TSMC Bets On AI-Powered Design To Make Next-Gen Nvidia And AI Chips Far More Energy Efficient

Taiwan Semiconductor Manufacturing Company (TSMC), the world’s leading contract chipmaker, has revealed a bold new approach to making artificial intelligence hardware more energy efficient: letting AI design the chips themselves.

At its Silicon Valley technology conference, TSMC showcased a strategy aimed at improving the energy efficiency of advanced AI chips by as much as tenfold.

The company, which produces processors for NVIDIA and other global leaders, demonstrated how AI-driven software tools are finding smarter, faster solutions than human engineers.

Nvidia’s high-performance AI servers today can draw up to 1,200 watts under heavy workloads — a demand so large that, if scaled, it would equal the continuous electricity usage of 1,000 US households.

Chiplets and AI-Driven Design

The efficiency push centres on packaging multiple “chiplets” — smaller components made with different technologies — into a single unit.

This modular approach promises greater performance but also requires sophisticated design techniques.

To meet that challenge, chip designers are turning to AI-powered software developed by companies such as Cadence Design Systems and Synopsys, which unveiled their latest tools at the event.

These platforms, built in close collaboration with TSMC, outperformed TSMC’s own engineers in certain chip design tasks.

“That helps to max out TSMC technology’s capability, and we find this is very useful,” Jim Chang, deputy director at TSMC for its 3DIC Methodology Group, said during a presentation describing the findings, Reuters reported.

Overcoming Physical Limits

Chipmakers are also running into the physical barriers of existing manufacturing techniques.

Moving data across chips using traditional electrical connections is reaching its limits.

Engineers are now experimenting with optical connections to transfer data at scale, though these systems must be proven reliable for use in vast data centres.

The Bigger Picture

As AI adoption accelerates worldwide, the industry faces mounting pressure to curb the enormous energy demands of training and running models.

TSMC’s embrace of AI-assisted design marks a shift toward using the very technology that consumes so much power to help solve the problem.

If successful, this could shape the future of both chip design and sustainable AI infrastructure.

Please click here to add Swarajya as your preferred and trusted news source on Google.

Also Read: China Opens Anti-Dumping Investigation Into Chip Trade With USA, Causing Rally In Domestic Semiconductor Stocks