ON Semiconductor Corporation ON

Key data

The business

onsemi sells power and sensing chips, and two lines matter. The Power Solutions group is the core: silicon carbide modules, power switches and IGBTs that move electricity around inside electric vehicles, fast chargers, solar arrays and factory equipment. The second is the sensing group, the CMOS image sensors behind the cameras that let a car see the road. Automotive is roughly half the company, industrial most of the rest. The moat is the design-in. A power module gets qualified into a car platform years before the car ships, and once it is in, swapping it means requalifying the whole system, so the supplier tends to hold the socket for the life of the model. What you are holding is a SiC module bolted to an EV traction inverter.

The numbers

The shape of the last three years is a boom and a bust. Read the years first.

Revenue fell from $8.25B in 2023 to $6.00B in 2025, and diluted earnings collapsed further, from $4.89 to $0.29, because a fixed cost base does not shrink as fast as sales and 2025 carried heavy restructuring charges. That is why the trailing multiple is 47 times. The earnings, not the price, are what is depressed.

The quarters show the turn. Q1 2026 was an outright loss, dragged down by a one-time charge, but Q2 2026 put up $1.60B in revenue, up 9.2% from a year ago and 6% sequentially, with diluted earnings of $0.56, up 37%. That is the first genuine growth after two years of shrinking. The other line to watch, inventory, has not cleared. Days of inventory sit near 193, so the chips are selling again while the warehouses stay full.

Free cash flow was $1.42B in fiscal 2025, but that comes from the capex cut, from $1.54B in 2023 to $341M in 2025. A power-chip maker cannot underinvest in fabs forever. Normalize capex toward $700M and free cash flow is nearer $1.0B, and on normalized earnings the multiple is about 25 times, not 47. The market is treating the Q2 turn as the start of a clean recovery. The number that would prove it wrong is gross margin, 37.4% against a peak near 47%. This memo's view is that volume returns before margin does, and the next gross-margin line settles it.

Management

Read the record, not the bios. The buyback is the good news. onsemi spent $1.38B repurchasing stock in 2025 at an average price near $49, well below today's $74, and has shrunk the diluted share count 8% in three years, capital allocation timed with the cycle rather than against it. The pay line is less flattering. CEO Hassane El-Khoury took home $17.7M in 2025, nearly 15% of net income, though that ratio is inflated by earnings falling to almost nothing rather than by a raise. Insiders sold, 14 open-market sales worth $20M over the year and no purchases, with the CFO responsible for about $6.5M. The filings here do not separate scheduled 10b5-1 sales from discretionary ones, so read it as a tilt, not a signal.

How it fails or surprises you

The inventory clears and silicon carbide re-accelerates (right tail). Days of inventory near 193 are still bloated, and gross margin at 37.4% sits ten points below its peak. If automotive restocks into real end demand and the SiC design wins ramp, revenue and margin recover together, and the operating leverage in a fab-heavy model is violent on the way up. At 25 times normalized earnings that recovery is assumed but its full slope is not paid for.

A large automotive customer in-sources or dual-sources. Half the company rides on cars, and the biggest EV makers have both the volume and the motive to bring silicon carbide in-house or split it to a second source. Lose a flagship socket and the design-in position that justifies the multiple weakens quickly.

Margin does not follow volume. The fact the recovery read explains least is that Q2 sales grew while gross margin barely lifted off the floor, and Q1 was a loss. If 37% is the new normal rather than a trough, this is not a cyclical dip waiting to snap back, it is a structurally lower-return business wearing a growth multiple.

Closing thoughts

Gross margin in the next quarter settles it. The evidence points to a business one or two quarters into a cyclical recovery, priced as though the recovery is largely done. The fatter tail is up if margin follows volume, because the earnings base the 25 times is measured against is itself depressed. The left tail is a stalled auto cycle and a stuck 37% margin, which would leave you paying a growth multiple for a no-growth cyclical. If the downside linchpin breaks, the whole normalization case goes, roughly a third of the price. If margin normalizes, mid-cycle earnings return and the multiple looks cheap against them. Weight it, in judgment not arithmetic, modestly toward the upside, contingent entirely on margin.

The bet is still that carmakers keep building electrified cars and keep buying onsemi's silicon carbide, and that onsemi is still the supplier they designed in. What breaks it is a stalled car market or an automaker taking the chip in-house. The one pair of numbers that tells you first is gross margin and inventory days in the next report. If Q3 shows revenue up again but gross margin still near 37% and inventory days still near 190, the recovery is refilling shelves, not returning.