Jensen Huang warns TSMC needs to 'work very hard' to meet AI demand — Nvidia CEO says its demand alone may force doubling its capacity over the next decade

Nvidia CEO Jensen Huang is continuing his tour of Taiwan, and in his latest musings with the press had warned that Taiwan Semiconductor Manufacturing Company (TSMC) has its work cut out to keep up with demand. SCMP reports that Huang said Nvidia's demand for wafers alone could force TSMC to "more than double its capacity in the next decade."

According to the report, Huang made the statement following a Saturday evening banquet of vital supply chain partner executives, including TSMC Chairman and CEO C C Wei, and Foxconn Chairman Young Liu. SCMP says Huang told reporters outside the restaurant that TSMC must "work very hard" to meet its insatiable demand, stating Nvidia "needs a lot of wafers," and that TSMC needs to boost its output this year.

"TSMC is doing an incredible job, and they’re working very, very hard," he reportedly said, adding, "We have a lot of demand this year." Huang's comments won't come as a surprise to the Taiwanese silicon giant. Despite recently admitting the company was "very nervous" about an AI bubble, the company said that it remains assured of demand. Naturally, that means the pain point for AI going forward is likely to be supply. In November, Wei remarked that TSMC's capacity was "Not enough, not enough, still not enough," stating that its advanced-node capacity falls "about three times short" of AI demand.

Huang's comments about TSMC needing to double its capacity in the next decade also serve as context for recent rumblings about plans to move 40% of Taiwan's chipmaking capacity to America. Huang assured the island nation that Taiwan's silicon shield will be preserved, namely, because this 40% is all new capacity. As per its last Annual Report from 2024, TSMC currently boasts an annual capacity of 17 million 12-inch equivalent wafers in 2024, a number that has likely increased with production added in 2025.

In May, the company revealed it planned to spend $42 billion on expansion in 2025 alone, and is also trying to expedite production across new sites, such as its Fab 21 phase 2 site in Arizona, where equipment will reportedly start arriving next summer, with mass production targeted for 2027, ahead of the original 2028 schedule.

Back in Taiwan, Huang also reportedly discussed the importance of memory chips and also touched on Nvidia's deal with OpenAI, resetting expectations a little by telling reporters the $100 billion deal was "never a commitment," but rather an invitation to invest.

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.