While AI's explosive growth hasn't been good news for consumers, because it's directly caused the global memory crisis, but it sure does keep manufacturers exceedingly busy. Case in point, TSMC has revealed profits that both blow past its original projections and smash earlier records.
TSMC had previously issued guidance to its investors predicting that it would make somewhere between $34.6 and $35.8 billion. That's no small potatoes by any means, but between the start of the year and the end of March, the company did in fact make $35.90 billion in net revenue.
During a recent earnings call, chief executive C.C. Wei described AI-related demand as "extremely robust." TSMC also said that 3-nanometre wafers (which are becoming more frequently used for AI chips thanks to Nvidia's imminent Rubin generation) accounted for "25% of total wafer revenue." The bulk of revenue still came from wafers made on earlier processes, though, with 5 nm wafers, for instance, accounting for 36% of total wafer revenue.
It's important to remember that not all of TSMC's revenue comes from fabricating silicon for AI chips. However, it's not hard to see why the company spent $165 billion alone on its plans for a 'gigafab' cluster in Phoenix, Arizona. This US-based chip fab project is already set up to make four-nanometre wafers, but 3 nm is planned to get underway next year.
'Extremely robust' almost feels like an understatement—TSMC is reportedly booked up until 2028, and even the next-gen Arizona fabs have had their projected capacity fully booked before they've even finished construction. In other words, don't expect the supply crisis or, by extension, consumer pricing to ease any time soon.









English (US) ·