Samsung Electronics shares appear appealing following a steep pullback, Morgan Stanley analyst Shawn Kim wrote, with the sell-off presenting what the firm views as a buying opportunity ahead of an anticipated change in AI memory architecture.
In its note, Morgan Stanley said Samsung's shares have fallen roughly 20 percent week to date, versus about a 17 percent decline in the KOSPI. The firm links the move to broader market pressures but underscores the tactical window it creates for investors.
Kim argued the memory landscape for AI inference is evolving toward a hybrid approach as chips become increasingly specialized. While high-bandwidth memory, or HBM, remains the dominant solution for many workloads, the analyst noted that "SRAM is carving a niche for workloads where latency matters more than throughput density."
Morgan Stanley expects Nvidia to unveil a new inference chip at its upcoming GPU Technology Conference that will adopt a Language Processing Unit, or LPU, architecture relying on substantial on-chip SRAM. The firm described that architecture as "purpose-built for the sequential speed of inference" and attractive to customers "willing to pay for speed."
Rather than framing the future as a contest between SRAM and HBM, Kim suggested the market is more likely to support a hierarchical arrangement combining SRAM for hot-path execution with HBM for scalable capacity. The note also highlighted a potential supply-chain implication: LPU designs could sidestep current bottlenecks in HBM production and CoWoS packaging.
Despite uncertainty over how the AI memory market ultimately fragments, Morgan Stanley kept Samsung Electronics as its top pick. The firm cited several factors supporting that stance, including HBM4 qualification, SRAM capabilities, flexibility in foundry operations and exposure to a wider commodity upcycle. Morgan Stanley also reiterated a favorable view on SK hynix.
Kim concluded by noting that "historically, such corrections have offered a good opportunity to buy," and said earnings expectations have "a lot of room for recovery." The analyst's view frames the recent price weakness as a potential entry point for investors focused on semiconductor memory dynamics and AI-related demand shifts.
Sectors affected: semiconductor memory, AI hardware, foundry services, and broader technology supply chains.