NVDA and TSMC seen cheap

Investors pitching the artificial-intelligence chip buildout are still circling two names: Nvidia for the processors and Taiwan Semiconductor Manufacturing for the factories that make them. (investor.nvidia.com) (investor.tsmc.com) The shorthand behind that trade is earnings power a year or two out, not last quarter’s headline multiple. In the market talk cited this week, Nvidia was framed at about 18 times estimated 2027 earnings per share, and TSMC’s Taipei-listed shares at about 15.5 times. (x.com) That argument lands after both companies posted another round of fast growth. Nvidia said on February 25, 2026 that fiscal 2026 revenue rose 65% to $215.9 billion, with data-center revenue reaching $62.3 billion in the January quarter alone. (investor.nvidia.com) TSMC said in its April 2026 first-quarter results that revenue reached $35.90 billion, gross margin was 66.2%, and second-quarter revenue was guided to $39.0 billion to $40.2 billion. The company said high-performance computing, the bucket that includes artificial-intelligence chips, made up 59% of first-quarter revenue. (investor.tsmc.com) The basic bet is simple: Nvidia designs the most in-demand artificial-intelligence accelerators, and TSMC manufactures the leading-edge chips and advanced packaging those systems need. If cloud providers and model builders keep spending, both companies sit near the center of that order flow. (investor.nvidia.com) (investor.tsmc.com) The harder question in 2026 is no longer whether companies want artificial-intelligence capacity. It is whether the customers buying tens of billions of dollars of chips can turn that spending into software sales, cloud revenue, or cost savings fast enough to justify the next wave of orders. (investor.nvidia.com) (investor.tsmc.com) (x.com) That is why investors lean on forward earnings estimates. Nvidia closed at $201.68 on April 17, 2026, while TSMC’s New York-listed shares closed at $370.50, prices that leave the debate centered on how much profit each company can produce by 2027 rather than what they earned over the last 12 months. (finance.yahoo.com 1) (finance.yahoo.com 2) There is also a structural difference between the two. Nvidia captures the economics of the finished computing platform, while TSMC sells manufacturing capacity across many customers, which can make TSMC look cheaper on earnings multiples even when demand is strong. (investor.nvidia.com) (investor.tsmc.com) TSMC’s 2025 annual report said robust artificial-intelligence demand ran through the year, and the company expanded capacity to more than 17 million 12-inch-equivalent wafers. Nvidia, for its part, has kept posting record data-center sales as customers build larger clusters of graphics processing units, or GPUs, for training and running models. (investor.tsmc.com) (investor.nvidia.com) So the “cheap” case is not really about low absolute prices. It is a claim that, if artificial-intelligence demand keeps converting into real earnings, the two companies at the core of the hardware stack still do not look expensive on 2027 numbers. (x.com)