TSMC posts a record quarter
TSMC reported a 35% year‑on‑year jump in first‑quarter revenue to a new record, with AI chip demand from customers like Apple and Nvidia cited as the main driver. The result reinforces that advanced manufacturing capacity remains the market’s choke point and a key commercial differentiator for chip designers. (cnbc.com)
Taiwan Semiconductor Manufacturing Company just put up a quarter so big that one month did a lot of the heavy lifting: March 2026 revenue hit about New Taiwan dollar 415.2 billion, up 45.2% from a year earlier, and first-quarter revenue reached a record New Taiwan dollar 1.134 trillion. (tsmc.com) That company sits in the middle of the chip industry because it does not mostly design chips itself; it manufactures chips that other companies design. In 2025, it said it served about 465 customers and had logic capacity above 17 million 12-inch-equivalent wafers, which is why names like Apple and Nvidia end up tied to its factories. (tsmc.com) The reason one manufacturer can move the whole market is that the newest chips are built on a handful of extremely advanced production steps that very few companies can run at scale. Taiwan Semiconductor Manufacturing said its 3-nanometer and 5-nanometer technologies made up 58% of wafer revenue in the fourth quarter of 2025, showing how much of its business is concentrated in the most advanced layers of the market. (tsmc.com) Artificial intelligence spending is what turned that concentration into a record quarter. The company’s January 2026 guidance said first-quarter revenue would land between $34.6 billion and $35.8 billion, and the actual result came in at the top end of that range at roughly $35.7 billion. (tsmc.com) The bottleneck is not only making the chip wafer itself. Taiwan Semiconductor Manufacturing told investors in January 2026 that demand for Chip on Wafer on Substrate packaging, the method used to connect many high-performance chips and memory parts into one package, would keep being tight through 2026. (tsmc.com) That packaging step matters because an Nvidia artificial intelligence processor is not one slab of silicon dropped into a server like a toaster into a wall socket. It is a bundle of compute chips and high-bandwidth memory that has to be stitched together with advanced packaging before cloud companies can buy it. (tsmc.com) Taiwan Semiconductor Manufacturing is spending accordingly. In January 2026, it raised planned capital spending for 2026 to between $38 billion and $42 billion, with about 70% aimed at advanced process technologies and another 10% to 20% aimed at specialty technologies that include packaging and other support pieces. (tsmc.com) The geography is shifting too, but slowly. The company’s Arizona subsidiary said its first fab entered high-volume production in the fourth quarter of 2024, while a second fab is planned for 3-nanometer technology and a third fab is planned for 2-nanometer and A16 technology, with total U.S. investment announced at $165 billion. (tsmc.com) So this quarter was not just a sales beat on April 10, 2026. It was a reminder that in chips, the scarce asset is still factory time on the newest lines, and the companies that control that time are the ones deciding how fast the artificial intelligence boom can actually ship. (cnbc.com)
