TAIPEI (Taiwan News) — AI’s growing power demand is making energy efficiency a major challenge for future chip development, TSMC Senior Vice President of Business Development Kevin Zhang (張曉強) said Thursday.
Speaking to Reuters at a conference in Amsterdam, Zhang said customers want stronger performance without using much more electricity. He said energy efficiency is now a top demand for smartphones, mobile devices, Internet of Things devices, and AI data centers.
The shift marks a major change for the chip industry. Companies can no longer depend only on adding more transistors to improve performance.
TSMC is the world’s largest contract chipmaker. It makes AI chips for Nvidia and AMD, as well as custom processors for major cloud companies including Google, Amazon, Meta, and Microsoft.
Zhang said transistor density remains important to TSMC’s plans. However, advanced packaging, chip stacking, and photonics are becoming more important for improving efficiency.
He said TSMC expects chips to use up to 30% less power with its current N2 technology compared to its A14 generation. The A14 generation is expected around 2028 and should also offer more than 20% improved computing performance.
Other chip companies are also looking for new ways to improve performance. Hong Kong Economic Times reported that Chinese tech company Huawei announced this week a “Tau Scaling Law” plan aimed at speeding up data movement inside chips.
Zhang said the idea has existed in the industry for a long time. He described it as mainly depending on closer integration of components, including through 3D stacking.
Huawei’s approach comes as Chinese companies face US-led export controls that block access to advanced extreme ultraviolet lithography machines made by Dutch company ASML.
TSMC is a major buyer of ASML’s extreme ultraviolet systems. However, the company said in April it would delay using ASML’s next-generation EUV technology for several years, highlighting why chipmakers are looking for other ways to make AI chips faster and more energy-efficient.
