ON Semiconductor is positioned to benefit as AI spending shifts from building data centers to running inference, the ongoing operational cost that will eventually dominate AI spending.

What happened: ON Semiconductor, a power semiconductor company, has data center revenue that grew 30% in the first quarter and reached $250 million（約400億円） on $6 billion（約9600億円） in total 2025 sales. The company supplies power technology for both new data centers with Nvidia and diverse environments including hyperscaler data centers, businesses, and edge inference (inference close to the data source, such as in EVs, autonomous driving, smart manufacturing, and healthcare diagnostics).

Why it matters: AI spending is currently concentrated on building infrastructure—data centers and training models—but inference spending (running AI models for real-world use) is set to surpass data center infrastructure spending in a few years. Since inference is power-hungry and requires thermal management, power semiconductor makers like ON Semiconductor stand to benefit as this spending category scales. The company is already on a growth track from its EV and industrial markets, so data center revenue could accelerate the next stage of its expansion.

What to watch: ON Semiconductor's data center revenue, which is fast-growing, will be key to tracking whether the company can capture the shift from infrastructure to inference spending as hyperscalers roll out AI applications.