onsemi’s AI growth rate is not a segment margin

On November 2, 2026, onsemi reported third-quarter revenue growth of 11%. CEO Hassane El-Khoury said AI data-center revenue rose 175% year over year. Results for the quarter ended October 2

The two rates answer different questions. One describes an entire company; the other describes a destination for some of its products. Reading the faster number as an explanation of group profitability skips the most important intermediate step: how much business belongs to that destination, and what economics accompany it?

Two maps of the same business

The company’s second-quarter filing identifies PSG, AMG and ISG as its reportable segments, with segment gross profit used in management’s evaluation. These are not a separate AI-data-center reporting category. Segment note

An end-market map asks where products are used. A segment map asks how a company organizes and evaluates its businesses. These maps can overlap without being interchangeable. A product family may serve several applications, while one application may require several product families. Adding an application estimate to segment revenue can therefore count the same sale twice.

The distinction has practical force for power electronics. onsemi’s September 16 Embedded Power Platform announcement describes intended applications spanning automotive, industrial and AI data centers. Platform announcement This is a product-positioning example, not proof of realized sales or profits in any of those markets.

For analysis, the useful representation is a matrix: product groups in one direction, customer applications in the other. A number placed on one axis does not fill the cells between them. Until that mapping is available, treating an entire power-related business as AI exposure would replace evidence with a label. Conversely, assigning all revenue from an application to a single product organization could omit the other products that application purchases.

Growth needs a starting weight

Consider an explicitly hypothetical company with total sales of 100 units. If a niche starts at two units and doubles, it contributes two additional units. If it starts at twenty and doubles, it contributes twenty. The same growth rate can therefore imply very different effects on the whole. These numbers illustrate arithmetic; they are not estimates of onsemi’s business.

A smaller category can be strategically important before it becomes financially dominant. It may establish customer relationships, demonstrate a new capability or create options for later products. That is a credible reason to watch it. It is not a reason to treat its percentage growth as the growth rate of the consolidated company.

Nor can the rest of a business be assumed static. Faster growth in one application can coexist with recovery, contraction or product exits elsewhere. A bridge from application growth to company growth must include those other movements. Otherwise, a headline theme receives credit for changes it did not necessarily cause.

This distinction does not diminish a new business; it makes the claim more testable. If the hypothesis is that an application will reshape the company, subsequent evidence should show its revenue weight changing, not merely another large percentage growth figure each quarter. The smaller the starting point, the more important it is to show speed and scale together.

Margin adds another layer

Even a correct revenue weight does not establish the profit contribution. Consolidated gross margin depends on the mix of sales and the margins earned on them. It can change because products become more profitable, because higher-margin products become a larger share, or because both occur together. The fastest-growing category need not be the one with the highest margin.

Manufacturing creates a further complication. More output can spread fixed production costs over more units. Yet higher utilization, better yield, pricing changes and a shift toward different products are distinct mechanisms. A common direction in sales and margin is consistent with several of them; it does not identify their relative contributions. These are analytical possibilities, not a claimed decomposition of this quarter.

A supplier might also accept development work or early production costs to enter an attractive market. Conversely, an established product can remain profitable despite slower growth. Judging either solely by its demand label loses the difference between winning a design, ramping production and earning an attractive return on the resources committed.

Operating profit requires another bridge beyond gross profit. Research, commercial support and other operating expenses still need funding. A product can contribute positively at the gross-profit level while requiring investment elsewhere in the organization. This is why a market narrative should not quietly substitute for a complete earnings explanation.

What would change the assessment?

The optimistic interpretation is that rapid adoption in an important application could eventually change the company’s sales mix. A more cautious interpretation is that percentage growth may remain impressive while its absolute contribution is still limited, or that other businesses explain much of the aggregate change. Both are plausible hypotheses, not conclusions established by a growth rate alone.

Useful additional evidence would connect end-market revenue to a consistent base, show the corresponding product mix, and explain the contribution of cost and utilization changes. Repeating those disclosures over time would help distinguish a durable shift from the timing of individual orders. Lack of that detail is an analytical limit, not evidence of misconduct.

The right response is neither to dismiss AI demand nor to make it explain every favorable number. Track adoption as an application story, evaluate profitability through comparable financial categories, and connect the two only where disclosure supports the connection. That produces a more testable business thesis than assigning an entire company the economics of its fastest-growing label. This is business analysis, not an investment recommendation.