China semiconductor self reliance efforts accelerate CXMT progress in high bandwidth memory as Nvidia AI PC chips launch to drive Samsung and SK Hynix sales
Government efforts in China toward semiconductor self reliance have resulted in swift progress for the nation in the high bandwidth memory market. According to a report published in Seoul Economic Daily, the technological difference between the two leading market firms, Samsung Electronics and SK Hynix, and China's ChangXin Memory Technologies, or CXMT, now is approximately three years. The accelerated progress comes as the Chinese state is actively channeling massive amounts of capital and resources toward Chinese memory manufacturers to challenge South Korea’s dominance in artificial intelligence hardware supply chains.
It was once assumed by market analysts that ChangXin Memory Technologies halted its progress toward the development of the fourth generation HBM3 memory in light of major limitations in production yield rates. Data collected so far demonstrates that the Chinese manufacturer has successfully developed the relevant technology which nearly matches the technical specifications of South Korean HBM3 while still grappling with the low manufacturing yield rates. In order to achieve such a quick turnaround, the Chinese state has pressed CXMT to speed up the manufacturing scale significantly. The company is on pace to reach an output of 300,000 twelve inch wafers per month, which is fourteen percent of total output volume worldwide, ranking it nearly equal with US based Micron Technology.
Despite the massive gains in overall output volume, South Korea's dominant market share persists for high end products. Samsung and SK Hynix have already mass producing the sixth generation HBM4 memory. Since the Korean companies already made the leap to this developmental stage, the effective market ready technical lead, on average, for top tier products still appears to be approximately three generations ahead of China’s efforts.
Further driving South Korea's memory sales is the latest announcement by Nvidia’s CEO Jensen Huang at the GTC Taipei conference. Nvidia announced its first artificially intelligent processor for personal computers, dubbed N1X. The chips were previewed alongside a brand new line of artificial intelligent personal computers known as Nvidia RTX Spark. The products were made in partnership with Microsoft.
These Nvidia N1X chips are expected to utilize high power, low voltage LPDDR5X memory chips made by Samsung and SK Hynix, demonstrating that South Korean companies will continue to supply critical contracts to consumer AI hardware despite volume advantages held by Chinese counterparts.
