields. We believe themore challenging manufacturing process constitutes a barrier to entry for AI server MLCCs. We recognize Murata, Samsung Electro-Mechanics, and Taiyo Yuden asthe three main suppliers of high-capacitance MLCCs for AI servers, although webelieve Murata has the highest market share (c50%) thanks to its superiorproduction yield. The increase in capacitance per board leads to a significant increase in the MLCCproduction load. Murata has announced an additional capacity expansion of 20%on top of its normal annual expansion (around 10%). However, further increasesmay be necessary, depending on demand. The increase in investment for AI serverMLCCs can be passed on to MLCC prices, and we believe the announcement of theadditional investmen