全球内存：新内存长期协议第三部分——多空辩论

Global Memory: New Memory LTAs Part 3 - Bull vs Bear Debate -Webinar slides, transcript and summary Following our deep dives into New Memory LTAs (seePart 1andPart 2), on July 9thwehosted the webinar New Memory LTAs - glass half full or half empty? This note summarizesthe webinar with slides and Q&A. A replay is availablehere. Mark C. Newman+1 212 845 7822mark.newman@bernsteinsg.com Mark Li+852 2123 2645mark.li@bernsteinsg.com LTAs won’t remove memory cyclicality, but they can meaningfully reduce theseverity of downturns by raising the earnings floor and improving revenue and profitvisibility.LTAs effectiveness ultimately depends on whether the financial guaranteescharged are substantial enough to offset the savings customers could realize by purchasingmemory at prevailing lower market prices. LTAs can provide meaningful downsideprotection, but remain vulnerable to abandonment if market prices fall substantially. April Li+1 917 344 8339april.li@bernsteinsg.com Phoebe Sun+1 917 344 8481phoebe.sun@bernsteinsg.com While LTAs are unlikely to penetrate into the entire end market, the customers mostinclined to enter these agreements - hyperscalers and AI infrastructure providers- are also among the industry's higher-quality counterparties.With strong balancesheets, diversified revenue streams, and strategic incentives to secure long-term supply,these customers may be better positioned to honor their commitments. Micron takingequity stakes of Anthropic is however an example of vendor financing & risk amplification.In addition, unlike prior semi-cycles that were inflated by inventory hoarding or double-ordering, today's demand is more supported by genuine infrastructure deployment for AI. Edward Hou, CFA+852 2123 2623edward.hou@bernsteinsg.com Yipin Cai, CFA+852 2123 2669yipin.cai@bernsteinsg.com Although financial guarantees appear modest relative to the massive multi-yearrevenue streams they are designed to protect,these New Memory LTAs becomeincreasingly protected over time as the guarantees are back-end weighted.Asremaining purchase obligations decline, the financial cost of walking away increasesrelative to contract value. Hence, downside protection can improve materially in the lateryears of the agreements (when it is needed most!). On industry fundamentals, we remain constructive on near-to-medium-termmemory supply-demand dynamics.We do not expect Chinese entrants to materiallydisrupt the market over the next two years, with industry supply growth driven primarilyby capex decisions at mature incumbents. We also expect HBM pricing to move higher asconventional memory becomes increasingly profitable. Where views diverge is on NAND:Asia team is more cautious, arguing that AI demand disproportionately benefits DRAM andthat Chinese competition poses a greater long-term risk in NAND. U.S. team believes AIworkloads are becoming increasingly NAND-intensive through larger context windowsand growing storage requirements, creating a favorable supply-demand backdrop amidconstrained supply. We highlight a clear divergence in how memory suppliers are approaching LTAs.Sandisk and Micron have been the most proactive in pursuing agreements backed byfinancial guarantees, while Asia based suppliers have taken a more selective approach.Looking ahead, both teams view further LTA disclosures, memory ASP, and Sandisk'supcoming Analyst Day as key catalysts that could shape investor perception of both LTAsand memory sector valuations. BERNSTEIN TICKER TABLE INVESTMENT IMPLICATIONS We rateSNDKOutperform with PT of $3000. Samsung Electronics:We rate Samsung Electronics Outperform with price target of KRW 440,000. SK hynix :We rate SK hynix Outperform with price target of KRW 3,300,000. Micron:We rate Micron Outperform with price target of US$1,300.00. KIOXIA: We rate KIOXIA Underperform with price target of JPY 40,000. Table Of Contents Part I: Bull vs Bear................................................................................................................................................................................................................... 3The Bear Thesis: LTA Limitations and Constraints................................................................................................................................................3The Bull Thesis: The New Memory LTA’s Benefits and Protections.............................................................................................................11Part II: Q&A.............................................................................................................................................................................................................................. 18 DETAILS Following our deep dives into New Memory LTAs (seePart 1andPart 2), on July 9thwe hosted the webinar New Memory LTAs -glass half full or half empty? This note summarizes the webinar with slides and Q&A. A replay is available here. The webinar feature