Energy use forcing rethink of AI chip design, TSMC says

A senior TSMC executive said on Thursday that surging electricity demands from AI are making energy efficiency rather than computing power the main constraint shaping future computer chip development.

Kevin Zhang, Senior Vice President of Business Development, said customers across smartphones to AI data centres are increasingly prioritising performance gains that do not drive up power use, as operators contend with the cost and availability of electricity.

"The area customers most want improvement in is energy efficiency. This is true across the board, whether you are the edge guy, smartphone, mobile, IoT application, or high-performance AI data center," Zhang told reporters at a conference in Amsterdam.

The shift is part of a broader turning point for the semiconductor industry, where simply packing more transistors onto chips is no longer enough to sustain performance gains for energy-hungry AI workloads.

TSMC, the world’s largest contract chipmaker, makes AI chips for Nvidia and AMD, as well as custom AI processors for major cloud companies including Google, Amazon, Meta and Microsoft.

Zhang said improvements in transistor density remain central to TSMC’s roadmap, but other approaches — such as advanced packaging, chip stacking and photonics — are becoming increasingly important to boost efficiency.

He said TSMC expects its chips to cut power consumption by up to 30% between its current N2 technology and its A14 generation, due around 2028, while delivering more than 20% higher computing performance.

The comments come as rivals also explore alternative ways to keep improving chip performance.

Chinese competitor Huawei unveiled its 'Tau Scaling Law' plan this week to improve performance by speeding up data movement within chips.

“The concept has been around in this industry for long enough,” Zhang said, describing it as largely dependent on integrating components more closely, such as through 3D stacking.

Huawei’s approach reflects constraints facing Chinese firms, which are barred by U.S.-led export controls from accessing extreme ultraviolet (EUV) lithography machines made by Dutch ASML - advanced tools for printing smaller circuits.

TSMC, a major buyer of ASML’s EUV systems, said in April it would delay adoption for several years of the next generation of the technology, highlighting how design features improving energy efficiency are becoming more urgent than smaller circuitry for its coming generation of AI chips.