Samsung to More Than Double HBM4 and HBM4E DRAM Output Next Year

Samsung Electronics is preparing a major ramp-up of its next-generation high-bandwidth memory, with supply chain sources indicating the company intends to more than double output of HBM4 and HBM4E DRAM next year. The move signals that the Korean giant is betting heavily on an AI-driven memory supercycle that shows little sign of cooling.

Why the HBM4 ramp matters now

HBM4 is the memory standard expected to sit alongside next-generation AI accelerators, and HBM4E pushes bandwidth and capacity even further. Doubling production is not a routine capacity tweak; it reflects how aggressively hyperscalers and chip designers are queuing up for stacked DRAM that can feed ever-larger AI models. Samsung's decision to scale both HBM4 and HBM4E simultaneously suggests it wants to serve both the current accelerator generation and the one that follows.

The timing is significant. Memory supply and demand have remained unusually tight, and analysts argue that tightness will persist well into the next cycle. That environment gives Samsung leverage on pricing and lets it prioritise high-margin HBM over commodity DRAM. Competitors are moving in parallel, but the scale of Samsung's planned expansion is one of the clearest signals yet that the AI memory race is entering a new phase.

Goldman Sachs and the 490,000 KRW target

Goldman Sachs Research has weighed in with a target price of 490,000 KRW for Samsung, framing the memory market as still supply-constrained. The note reinforces the view that the company's earnings trajectory is tied less to consumer electronics and more to the data-centre buildout. For investors, the combination of a doubling of HBM4 output and a bullish analyst target paints a picture of a company repositioning itself around AI infrastructure.

Not everyone is equally enthusiastic about the broader memory trade. Some analysts have described rival memory names as likely dead money from here, arguing that much of the easy upside has already been priced in. That scepticism makes Samsung's production plans more important, not less: execution on HBM4 and HBM4E will determine whether the company converts tight supply into durable profit.

What to watch next

The key questions are qualification timelines, yield rates on advanced stacking, and how quickly customers commit to HBM4E. If Samsung hits its doubling target, the memory landscape could look very different by the end of next year, with consequences for pricing, competitor roadmaps and the cost of training frontier AI models. Readers tracking the AI hardware cycle should keep a close eye on Samsung's production updates and analyst revisions in the coming quarters.