How many years will the hunger for computing power in the AI era last? The answer to this question might be glimpsed from the pace of global chip capacity expansion.

Recently, an important signal from the industry: major global semiconductor manufacturers announced new wafer fab projects, mainly focused on expanding high-performance storage chips—especially DRAM capacity. This will become the largest chip manufacturing facility in the region.

The problem is that the production start date has been pushed to 2030.

What does this mean? Currently, the demand for storage and computing chips for AI model training and data center expansion is extremely tight, but new capacity won't be available for another 5 years. How long this supply-demand gap will last and how the market will respond have become core issues that the industry chain upstream and downstream are contemplating. The lengthening of the chip cycle may redefine the landscape of hardware costs and the economics of computing power in the coming years.