Memory chips, including RAM and NAND, have become some of the most valuable commodities in the technology industry because of growing demand from AI data centres. However, Elon Musk believes the situation could become even more challenging once humanoid robots begin to be manufactured on a large scale.
Musk responded to a discussion on X that referenced a Counterpoint Research report projecting a sharp increase in RAM requirements for humanoid robots. According to the report, global DRAM demand from these robots is expected to grow more than 20-fold, rising from approximately 8,300TB in 2026 to around 172,000TB by 2030.
Musk dismissed the analysis, arguing that the memory requirements of humanoid robots would be considerably higher. "Silly Analysis. Robot RAM needs will exceed everything else on Earth, as you need the robots to be on physical substates in order to shape atoms," he wrote.
Demand for memory chips from AI data centres has already put considerable pressure on the industry. RAM prices have risen sharply over the past year, pushing up the cost of consumer electronics, including smartphones and laptops. The shortage has affected not only smartphone manufacturers but also Indian businesses. Zoho founder Sridhar Vembu has spoken about the difficulties companies are facing as memory prices continue to climb.
But there may be some good news
Musk has also outlined a vision that could potentially reduce some of the pressure associated with the expansion of AI and humanoid robotics. In the same post, he said superintelligence, a term also used by US President Donald Trump when discussing advanced AI, would depend on computing infrastructure located in space.
"Super intelligence needs will vastly exceed that, but \~100 per cent of SI will be deep space compute. Obviously," Musk added.
Although Musk did not explain whether this approach could help stabilise RAM prices, moving computing infrastructure into space could theoretically reduce the need for data centres on Earth.
Returning to humanoid robots, Musk appears to believe that their memory requirements will be significantly greater than the estimated 39GB of DRAM capacity per robot cited in the report.
Humanoid robots need to process real-time information from cameras, LiDAR systems, tactile sensors and other sources to understand their surroundings and determine how to move. More sophisticated AI systems, including world models designed to predict changes in the environment, would require additional memory and computing power.
Memory manufacturers are also paying attention to the robotics sector. Micron chief executive Sanjay Mehrotra highlighted the potential requirements during an earnings call.
"Level 4 or higher autonomous vehicles typically require over 200GB of memory and several terabytes of storage. Humanoid robots will likely need similar levels of memory and storage," he said.
Humanoid robotics is already advancing, with Chinese companies such as Unitree making significant progress in developing fast-moving robots. Elon Musk's Tesla has also been working on its Optimus humanoid robot for some time.
Only time will reveal whether Musk's prediction proves accurate. For now, however, the ongoing RAM shortage suggests that pressure on the memory market may persist for the foreseeable future.
