It's impossible to ignore the massive AI infrastructure deals that have been announced recently. Many of these agreements will cover multiple years of hardware deliveries, with values in the $100 billion or greater range.
While Nvidia has so far been the go-to company for AI computing hardware (and still holds a commanding lead in the space), deals for large quantities of Broadcom's custom AI accelerator chips and AMD's GPUs have begun to become more common. This may make it harder for investors to assess which one of these chip designers will be the biggest success story moving forward.
However, all three of these companies are fabless businesses, which means that while they design chips, they don't manufacture them. Instead, they outsource the majority of their chip production to Taiwan Semiconductor Manufacturing (TSM +6.30%), which is a neutral player in this industry. I think it's one of the best-positioned stocks to benefit from the ongoing increase in spending on AI infrastructure, and investors should scoop up shares of it before 2025 is over in anticipation of what's coming in 2026 and beyond.
Image source: Getty Images.
TSMC is the semiconductor foundry market leader
There aren't a lot of advanced semiconductor factory operators in the world, mainly because TSMC has been so dominant. It's the leading chip foundry in terms of revenue by far. It also has a wide technological lead, and that lead is about to get even wider.
Currently, the most advanced chips anyone in the world makes are 3-nanometer (nm) chips. That figure reflects the sizes of certain features on the chip, and the smaller those features are, the more transistors can fit on it. Smaller features thus result in more powerful and faster chips.
While a foundry capable of producing 3nm chips is impressive, TSMC never rests on its laurels. Its practice is to be developing processes for its next-generation node even before it's done fully integrating the last one. Later this year, TSMC is slated to begin producing 2nm chips for its clients.
NYSE: TSM
Key Data Points
These chips could be a major upgrade for AI companies, as they will consume 25% to 30% less power when configured for the same speed. Given how energy consumption has become a pain point for AI companies, that would be a welcome innovation. It could also drive even more business to TSMC as companies like Nvidia, Broadcom, and AMD design chips around this new manufacturing technology.
TSMC's next process node beyond 2nm, which it has dubbed A16, is slated for launch in 2026. Those chips will use 15% to 20% less power than the 2nm chips. Beyond that, its A14 process node chips are expected to use 25% to 30% less power than the 2nm.
TSMC is always innovating, and the chips that deliver these efficiency improvements will generate more revenue for the company. Furthermore, TSMC has locked in its leadership position, making it about as no-brainer of a stock pick as it gets. Moreover, on a valuation basis, the stock is still quite cheap compared to its peers despite its excellent growth and outlook.
TSMC's stock isn't as expensive as its peers
When compared to AMD, Nvidia, and Broadcom, TSMC's growth rate is now one of the best in the group.
TSM Revenue (Quarterly YoY Growth) data by YCharts.
Yet its stock trades at a significant discount to those chip designers on a forward price-to-earnings basis.
TSM PE Ratio (Forward) data by YCharts.
At 31 times forward earnings, TSMC's stock isn't necessarily cheap, but it's far more reasonable than the other stocks.
I think this combination makes TSMC a much better buy than some of the other AI computing hardware providers. While the other three will battle for chip design supremacy, all three of them will be using TSMC to make their chips. This keeps it positioned nicely to thrive regardless of which designers are in the lead at any given time, and I think it's one of the safer picks in this investing sector.
