Power Integrations: the EPS beat was expense timing, but the actionable signal is GaN mix into a Q1 revenue reset

The right way to trade this print is to separate what was already in the stock from what actually changed. What was priced in was a soft Q4 demand quarter: Street revenue was $103.1 million and Power Integrations reported $103.2 million, a +0.1% surprise that does not justify a broad estimate reset by itself. What surprised was below the revenue line: EPS of $0.23 beat the $0.19 estimate by +21.1%, and the call makes clear the mechanism was not hidden operating leverage from a demand inflection but spending and tax. CFO Nancy Erba said non-GAAP operating expenses of $45 million were “well below the outlook of $47 million,” and non-GAAP net income included “a benefit of about $0.02 from the lower-than-expected tax rate.” That distinction matters because a +21.1% EPS surprise on a +0.1% revenue surprise is not the same as a bookings-led upcycle; it is an earnings bridge that gives management time for the higher-quality mix story to show up in revenue.

That bridge is credible because the revenue trajectory looks worse in the quarter than in the sequencing into Q1. Q4 revenue of $103.2 million was down -13.2% QoQ and -1.9% YoY, taking the company back below the $105.5 million in Q1 FY2025, the $115.9 million in Q2 FY2025, and the $118.9 million in Q3 FY2025. On the surface that is exactly the type of print a PM should fade in power discretes. The counterpoint is that the next quarter already in the history table shows $108.3 million in Q1 FY2026 revenue, +4.9% QoQ and +2.6% YoY, and the call guide was $104 million to $109 million. The company did not guide a snapback to the $118.9 million Q3 level, so the bullish case cannot be “demand is back.” The more defensible thesis is narrower: the trough-like Q4 revenue base and the Q1 guide leave room for the market to underwrite a modest recovery while the mix indicators, especially industrial, GaN, and high-power, carry more of the valuation burden than raw unit growth.

The margin path reinforces that this is a mix debate rather than a volume debate. Reported gross margin fell from 54.5% in Q3 FY2025 to 52.9% in Q4 FY2025, and Q1 FY2026 came in at 52.6%, so there is no reported gross-margin breakout in the quarterly history. But on the company’s non-GAAP basis, Erba gave a different lens: “Non-GAAP gross margin was 55.1%, up 70 basis points from the prior year, mainly driven by higher industrial revenues as a percentage of our mix with some additional benefit from higher back-end manufacturing volumes.” The useful part of that quote is the attribution. Industrial was 37% of Q4 revenue mix, ahead of consumer at 34%, communications at 15%, and computer at 14%, and management tied the higher non-GAAP margin to industrial share rather than to broad end-market volume. If Q1 revenue is only $104 million to $109 million and non-GAAP gross margin is only 53% to 54%, the stock should not be valued on a near-term margin surge. It should be valued on whether the 37% industrial mix can become more durable as PowiGaN and high-power design wins convert.

The market may be missing that the design-win data is more specific than the revenue line. Full-year revenue was up 6%, non-GAAP EPS grew 8%, and cash flow from operations was $112 million, up $30 million from the prior year, but those figures alone do not resolve the end-market cycle. The more investable data points are that design win value grew 10% in 2025, PowiGaN product revenue grew more than 40% for the year, and revenue outside of cell phone applications has averaged 12% growth over the past 2 years. That combination says the company is not merely waiting for cell phones or consumer adapters to recover. It is redirecting the mix toward applications where voltage, efficiency, and integration matter more than commodity pricing. The risk is that design win value is not revenue, and the Q4 revenue decline of -13.2% QoQ proves conversion is not yet visible at the consolidated level. But that is precisely the variant perception: the reported quarter gives little demand gratification, while the design-win mix offers a path for better revenue quality before the headline revenue growth rate looks clean.

The second-order implication for customers is concentrated around architectures rather than named order books, because the supply-chain data lists no customers of POWI and no suppliers to POWI. The one named customer in the call excerpts is NVIDIA, and management placed the engagement in data center auxiliary power for a “next-gen 800-volt DC architecture,” where Power Integrations’ “1,700-volt GaN solutions” are positioned against silicon carbide. Those numbers are the magnitude: 800-volt DC at the architecture level and 1,700-volt GaN at the device level. The read-through for NVIDIA is not that Power Integrations has disclosed a revenue ramp, because it has not in the provided data; the read-through is that auxiliary power is becoming an attach point in high-voltage data center power distribution rather than a legacy low-voltage support function. The read-through for unnamed meter customers in India is similar: the company is seeing migration to 900 and 1,250-volt GaN products because voltage swings on India’s grid require added protection. Since the supply-chain section has no named suppliers, there is no defensible supplier beneficiary to call out from this data pack, and inventing one would be the wrong conclusion. The customer implication is still meaningful: 900, 1,250, and 1,700-volt GaN references move the debate from consumer-cycle recovery to high-voltage content expansion.

That content argument also changes how to compare Power Integrations with peers. The peers table shows broader power-discrete revenue growth that looks healthier than POWI’s Q4 -1.9% YoY, including DIOD at +22.1% on $405.5 million revenue and VSH at +17.3% on $839.2 million revenue. On raw growth, POWI did not lead. But POWI’s gross margin of 52.9% in Q4 FY2025 sits far above DIOD’s 31.8% and VSH’s 21.0%, and also above ROHCY’s 26.7%, 6504.T’s 31.0%, and 6503.T’s 32.2%. The comparative point is therefore not that POWI is outgrowing the power-discrete group; it is that POWI’s revenue base is currently under-earning on growth while retaining a margin structure peers do not show in the provided data. If the 10% growth in design win value and more than 40% PowiGaN revenue growth translate into even modest top-line growth, the incremental earnings sensitivity can matter more than at lower-margin peers. If they do not, the premium margin becomes less useful because reported revenue has already slipped from $118.9 million in Q3 FY2025 to $103.2 million in Q4 FY2025.

The call delivery supports that interpretation, but not without conflict. The tone history improved sharply on the surface: sentiment rose by +0.15 to 0.21, guidance_tone rose by +0.33 to 0.01, tone_confidence rose by +0.08 to 0.32, and prepared_sentiment rose by +0.33 to 0.18. Those are not euphoric absolute readings, but they mark a recovery from Q3 FY2025 when sentiment was -0.03 and guidance_tone was -0.02. The conflict is that uncertainty increased by +10.0 to 48.7, and QA sentiment slipped by -0.04 to 0.26 even as qa_evasiveness dropped by -44.5 to 66.9. That mix says management sounded more constructive and less evasive than the prior call, but it did not eliminate forecast ambiguity. In practical terms, the call tone confirms the thesis only to the extent that it matches the numbers: management is more confident about the forward setup, but the Q1 revenue range of $104 million to $109 million is still only a recovery from the Q4 trough, not a return to the $118.9 million Q3 FY2025 revenue level.

The spending guide is where near-term EPS expectations need discipline. Q4 non-GAAP operating expenses of $45 million were below the $47 million outlook, but Q1 non-GAAP operating expenses are guided to about $46 million, plus or minus $0.5 million. That means the Q4 EPS beat should not be extrapolated one-for-one, especially because the company disclosed a Q4 tax-rate benefit of about $0.02 and Q1 GAAP results will include a restructuring charge of between $3.5 million and $4 million. The full-year cost framework also carries a remaining item in the $3 million to $5 million range. For PMs, the investment question is whether Q1 revenue toward $109 million and non-GAAP gross margin toward 54% can absorb a $46 million operating-expense base. If revenue lands near $104 million and non-GAAP gross margin sits near 53%, the Q4 EPS beat will look like temporary cost and tax help rather than a base from which estimates rise.

Cash generation gives the company room to wait for conversion, which is important because the revenue proof is not yet decisive. Q4 cash flow from operations was $26 million and CapEx was $7 million, while 2025 cash flow from operations was $112 million and CapEx was $24 million, producing free cash flow of $87 million. The company returned $145 million through buybacks and dividends, or 167% of free cash flow, so capital return exceeded internally generated free cash flow for the year. That is not automatically negative, but it limits the tolerance for a prolonged gap between design wins and revenue. Inventory also rose by $2 million in Q4 while days of inventory on hand reached 313 because of the lower revenue number. The 313 days figure is the clearest balance-sheet warning in the print: if Q1 revenue of $108.3 million and the guide range of $104 million to $109 million are not followed by a stronger Q2 revenue trajectory, inventory will argue against a clean demand recovery. Cash flow lets management continue investing, but inventory tells investors not to declare the upcycle early.

The thesis therefore lands between the obvious bear case and the easy bull case. The bear case is that Q4 was a miss in disguise because revenue was effectively in line, gross margin fell to 52.9%, and EPS upside depended on $45 million OpEx versus a $47 million outlook plus a tax benefit of about $0.02. That is directionally right for the quarter. The bull case that needs more precision is that Power Integrations has a differentiated GaN pipeline that is beginning to matter in industrial, meters, and data center auxiliary power. The data supports that view through 10% design-win value growth, more than 40% PowiGaN revenue growth, 37% industrial revenue mix, and revenue outside cell phone applications averaging 12% growth over the past 2 years. What the market may misprice is the timing mismatch: the Q4 P&L still looks cyclical, but the product mix is already shifting toward higher-voltage GaN sockets that should not be valued like a generic consumer adapter recovery.

What to watch next is concrete. The thesis is confirmed if the next reported quarter holds revenue within or above the $104 million to $109 million guide, with the already reported Q1 FY2026 level of $108.3 million and +4.9% QoQ growth as the benchmark, while non-GAAP gross margin stays inside the 53% to 54% guide and operating expenses remain around $46 million, plus or minus $0.5 million. It is further confirmed if management repeats or raises the 2025 markers of 10% design win value growth, more than 40% PowiGaN product revenue growth, and 12% average growth outside cell phone applications over the past 2 years, because those are the numbers that turn a low-growth quarter into a mix-change story. It breaks if revenue falls back toward the Q4 FY2025 $103.2 million level, if gross margin remains closer to the reported 52.6% in Q1 FY2026 than the guided non-GAAP 53% to 54% range, or if inventories fail to improve from 313 days after revenue has already recovered from the Q4 trough. The next call should also show whether the tone improvement was real: guidance_tone at 0.01 and tone_confidence at 0.32 are low bars, and a durable turn should improve those without another rise in uncertainty from 48.7.