Power Integrations beat the quarter, but the order book says the cycle is rolling over before AI GaN can matter

The thesis from this print is that Power Integrations delivered a better-than-modeled Q3, but not a cleaner cycle turn. What was priced in was a modest recovery quarter: the Street sat at $111.3 million of revenue and $0.34 of EPS. What actually surprised was the near-term shipment level, with revenue of $118.9 million versus $111.3 million, a +6.9% surprise, and EPS of $0.36 versus $0.34, a +5.9% surprise. That matters, but it is not the whole trade. The variant perception is that the beat came before, not through, the order slowdown that management described, and the company’s own Q4 range of $100 million to $105 million reframes Q3 as a last good quarter in the current inventory and end-market sequence. A portfolio manager should treat the stock less as a clean cyclical recovery and more as a timing problem: near-term revenue is being cut by consumer and appliance weakness, while the AI data-center GaN opportunity has credible technical markers but production release planned for late 2026.

The revenue trajectory makes that timing problem visible because Q3 did not break out of the post-correction band. Revenue was $118.9 million in Q3 FY2025, up +2.6% QoQ and +2.7% YoY, after $115.9 million in Q2 FY2025 and $115.8 million in Q3 FY2024. That is enough to beat a $111.3 million estimate, but not enough to restore the $125.5 million level seen in Q3 FY2023. The sequential recovery also slowed from +9.8% in Q2 FY2025 to +2.6% in Q3 FY2025, while gross margin slipped from 55.2% in Q2 FY2025 to 54.5% in Q3 FY2025. The company’s own reported call basis tells the same story with slightly different presentation: Eric Verity said, “Third quarter revenues were up 3% sequentially to $119 million.” That wording is useful because it commits to growth, but only at a low-single-digit pace, which is not the demand acceleration implied by a +6.9% Street revenue surprise.

The margin line is the second reason not to extrapolate the beat. Gross margin in the quarterly history peaked at 55.2% in Q1 FY2025 and stayed at 55.2% in Q2 FY2025 before falling to 54.5% in Q3 FY2025, and management’s Q4 non-GAAP guide is lower again at 53.5% to 54.0%. Verity attributed Q3 non-GAAP gross margin of 55.1% to mix and cost timing, saying it was “down 70 basis points from the prior quarter, driven by higher input costs flowing through our inventory as well as smaller benefit from the dollar and exchange rate.” The number to focus on is not only the 70 basis points, but the sequence into Q4: even with non-GAAP operating expenses expected around $47 million, down slightly from Q3’s $47.4 million, the guide says the company is losing gross-margin support at the same time revenue is stepping down. That combination is rarely rewarded for long in power discretes unless bookings reaccelerate quickly.

The demand commentary explains why the Q4 reset should carry more weight than the Q3 beat. Jennifer Lloyd disclosed that July orders had slowed, with bookings “down about 20% compared to the monthly run rate of the first half of the year,” and then identified appliances as the pressure point, with orders “down about 40% in Q3 compared to the first half.” Those figures convert what could have been dismissed as channel noise into a specific end-market problem. The Q3 revenue mix was 42% industrial, 34% consumer, 13% computer and 11% communications, so the consumer category is large enough to drive the consolidated guide. Management did not leave that to inference: Lloyd said Q4 revenue should be $100 million to $105 million “with the consumer category driving a large portion of the decrease compared to the third quarter.” That is the core mispricing risk. If investors pay for the +6.9% revenue surprise but underweight a 34% consumer mix facing a 40% appliance-order decline, they are capitalizing stale shipments rather than current demand.

The Q4 guide also changes the full-year narrative because the midpoint implies the business gives back the Q3 upside quickly. Management expects $100 million to $105 million of Q4 revenue, and Verity said that at the midpoint “full year revenue growth would be about 6%.” A business growing about 6% for the year, with Q4 revenue below Q3’s $118.9 million and Q4 non-GAAP gross margin guided to 53.5% to 54.0%, is not being pulled into a broad power recovery yet. The quarterly history makes that sharper: after Q3 FY2025 revenue of $118.9 million, Q4 FY2025 revenue is listed at $103.2 million, down -13.2% QoQ and -1.9% YoY, with gross margin at 52.9%. Even if investors separate reporting bases and do not conflate the company’s call guide with later historical actuals, both point in the same direction: Q3’s upside did not persist into a higher run rate.

The offset is capital return, and it is real enough to support downside but not enough to answer the revenue question. Lloyd said the company generated $30 million in cash from operations in Q3 and is “on track for more than $80 million in free cash flow this year.” Verity gave the mechanics: cash flow from operations was $30 million, CapEx was $6 million, buybacks used $42 million to repurchase 919,000 shares, and dividends returned $11.8 million. Diluted share count was 56.2 million, down about 200,000 from the prior quarter, and Q4 share count is expected to come down by 400,000 to 500,000 shares compared to Q3, bringing it below 56 million. The board also increased the dividend by $0.005 to $0.215 per share effective in Q1 of 2026. That is a meaningful shareholder-return cadence, including nearly $150 million to stockholders this year through buybacks and dividends, but the completed buyback authorization reduces the immediate visibility of repurchase support unless a new authorization follows.

The strategic bull case remains high-voltage GaN, but the print makes clear that this is not a near-term earnings bridge. Lloyd said the company published work at OCP Global Summit showing advantages of its 1,250 and 1,700-volt GaN technologies in 800-volt DC AI data centers, and the specific commercial milestone is early samples by year-end with production release planned for late 2026. That is a credible roadmap for AI power architecture, but the dates matter. Late 2026 production release does not offset Q4 FY2025 revenue guided to $100 million to $105 million, nor does it neutralize appliance orders down about 40% in Q3 compared to the first half. The high-power gate driver business is the nearer proof point, with revenues up more than 30% year-to-date, but management did not provide its revenue base in the data pack, so investors should not assign it enough magnitude to outweigh a 34% consumer mix unless future disclosures quantify that base.

That customer and supplier read-through is unusually narrow because the data pack names no customers of Power Integrations and no suppliers to Power Integrations. The absence itself matters for portfolio work: there is no named customer or supplier to pair-trade from this print using disclosed POWI exposure. The only quantified end-market read-through is category-level, not account-level: 34% consumer mix in Q3, appliance orders down about 40% in Q3 compared to the first half, 42% industrial mix, 13% computer and 11% communications. For downstream electronics demand, the implication is that appliance-exposed power semis are seeing a sharper order correction than the consolidated +2.7% YoY Q3 revenue line suggests. For suppliers into Power Integrations, no named read-through can be made from this pack, and making one would overstate the evidence.

The peer comparison reinforces that the issue is not simply “power discretes are weak.” In the latest reported quarter, VSH printed $839.2 million of revenue, 21.0% gross margin and +17.3% revenue YoY, while DIOD printed $405.5 million of revenue, 31.8% gross margin and +22.1% revenue YoY. Power Integrations’ Q3 FY2025 revenue growth was +2.7% YoY with 54.5% gross margin. The comparison says POWI still owns a structurally richer margin profile than VSH and DIOD, but its top-line recovery is materially slower than their +17.3% and +22.1% revenue YoY growth rates. That is the relative-stock tension: POWI deserves credit for a 54.5% gross margin versus peer margins of 21.0% and 31.8%, but if the sector bid is based on revenue acceleration, this print does not give POWI the same evidence.

The call delivery confirms management was not trying to sell Q3 as a clean inflection. In the tone history, Q3 FY2025 sentiment fell to -0.03 from 0.03 in Q2 FY2025, guidance_tone fell to -0.02 from 0.18, and uncertainty rose to 78.4 from 47.2. QA evasiveness also rose to 61.4 from 48.4. Those changes line up with the substance of the call: the company beat Street numbers, but management had to explain a Q4 revenue range that is well below Q3, a consumer-led decline, and appliance orders down about 40% in Q3 versus the first half. The prepared sentiment score of 0.03 versus 0.02 in Q2 FY2025 shows management did not materially change the scripted optimism, but the lower guidance_tone and higher uncertainty show the forward setup became harder to defend.

The tone history also helps separate what is actionable from what is merely uncomfortable. Higher uncertainty at 78.4 in Q3 FY2025 is close to the 78.9 level in Q1 FY2025, but the current print has a more specific culprit: appliance orders down about 40% in Q3 compared with the first half and Q4 revenue expected at $100 million to $105 million. That specificity is investable because it creates a clean test. If consumer and appliance orders stabilize, POWI’s margin structure can reassert itself; if they do not, the company enters 2026 with a smaller revenue base and a GaN AI story still anchored to late 2026 production release. The call’s AI optimism score of 0.32 in Q3 FY2025 versus 0.25 in Q2 FY2025 shows more AI language, but not enough to dominate the transcript. That is appropriate because early samples by year-end are not the same as revenue conversion.

What to watch next is therefore numerical and near-dated. For Q4, the thesis is confirmed if revenue lands inside or below the $100 million to $105 million range, non-GAAP gross margin remains within 53.5% to 54.0% or below, and management still attributes the decline to consumer after a Q3 mix of 34% consumer and appliance orders down about 40%. It is also confirmed if full-year revenue growth stays around the “about 6%” midpoint framing and if share count only falls by the guided 400,000 to 500,000 shares without a new buyback authorization replacing the completed one. The thesis breaks if Q4 revenue clears $105 million with gross margin above 54.0%, bookings recover from the July level that was down about 20% versus the first-half monthly run rate, and management quantifies high-power gate driver or GaN AI revenue at a scale large enough to matter against $100 million to $105 million of quarterly sales. The key date marker is Q1 of 2026 for the $0.215 per share dividend and year-end for early samples of the rack-level AC to DC GaN product; production release remains planned for late 2026, which means the next quarter must stand on order recovery, not AI optionality.