More new logic fabs needed to fill the looming supply shortage in AI chips by 2030: McKinsey

Memory chips are also likely to face supply woes.

Three to nine new logic semiconductor fabrication plants are needed by 2030 to meet the additional demand for advanced chips powering artificial intelligence (AI), according to McKinsey & Company.

An analysis by McKinsey showed that the generative AI boom globally will require an additional 1.2 million to 3.6 million wafers produced using technology nodes equal to or less than three nanometers (nm) by 2030.

This will add to the existing demand of 15 million wafers by 2030 for non-gen AI applications. It said nearly half or seven million will be produced using more than 3nm, and the remaining eight million wafers will be produced using nodes equal to or less than 3nm.

With the current logic fab pipeline, McKinsey said only 15 million wafers using tech nodes of less than 3nm can be supplied by 2030, leaving a shortfall of one to four million wafers.

“To close the gap, three to nine new logic fabs will be needed by 2030,” it said.

READ MORE: Manufacturing is at a turning point as AI shapes Industrial 4.0

In terms of memory chips, it said the projected chip demand will largely be dependent on how big will be the standard memory capacity for future AI applications. 

McKinsey estimated that DRAM requirements from gen AI applications will require five to 13 million additional wafers under a “DRAM light” scenario, where AI accelerators remain memory light compared to GPU-based systems.

In a “DRAM base” case, in which AI systems catch up to GPU-based systems in terms of memory, McKinsey said DRAM demand could go as high as seven to 21 million wafers. That scenario would require the need for an additional six to 18 new memory fabs.

“The surge in demand for gen AI applications is propelling a corresponding need for computational power, driving both software innovation and substantial investment in data center infrastructure and semiconductor fabs. The critical question for industry leaders is whether the semiconductor sector will be able to meet the demand,” McKinsey said.

“Investment in semiconductor manufacturing capacity and servers is costly and takes time, so careful evaluation of the landscape is essential to navigating the complexities of the gen AI revolution,” it added.