Samsung, SK hynix, SanDisk, Kioxia, and Micron are building new manufacturing and packaging facilities to increase DRAM and NAND flash memory capacity [1].

This expansion is critical because artificial intelligence workloads are consuming a disproportionate share of the global memory supply. Without new capacity, the hardware necessary to power evolving AI systems could face severe shortages, stalling industry growth [1, 2].

Most major memory chipmakers are currently running at or near full capacity [3]. The surge in demand has been so aggressive that production slots for 2026 are already almost sold out [3]. To address this gap, the companies are investing in new or expanded fabrication plants to scale their output [1].

These new facilities are not expected to be fully operational until late 2027 or 2028 [1]. The timeline indicates a significant lag between the current demand for AI memory and the industry's ability to physically increase supply. While some reports describe the expansion as a race, others note that capacity is expanding slowly relative to the speed of AI adoption [1, 2].

The focus of the new construction includes both the manufacturing of the chips and the specialized packaging required for high-performance AI memory. This dual approach aims to optimize how data moves between processors and memory, which is a primary bottleneck for large-scale AI models [1, 2].

“2026 production slots are already almost sold out”

The gap between the exhaustion of 2026 production slots and the 2027-2028 operational dates for new fabs suggests a looming supply crunch. This may lead to increased costs for AI infrastructure and could force companies to prioritize high-margin AI clients over traditional consumer electronics, potentially inflating the price of standard hardware.