Elon Musk Weighs In on China’s Memory Chip Ambitions and the Challenge of 3D DRAM
Elon Musk says China remains one of the world’s strongest forces in scaling advanced manufacturing, but he believes the country may face tougher challenges as the semiconductor industry moves toward next-generation memory technologies such as 3D stacking, HBM-style architectures, and memory-on-logic designs.
Musk shared his views in a recent discussion on X, where he responded to questions about China’s ability to produce advanced RAM chips at competitive scale. His comments come as China’s memory chip industry continues to accelerate, with local manufacturers pushing into technologies that have traditionally been dominated by global leaders in South Korea, the United States, and Taiwan.
China’s CXMT recently attracted attention after outlining plans to debut DDR5 memory chips based on hybrid bonding and 4F² cell technology by the end of the year. The announcement stood out because 4F² is viewed as a major step forward for the DRAM industry. Instead of arranging memory cell components primarily across a flat surface, 4F² allows elements to be stacked more efficiently, reducing chip area and potentially improving performance.
That matters because modern memory chips are running into physical limits. As smartphones, AI servers, electric vehicles, and data centers demand more bandwidth and higher capacity, chipmakers are searching for new ways to increase density without dramatically increasing power consumption or cost.
Musk acknowledged that China’s manufacturing strength should not be underestimated. He pointed to industries such as lithium batteries and solar panels as examples where China has shown an extraordinary ability to scale production after refining the required manufacturing equipment and processes. In his view, China’s ability to ramp up production in certain sectors can be greater than the rest of the world combined.
However, he also suggested that memory chips may not be as simple to scale as solar panels or batteries. While China appears to be making fast progress on standardized memory platforms such as LPDDR5 and future LPDDR6 chips, Musk believes more complex architectures could be harder to master.
The technologies he highlighted include 3D DRAM, HBM-style memory, and memory-on-logic designs. These approaches are becoming increasingly important as artificial intelligence workloads place huge pressure on memory bandwidth and energy efficiency.
3D DRAM involves stacking memory layers vertically inside a chip, allowing manufacturers to increase density without relying only on shrinking transistor dimensions. High-bandwidth memory, or HBM, also uses stacked memory dies connected through advanced packaging, making it especially valuable for AI accelerators and high-performance computing chips.
Another emerging concept is memory-on-logic, where memory is placed very close to or directly above processing components. Qualcomm has explored an approach that places an AI accelerator beneath LPDDR memory, aiming to improve bandwidth, reduce latency, and lower overall system costs.
These technologies are difficult because they require more than just strong factory output. They depend on extreme precision in materials engineering, bonding, etching, thermal management, and chip packaging. Even small defects in stacked structures can reduce yield and make mass production expensive.
Musk also compared China’s scaling ability with SpaceX’s performance in aerospace manufacturing. He described SpaceX as an exception to the broader pattern, noting that his company has consistently outpaced China in the amount of payload delivered to orbit. Much of that launch volume has come from Starlink, SpaceX’s satellite internet network, which has placed thousands of satellites into orbit to expand global broadband coverage.
Still, when it comes to semiconductors, Musk left room for surprise. He suggested that RAM chip production could become another area where China manages to scale faster than expected, especially for memory products built around widely adopted global standards.
China has already shown signs of progress in advanced DRAM research. Earlier this year, Chinese semiconductor equipment company Naura reported a breakthrough involving a 64-layer 3D DRAM stack using silicon and silicon-germanium layers. The company demonstrated a two-step process designed to improve key manufacturing metrics, including etch selectivity and material loss rate.
Such developments show that China is not only working to manufacture existing memory technologies but is also attempting to build the equipment and process knowledge needed for future chip architectures.
The race for advanced memory chips is becoming increasingly important as AI systems, smartphones, autonomous vehicles, and cloud platforms demand faster and more efficient data movement. Processors may receive much of the attention, but memory is now one of the biggest bottlenecks in computing performance.
Musk’s comments reflect a broader reality in the semiconductor industry: China has proven that it can scale manufacturing at remarkable speed, but the next phase of chipmaking may test the limits of that advantage. Producing advanced memory at high volume, competitive cost, and strong yield will require deep expertise across design, materials, equipment, and packaging.
For now, China’s progress in DDR5, LPDDR, 3D DRAM, and advanced chip stacking remains one of the most closely watched developments in the global semiconductor market. If Chinese firms can turn research breakthroughs into mass production, the balance of power in the memory chip industry could shift significantly in the coming years.
