您的浏览器禁用了JavaScript(一种计算机语言,用以实现您与网页的交互),请解除该禁用,或者联系我们。 [野村]:内存价格持续走强;长期协议助力可持续性 2026年二季度内存价格将强于预期;中国厂商扩产暂未构成大风险 - 发现报告

内存价格持续走强;长期协议助力可持续性 2026年二季度内存价格将强于预期;中国厂商扩产暂未构成大风险

信息技术 2026-03-06 野村 灰灰
报告封面

Global memory EQUITY: TECHNOLOGY Research Analysts Strong memory prices continue; LTAs to help sustainability Asia Pacific Technology 2Q26F memory prices to be stronger than our expectation;China players’ ramp-up not likely a big risk yet CW Chung - NIHKcwchung@nomura.com+852 2252 6075 2Q memory prices likely to exceed our forecasts significantly; trend for LTAs mayaffect share prices positively Eon Hwang - NFIKeon.hwang@nomura.com+822 3783 2318 While price negotiations for 2Q26 are underway, we notice increasing likelihood ofmemory prices surpassing our previous forecasts. We previously argued that additionalprice increases from 2Q onwards are not desirable, as commodity memory prices havegrown more rapidly than our expectation in 1Q, and commodity DRAM/NAND OPMs arecurrently approaching 80%/60%. Moreover, significantly high memory prices are alreadyinducing PC/smartphone price hikes of 10-25%. Recently, buyers in the memory market,however, are seeking supplies at higher prices as they look forward to secure memoryvolume. While a “seller’s market” would usually mean that sellers determine pricing asdemand exceeds supply, recent trend would be a “buyer’s market” with a subtly differentnuance, with buyers’ willingness determining pricing as they are willing to secure memorysupply at higher prices while memory suppliers are rather passively taking higher pricesproposed by customers, in our view. We think it is likely that prices for both DRAM andNAND will increase by more than 30-50% q-q in 2Q26F, allowing OPM to increase to mid-80% level (vs. our earlier forecast of mid-single digit q-q increase for DRAM price and20% q-q increase for NAND price; see Fig. 12 in ourprevious report). Amid such a trend, customers are seeking long-term agreements (LTAs) of three years orlonger. Historically, supply contracts used to be closer to MOUs with weak binding, andDRAM suppliers were also hesitating to make binding deals as they might have to forgoupside potentials during price up-cycles. Even when LTAs were made, they did not havesignificant implications in the past, as many contracts were cancellable without penaltieswhen market situations change. Recently, however, the LTAs for certain portion ofvolumes for key customers are being made. While the fixed pricing for DRAM was madeon a biweekly basis 30 years ago, currently the terms are getting longer to monthly andquarterly basis, and we now even see yearly-basis supply contracts for HBMs (high-bandwidth memory). Although such a trend is in its early stage, we believe memoryplayers’ valuation multiples could approach foundry players’ levels as such businessmodels get widespread, and we think this would be highly positive to the share prices ofmemory players by reducing the valuation discounts on earnings. China players’ endeavor for capacity ramp-up amid memory super cycle; memorymarket to be bipolarized Along with such positive trends, we see China players such as CXMT (unlisted) andYMTC (unlisted) raising capex. CXMT is constructing a fab with 50k/month (phase 1)capacity in Shanghai, in addition to its Hefei fab (300k/month), and is trying to bringforward the fab opening and equipment installation schedule to 2H26 from the originalplan of 2027, according to amedia report. We believe CXMT will likely utilize itsShanghai fab to produce HBM2E and HBM3 to meet China domestic customers’ demand,considering that the company has been developing HBM2E andannouncedplans toproduce HBM3 in 2026. Despite the US government’s restrictions on semiconductorequipment exports, CXMT plans to procure equipment for capacity ramp-up throughforeign equipment already procured in advance, equipment localization (we estimatelocalization rate to be 30%) and efficiency improvement for existing equipment. Weexpect the yield rate for HBM would be very low and only dedicated to certain localcustomers; hence, the impact on the global market would be very limited for a while. YMTC, a NAND supplier, is also trying to bring forward the opening of the Wuhan 3 fab(phase 1: 50k) from 2027 to 2H26. While YMTC used to be a NAND player, it is nowplanning to allocate a portion of theWuhan 3 fab for DRAMproduction. As YMTC is alsosubject to US equipment export restrictions, we believe it is likely trying to build capacityby utilizing local equipment vendors (50-60% localization for NAND) and foreignequipment already procured, similar to CXMT’s strategy. We assume that YMTC plans toproduce in a manner that it does not to compete with major DRAM companies, but toproduce some legacy DRAM to be used for its own SSD production. Currently it is uncertain how fast China players could expand capacity under USgovernment restrictions, and how much they could improve yields using China-madeequipment. However, thanks to currently super-high commodity memory prices andsevere shortage, we believe China memory players will be able to generate profits evenwith 1Znm-1Anm technology (2-3 generations lagging vs. majo