Demand for DRAM used in humanoid robots could rise more than twentyfold by 2030, creating a potential new market for South Korean memory producers. Counterpoint Research forecasts that total demand from the sector will increase from about 8,300 terabits in 2026 to 172,000 terabits in 2030, a 20.7-fold expansion.
The forecast reflects growth in both the number of humanoid robots and the amount of memory installed in each machine. Average DRAM capacity per robot is expected to more than double from 19 gigabytes this year to 39 gigabytes by 2030. These estimates remain projections and will depend on how quickly humanoid systems move into wider use.
Memory requirements are rising as robots process information from cameras and sensors in real time. More advanced artificial intelligence models need sufficient capacity to interpret those inputs and support autonomous decisions, linking the sophistication of a robot's software directly to its onboard memory needs.
Battery limits favor low-power memory
Counterpoint expects power-efficient DRAM, particularly low-power DDR products, to benefit from this shift. Because humanoid robots operate on batteries, power consumption is a significant constraint. The research firm therefore sees stronger demand for low-power DRAM and high-capacity NAND flash than for high-bandwidth memory, which consumes more power.
The projected expansion offers Samsung Electronics and SK Hynix another source of demand beyond established computing markets. Both South Korean companies are seeking an early position in memory for humanoid robots, although the forecast does not establish how much of the future market either company will capture. The key indicators to watch are robot deployment volumes, memory capacity per unit and whether manufacturers favor low-power components as expected.