When a company grows by 35 percent year over year, exceeds consensus revenue estimates, reports gross margins above 62 percent, and still trades at a valuation discount to its own industry peers, markets experience a specific type of cognitive dissonance. Taiwan Semiconductor Manufacturing is currently in this predicament, which has subtly reopened one of the more intriguing debates in investing: what does it really mean for a technology stock to be cheap?
In addition to exceeding Wall Street projections of approximately $35.4 billion, TSMC’s preliminary first-quarter 2026 revenue of $35.71 billion demonstrated that the AI-driven demand driving orders through its factories in Hsinchu and Tainan hasn’t shown any discernible signs of slowing. Sales in March alone totaled $13.07 billion, up more than 45% from the same month last year. These figures don’t represent a business struggling through a challenging cycle. These are the figures of a business that is operating at almost full capacity on the most sophisticated chips that are currently manufactured, seeing order books fill more quickly than it can expand its facilities.
The context surrounding the valuation discussion is what makes it truly fascinating, as opposed to just another analyst note recycled as market commentary. At about 25 times, TSMC’s forward price-to-earnings ratio is lower than the median of roughly 29 times for the semiconductor industry. Its trailing P/E ratio of about 34 times is still slightly lower than the average for the entire tech sector, which is 35. Due to its own rapid earnings growth, Nvidia, whose chips TSMC produces at the 3-nanometer node and below, trades at a forward P/E of about 23 times. The picture has truly become confusing for investors who want to use valuation as a meaningful signal instead of just a number on a screen. In the global AI supply chain, the most significant company trades like a reasonably priced industrial enterprise.
It’s difficult to ignore the fact that this discrepancy between market pricing and operational reality has existed for years rather than just a few months. In September 2025, analysts on Seeking Alpha made the same claim, citing TSMC as a stock that was trading at a discount to its earnings potential. In 2022 and 2023, Morningstar posed the same question, citing geopolitical risk—the constant threat of cross-strait tensions—as the discount factor that markets just won’t stop using. That discount isn’t irrational; it’s real.
A type of risk that is absent from a typical discounted cash flow model is carried by a business whose most important fabrication facilities are located on an island that China regards as its own sovereign territory. The counterargument, however, is worth considering: TSMC has been operating in that environment for decades, has continued to grow its capabilities through it, and is currently constructing factories in Arizona and Japan because its clients, including Apple, Nvidia, AMD, and Qualcomm, require some of that geopolitical exposure to be diversified away.
The question of how investors are pricing the AI wave in general is at the heart of the larger discussion that TSMC’s rise has reignited. There are concerns about whether the initial wave of market enthusiasm overshooted fundamentals because several technology stocks that benefited from the initial AI enthusiasm have retreated or stalled. TSMC, on the other hand, sits downstream of everything; whether Nvidia, AMD, or a dozen smaller chip designers win the market for AI accelerators, it doesn’t matter because TSMC produces the chips for the majority of them. The valuation argument is so persistent and, depending on your risk tolerance, persuasive because of this structural position.
Although it tells a different kind of story, Alphabet provides an intriguing parallel. It appears cheap in comparison to the AI excitement surrounding smaller names, with a P/E ratio of about 28 times. By the end of 2025, its Gemini chatbot had grown by about 67% in just six months, reaching 750 million monthly active users. The market appears to still be pricing in a new revenue dimension that was introduced by a multiyear agreement with Apple to power a Siri update. After clocking an algorithm 13,000 times faster than a supercomputer on its Willow chip last year, Alphabet is also making significant progress toward developing a large-scale quantum computer. All of that doesn’t seem like a business that ought to be valued lower than the S&P 500 average, yet here we are. Investors seem unsure of Alphabet’s ability to resist OpenAI and Anthropic’s intrusion into its primary search business, which could be stifling a multiple that would otherwise be significantly higher.
All of this results in an investment environment where the conventional shorthand—high P/E means expensive, low P/E means cheap—has truly lost its credibility as a stand-alone assessment. By traditional measures, TSMC appears inexpensive and is expanding at a rate of 35%. Due to growth that has since faltered, some smaller AI software companies were trading at 80 or 90 times forward earnings last year. In many instances, the narrative-driven premium given to AI application software has lagged behind the companies actually producing the physical infrastructure of the AI era, such as chips, sophisticated packaging, and the equipment needed to make them. It’s still unclear if and which way that relationship will correct.
The 14 out of 18 analysts who gave TSMC a Strong Buy rating and a mean price target of $419.38 indicate that Wall Street thinks the discount is too steep. The bull case is based on a fairly straightforward premise: TSMC’s 2026 forecast of 30% revenue growth will probably prove conservative if AI spending keeps up and hyperscalers and chip designers keep increasing their orders. It’s possible that the market is just taking longer than normal to accept the sustainability of a positive development. This wouldn’t be the first time.
