Cohu (COHU), which makes semiconductor test and inspection equipment, posted Q2 revenue of $149 million, up 38% from the same period a year earlier (about $108 million) and a step up from the prior quarter's $125 million. The figure is only modestly above the $144 million consensus, but the wide year-over-year growth gap makes it easy to read as a sign of a cyclical recovery.

On the earnings side, adjusted (non-GAAP) net income came in at $14.1 million, or $0.26 per share, well ahead of the $0.14 estimate. Given that Q1 adjusted EPS was just $0.01, the speed of the quarter-to-quarter profitability recovery stands out. Gross margin was reported at 45.4% on a GAAP basis and 45.5% on an adjusted basis.

Management cited broad improvement across the end market and said estimated test cell utilization rose sequentially to roughly 80% by the end of June. The pipeline for AI-driven computing opportunities was lifted to about $850 million annually, with management highlighting adoption of the Eclipse test handler for high-power processors used in data centers as a growth driver.

Even though the adjusted figures showed a wider profit, on a GAAP basis the company was essentially at break-even with a Q2 net loss of about $0.2 million, or $0.00 per share. When comparing directly with the $0.14 adjusted consensus, the adjusted figures are the right reference; looking only at GAAP can give a different feel.

Stock-based compensation, acquired intangible amortization, restructuring charges, and similar items are excluded from the adjusted figures, so the gap between the two measures can persist structurally. Beginner investors should be cautious about treating "$0.26 per share" as the same as cash earnings.

Given the business mix, the semiconductor capex cycle and a concentrated customer base can amplify earnings swings. The midpoint of Q3 revenue guidance is higher than Q2, but if shipments or revenue recognition slip, or if a key customer's orders wobble, results could tilt toward the lower end of guidance (around $163 million).

Management's tone leans more toward confidence than defensiveness. President and CEO Luis Müller pointed to a recovery across the end market and improving test cell utilization, citing not just the headline numbers but also utilization as a real-time read on the field. He put AI and data center customer momentum front and center, raised the annual high-performance computing revenue estimate, and tied differentiated product adoption — the Eclipse handler and T-Core active thermal control — to specific drivers.

What the market welcomed was less the earnings beat itself and more the interpretation that the beat is tied to the AI equipment demand and rising utilization story rather than a one-off bounce. That said, the next-quarter guidance the company disclosed is centered on revenue, and per-share guidance is not in this release. Investors should keep the company's outlook and sell-side EPS estimates separate rather than mixing them.

The after-hours surge looks less like a reaction to a slightly beat quarter and more like the combination of simultaneous revenue and adjusted earnings surprises, a step-up in next-quarter revenue expectations, and concrete numbers (a higher annual revenue estimate and a larger opportunity pipeline) reinforcing the AI and high-performance computing story. With utilization at 80% thrown in as a real-time field metric, expectations that operating leverage can accelerate after a cyclical bottom appear to have been priced in.

Still, the stock had already rallied in the regular session on the report day before reacting again after hours, so at the next open both profit-taking and "guidance already priced in" debate could surface at the same time. Beginner investors should treat after-hours volatility as confirmed gains; better to track whether next-quarter revenue actually lands on track and whether AI-linked revenue becomes more visible in the numbers.