Samsung Warns Memory Chip Shortage Could Deepen Through 2027 and Persist Into 2028

Samsung Electronics stated that the ongoing shortage of high-end memory chips, exacerbated by skyrocketing demand for artificial intelligence (AI) hardware, is likely to persist through 2027 or possibly even 2028. “As the number of AI data centers and high-performance computing systems expand rapidly, demand for high-end memory products used for AI-centric powerful hardware is surging at an unprecedented level,”Samsung Electronics said, referring to high-bandwidth memory (HBM) and premium DRAM chips. “Memory chip has become the most essential component for today’s AI systems, Unlike general-purpose workloads which perform tasks with slower speeds at low-speed memory, large language model development, as well as training, require a significant amount of high-speed memory that transfers data to processors more quickly because LLMs need enormous capacity of memory to overcome information gaps in a larger number of tasks” the South Korean tech giant explained.

Given its status as one of the world’s largest memory suppliers, Samsung acknowledges that the supply situation may remain tense.

Samsung believes that even though multiple semiconductor companies have ramped up their efforts to scale up capacity, it will take several years for new fabs to be built. It also acknowledged that the demand spike has been heavily driven by AI hardware. In addition to Samsung, rival memory chipmakers have also expanded output but industry experts indicate that even with this growth, capacity is still anticipated to struggle to keep pace with projected AI requirements in the coming years, which could potentially impact the delivery of AI infrastructure. However, experts added, “this shortage will be the strongest among premium memory products.

While the cost of cloud computing may be impacted, sustained demand should provide opportunities for memory chip makers to invest in supply.”

Furthermore, governments worldwide have announced initiatives to bolster their respective domestic semiconductor industries through incentives for building fabrication plants. Even with investments from all sides-including significant government support-Samsung expects equilibrium in the market will be a protracted process, “due to the nature of long time-lines for construction and increasing application of AI and its services.” The company anticipates significant growth opportunities in the semiconductor market through 2028, all driven primarily by the acceleration of the adoption of generative AI across various enterprises globally, although the memory crunch itself has made this difficult.