The artificial intelligence (AI) infrastructure market is booming. The five largest spenders on data center infrastructure are forecast to invest more than $700 billion this year. That is more than the 2025 gross domestic product (GDP) of all but 24 countries. Meanwhile, there are no signs that this spending will slow down anytime soon.
That said, spending is starting to shift. Nvidia's graphics processing units (GPUs) dominate the AI data landscape, but Advanced Micro Devices recently struck some sizable partnerships for its GPUs for inference use. Meanwhile, Alphabet's popular Tensor Processing Units (TPUs) are seeing tremendous momentum, while Amazon's Trainium chips are gaining traction. This is leading to other hyperscalers looking to develop their own custom AI application-specific integrated circuits (ASICs) with the help of companies like Broadcom and Marvell Technology.
Image source: Getty Images.
Meanwhile, with the rise of agentic AI, demand for high-performance central processing units (CPUs) is also starting to boom. The data center CPU opportunity is so big that Arm Holdings decided to forgo its typical licensing and subscription business model and will develop its own data center CPUs.
All of this bodes very well for one company: Taiwan Semiconductor Manufacturing (TSM +0.62%).
NYSE: TSM
Key Data Points
The foundry leader
TSMC has established a virtual monopoly in the manufacturing of advanced semiconductor chips. It has a strong technological expertise and the ability to make these chips at scale with few defects. This has made it the go-to partner for chip designers. The overall growth of the AI chip market is a major driver for the company, and the fact that hyperscalers are diversifying their suppliers only strengthens TSMC's position. TMSC has already shown strong pricing power, and more companies clamoring to gain capacity only benefits it.
At the same time, the need for high-performance CPUs to handle agentic AI adds further demand for its services. The ratio of GPUs to CPUs in AI data centers is expected to significantly narrow, giving TSMC another growth lever. The rise of autonomous driving, robotaxis, and robots in the future also adds potential opportunities.
With $10,000, you can buy around 26 shares of TSMC. Given that the company looks destined to be an AI winner, no matter what direction the AI infrastructure landscape takes, it looks like a no-brainer buy. If Nvidia remains the AI chip king, TSMC wins. If AI ASICs overtake GPUs, it doesn't matter; TSMC still wins. That's the type of AI stock you want to hold for the long term.
