It’s no secret that spending on artificial intelligence (AI) continues to exceed expectations, and based on hyperscalers’ spending plans, that trend is likely to continue over the next few years.
However, the nature of AI spending is changing, with spending on inference expected to exceed spending on data center infrastructure within a few years. That’s why stocks like power semiconductor companies ON Semiconductor (NASDAQ:ON) ideally placed to be the primary beneficiary. Here’s why:
Forgot Nvidia in 2009? This unusual signal is flashing again. In 2009, a “double down” signal flashed for a little-known chipmaker called Nvidia. For the first time in years, a company 100 times smaller than Nvidia is flashing the same “full conviction” signal. Continued “
Inferential spending takes over
To understand why, it’s important to distinguish between inference spending and data center infrastructure spending. The two are clearly related and one cannot exist without the other. But the reality is that infrastructure spending is currently skyrocketing as hyperscalers rush to build infrastructure to support future growth. You can think of this spending as your AI capital budget for building data centers and training models.
By comparison, inference costs (running AI models for use in the real world) can be thought of as ongoing operational costs. After the infrastructure is initially built, inference can account for a large portion of the expense, outweighing maintenance and expansion.
However, inference spending is power intensive, requires thermal management, and necessarily scales over time. This is good news for power semiconductor companies, including: Nvidia Partner of ON Semiconductor.
Make no mistake, the company benefits from spending on data center infrastructure and inference, but it’s the latter that really drives long-term profits.
ON Semiconductor is ideally located
The company is best known for its power and sensing chips for electric vehicle (EV) and industrial markets. They are attractive enough in their own right. And I selected the company as a top stock to buy in 2026 on the basis that the EV market will pass an inflection point while the industrial end market will slowly improve. Those things happened and the company is now back on a growth trajectory.
But fast-growing data center revenue (up 30% in the first quarter, equating to $250 million on revenue of $6 billion in 2025) could be the key to the next phase of the company’s multi-year expansion. This is an opportunity driven by Nvidia’s ability to deliver power technology to both new generation data centers and diverse environments such as hyperscalar data centers, business, and edge inference. The latter refers to inferences that are close to the data source, such as EVs, autonomous driving, smart manufacturing, and healthcare diagnostics.
