Demand for DRAM for Robots to Surge 20-Fold by 2030

Demand for memory (DRAM) used in humanoid robots that behave like humans is projected to grow roughly 20-fold by 2030 compared with 2026. In a report released on the 11th, market research firm Counterpoint Research attributed this to the combined effects of rising robot shipments and increased memory capacity per robot. DRAM (dynamic random-access memory) is a semiconductor that temporarily stores data and is essential for a robot's brain to process information quickly.

Average DRAM capacity per robot is also expected to more than double, from 19GB in 2026 to 39GB in 2030. This is because robots need more memory to perform more AI computations. The reason lies in the robot's 'brain,' its computing unit. Processing in real time the data generated by cameras, LiDAR, and tactile sensors requires large-capacity storage and compute memory. In particular, memory requirements have grown even larger with the adoption of end-to-end approaches that process multimodal inputs in a unified manner, as well as world models.

The memory configuration is dictated by the fact that robots run on batteries. Since power efficiency is critical, low-power DRAM of the LPDDR family paired with high-capacity NAND flash is mainly used today. The shift to 3D DRAM is cited as a key challenge going forward.

"Level 4 and higher autonomous vehicles typically carry more than 200GB of memory and several terabytes (TB) of storage," Micron CEO Sanjay Mehrotra said at the company's fiscal 2026 fourth-quarter earnings announcement on October 30 (local time), adding that "humanoid robots will require a similar level of memory and storage."