June 6 News:

    According to semiconductor equipment manufacturers, the main reason for Nvidia’s AI GPU shortage is that Nvidia and TSMC’s previous forecasts for annual production capacity and orders were quite conservative. 

    Unexpectedly, the demand for HPC chips such as AI GPUs erupted, and the original CoWoS wet process with low production capacity was advanced. Encapsulation equipment is difficult to meet. In addition, the delivery time of some packaging equipment and components is as long as 3 to 6 months, and the new production capacity will be slowly in place at the end of the year at the earliest, which means that the next six months will be out of stock. . 

     Wei Zhejia, president of TSMC, also revealed for the first time that he recently had a conference call with a customer and asked to expand CoWoS orders.