The next great wave of demand for artificial intelligence chips could come not from chatbots, but from humanoid robots caring for ageing populations. That is the prediction of Wendell Huang, chief financial officer of TSMC, the Taiwanese company that manufactures the world’s most advanced semiconductors. As countries grapple with rapidly ageing societies, Huang sees robot carers and autonomous vehicles as major commercial frontiers beyond the current boom in AI data centres.

TSMC is already struggling to keep pace with demand. Huang says the company is expanding as fast as it can across Taiwan, the United States, Japan and Germany, but new fabrication plants take two to three years to build and a further year or two to reach full production. Despite concerns about overinvestment, he rejects the idea that AI is a bubble, describing it as a “multi-year structural megatrend” backed by the financial strength of the world’s biggest cloud and technology companies.

The most advanced chips will continue to be ramped up in Taiwan, Huang says, because research and manufacturing teams need to work in close proximity. Recreating Taiwan’s semiconductor ecosystem in the US will take at least five to ten years, even though TSMC’s Arizona lab has now matched the yield of its mother lab in Taiwan.

Huang is also pointed about Elon Musk’s stated ambition to manufacture chips. “There’s no shortcut in semiconductor manufacturing,” he says, arguing that government subsidies alone cannot guarantee success in the foundry business. TSMC’s advantage, he suggests, rests on technology, execution and nearly four decades of customer trust.

Geopolitics remain unavoidable. TSMC sits at the centre of US-China tensions over technology and Taiwan, but Huang declines to be drawn on the politics, insisting the company builds capacity according to customer demand rather than government instruction. On export controls and reports of chips reaching China through third parties, he says TSMC has robust compliance systems, while acknowledging the limits of tracing products once they leave its facilities.

Presenter: Suranjana Tewari

Picture Courtesy of Taiwan Semiconductor Manufacturing Company, LTD

3:10 The AI chip landscape