China's AI compute race is shifting to supernodes, as cloud providers and model developers seek domestic infrastructure capable of handling surging large-model training and inference demand.
Chinese companies are shifting more AI accelerator spending away from Nvidia and toward domestic suppliers, a sign that US-China technology tensions are no longer just reshaping chip exports, but the buildout of China's AI infrastructure itself.
The AI data center boom is reshaping the memory supply chain, giving Samsung Electronics, SK Hynix, and Micron greater pricing power while pushing cost pressure into PCs, smartphones, cars, and other end markets.
Samsung Electronics has started mass production of its PM1763 enterprise SSD, a PCIe 6.0-based drive built for AI infrastructure and slated for Nvidia's next-generation Vera Rubin platform, expanding the company's AI memory strategy beyond HBM into high-performance server storage.
China's falling robotaxi supply chain costs could help push the global autonomous taxi market to US$1 trillion by 2040, with Waymo and Tesla leading worldwide as Baidu, Xpeng and WeRide scale rapidly in China, according to Morgan Stanley.
A Chinese research team has developed a phase-change memristor-based neural dynamical system chip, offering a potential hardware path for real-time brain modeling, brain-computer interfaces, and brain disease diagnosis.
For years, Washington and Beijing focused much of their tech rivalry on silicon. Now the front line may be moving to the AI models themselves.
