Artificial intelligence spending has continued to run ahead of expectations, and industry forecasts suggest that the buildout phase will eventually give way to a larger, more persistent wave of “inference” demand. That shift matters for investors in power semiconductors such as ON Semiconductor, whose business spans both data center infrastructure and the electricity-intensive workloads that increasingly power AI at the edge and in the cloud.
While the market has largely tracked data center capex over the past year, the long-run profit outlook may hinge on inference—an operating-cost category that is expected to scale as AI systems are deployed in real-world applications, from hyperscaler data centers to vehicles, industrial automation, and healthcare.
Key takeaways
- Price move: The article does not provide a specific market move for ON Semiconductor shares.
- Catalyst: The central driver is the expected evolution of AI spend from data center buildout to inference workloads.
- Why it matters: Inference is described as more power intensive and likely to account for a larger share of spending after early infrastructure is installed.
- Implication for ON Semiconductor: The company is positioned to benefit from scaling power and thermal-management needs tied to inference, alongside continued data center revenue growth.
Why AI spending is shifting from buildout to inference
The article draws a distinction between two phases of AI investment. Data center infrastructure spending refers to the capital required to build out the systems needed for model training and future capacity expansion. Inference spending, by contrast, represents the ongoing cost of running AI models once they are deployed for real-world use.
According to the article, infrastructure spending is currently the focus as hyperscalers expand capacity to support anticipated growth. But after early buildouts are completed, inference is expected to become the dominant spending category—above maintenance and incremental capacity additions. The key investment takeaway is that inference may scale for a longer period, creating sustained demand for power electronics and thermal solutions.
The article argues that inference workloads are inherently power hungry and require robust thermal management. As AI usage grows, that combination could increase the addressable market for power semiconductor suppliers, including ON Semiconductor.
How ON Semiconductor’s revenue mix ties to the AI cycle
ON Semiconductor is described as supplying power and sensing chips used across electric vehicles and industrial markets, but the article emphasizes a growing data center component. It cites data center revenue growth of up 30% in the first quarter and states that data center revenue was equivalent to $250 million on $6 billion in sales in 2025—figures presented in the article as part of the company’s momentum.
The article frames ON Semiconductor’s opportunity as driven by its ability to provide power technology across multiple environments tied to AI. It highlights support for the next generation of data centers as well as “edge inference,” where AI compute happens closer to the source of data. Edge use cases mentioned include electric vehicles and autonomous driving, smart manufacturing, and healthcare diagnostics.
In this view, ON Semiconductor benefits from both phases—data center infrastructure demand and inference demand—but the article’s emphasis is that inference should be the longer-duration driver of profits.
Company outlook and what investors will watch
The article points to guidance from ON Semiconductor’s CEO, Hassane El-Khoury, stating that the company expects its data center growth to “double year over year in 2026,” which the article translates into about $500 million in revenue for 2026. It also notes that Wall Street expectations for overall 2026 revenue are $6.47 billion—with the implication being that the data center segment could become a more visible contributor to results over time.
Beyond near-term demand from EVs and industrial automation, the article suggests that ON Semiconductor’s rapidly growing data center business could accelerate the next stage of its multiyear expansion. The argument hinges on the expectation that inference demand continues to rise as AI models move from experimentation into broader deployment—raising questions investors may want answered in upcoming updates, including how quickly inference workloads convert into measurable revenue.
Bigger picture: the market’s next question
AI infrastructure buildouts and inference deployment are linked, but they carry different investment profiles. The article’s central thesis is that as the industry transitions from building capacity to running models at scale, power semiconductor demand may become more structurally tied to AI consumption rather than only to capital spending cycles.
For investors, the immediate focus will likely be on ON Semiconductor’s ability to sustain data center growth while demonstrating that inference-driven demand is expanding. Monitoring management commentary around segment performance, customer qualification cycles for next-generation platforms, and the pace of inference-related deployments could help clarify how much of the AI spending shift ultimately flows through to the company’s financials.
Next, investors will likely look for fresh quarterly results, management commentary on data center order patterns and customer engagement, and any updates to guidance for 2026. Broader market sensitivity to AI infrastructure spending and the macro backdrop for interest rates and capital budgets may also influence how quickly investors reprice the relative importance of inference versus data center buildout.
