The memory market faces a potential shift as new demand sources emerge beyond traditional data centers. Counterpoint Research forecasts that DRAM demand could skyrocket to 172,000 terabits by 2030, up from 8,300 terabits this year, driven largely by the rise of humanoid robots. These advanced machines, expected to start arriving next year, will require significantly more memory, with average DRAM capacity per unit doubling from 19 gigabytes to 39 gigabytes over the same period. The demand is fueled by the complex processing needs of these robots, which handle image, video, and sound data simultaneously and use 'world models' to predict environmental changes, requiring faster processing and greater memory capacity.
Beyond robotics, the automotive industry is also poised to drive memory demand. Highly autonomous vehicles, which typically require over 200 GB of memory and several terabytes of storage, are expected to become more prevalent. Micron CEO Sanjay Mehrotra noted that humanoid robots will likely have similar memory and storage requirements, further intensifying demand. Meanwhile, Goldman Sachs analysts are closely watching Samsung's earnings report for insights into whether memory shortages will persist into 2028, as major memory firms continue to warn of tight conditions through next year.
The rise of physical AI, including humanoid robots, represents a significant shift in the memory market. While data centers have been the primary drivers of demand so far, the coming wave of AI-driven applications could reshape the landscape. Eric Sheridan, a Goldman Sachs analyst, recently raised his forecasts for humanoid robot shipments, predicting 890,000 units by 2030 and 6.5 million by 2035. This growth is expected to create a $138 billion market opportunity, signaling a new era of demand for memory chips. As these technologies develop, the pressure on memory suppliers to meet rising demand will intensify, potentially affecting global pricing and availability.