NVIDIA's Next-Generation AI Chips Enter Mass Production; TSMC's Capacity to Double in the Next Decade

According to Taiwanese media reports, NVIDIA CEO Jensen Huang revealed at a recent "Trillion-Dollar Banquet" that the company is now in full-scale mass production of its next-generation AI chips, Grace Blackwell and Vera Rubin. The latter includes six globally leading chips, placing a significant demand on TSMC's wafer and CoWoS advanced packaging capacity.

Huang pointed out that to support the rapid expansion of global AI infrastructure, TSMC's capacity is expected to more than double in the next decade. He emphasized that AI development is inseparable from memory support, and the demand for HBM and LPDDR will continue to rise, becoming a key component of the AI ​​ecosystem.

Meanwhile, NVIDIA has evolved from a GPU company into a full-fledged computing platform covering CPUs, network chips, switches, and intelligent data processors. Furthermore, Huang confirmed that NVIDIA is collaborating with MediaTek to develop a low-power, high-performance SoC for high-end AI PCs, further expanding its ecosystem.