CEVA’s beat was only 0.9%, but the mispricing is in the royalty mix, not the revenue line

The actionable read on this event is that CEVA’s Q4 was not a demand-inflection print in the usual semiconductor sense, because the street-comparison basis left little room for celebration: EPS was actual $0.18 versus estimate $0.18, a 0.0% surprise, and revenue was actual $31.3 million versus estimate $31.0 million, a +0.9% surprise. That was what was priced in: a record-ish quarter, a licensing-heavy model with high gross margin, and a normal post-holiday sequential decline into Q1. The actual surprise was more subtle and more investable. The company converted a quarter with only +7.1% revenue YoY and +10.2% revenue QoQ into a path where non-GAAP operating income and non-GAAP net income are expected to increase approximately 35% to 40% year-over-year, despite a foreign-exchange expense headwind described as around $5 million. In a royalty IP model, that spread between modest current revenue surprise and much larger profit-growth guidance is the variant perception: investors looking only at the 0.9% top-line beat will underweight the embedded royalty base and the operating leverage if licensing signed today converts into royalties over expected product lives.

The financial trajectory supports that interpretation because Q4 was the highest revenue point in the supplied quarterly series, yet not an outlier in gross margin. Revenue moved from $24.2 million in Q1 FY2025 to $25.7 million in Q2 FY2025, $28.4 million in Q3 FY2025, and $31.3 million in Q4 FY2025, with QoQ growth of +5.9%, +10.5%, and +10.2% after the Q1 seasonal drop of -17.0%. Gross margin, however, was 85.6%, 86.2%, 88.0%, and 88.1% across those same quarters, so the revenue recovery did not require a margin giveback. That matters because Q1 FY2026 then steps down to $27.0 million with gross margin of 85.8%, a -13.6% QoQ reset but still +11.5% YoY. The market was right to expect seasonality after Q4, but it is too easy to treat the Q1 decline as a failed breakout. The more relevant comparison is Q1 against Q1: $27.0 million in Q1 FY2026 versus $24.2 million in Q1 FY2025, with the lower seasonal royalty mix still producing gross margin of 85.8% rather than breaking the model.

That revenue and margin path is why the composition of Q4 matters more than the 0.9% beat. Per CFO Yaniv Arieli, licensing and related revenue increased 11% year-over-year and 9% sequentially to $17.5 million, representing 56% of total revenue, while royalty revenue increased 2% year-over-year and 12% sequentially to $13.8 million, representing 44% of total revenue. The licensing side is the near-term stabilizer, but the royalty side is the proof point that shipped-unit categories are no longer only a future story. CEVA-powered devices shipped in 2025 reached 2.1 billion units, up 6% year-over-year; within that, Wi-Fi shipments grew 48% year-over-year and cellular IoT shipments grew 42% year-over-year. Those category growth rates are far more important than total units because the mix points to connectivity IP attaching across devices that customers can ship at scale. The company’s own language on booked demand was unusually concrete, with CEO Amir Panush saying, “Based on these signed agreements and our insights into customers' road maps, we estimate that they represent an aggregated lifetime royalty potential of $125 million over their expected product lives.” The phrase matters because it ties the number to signed agreements and customer road maps, not a general market forecast.

The same mix also explains why the GAAP and non-GAAP stories diverge, and PMs should not dismiss the divergence as accounting noise without sizing it. On the street-comparison basis, EPS was $0.18 and exactly in line. On the company’s own non-GAAP account, Yaniv Arieli said non-GAAP net income and non-GAAP diluted income per share increased 86% and 71% to $4.9 million and $0.18 year-over-year, compared with $2.7 million and $0.11 in Q4 FY2024. GAAP tells a different but still improving story: the quarterly history shows diluted EPS of -$0.04 in Q4 FY2025 versus -$0.07 in Q4 FY2024, and the call noted GAAP net loss of $1.1 million versus $1.7 million in the prior-year quarter. The conflict is not about revenue quality; it is about the cost structure below gross profit. Taxes were approximately $2.2 million, higher than guidance of $1.8 million, and affected by an adverse tax asset write-off, withholding-tax limitations, and geographic relocation of revenue recognition. That $2.2 million tax line is one reason the GAAP loss persists even as the model shows non-GAAP operating leverage.

The operating leverage case is credible because management quantified both the headwind and the underlying discipline. Excluding currency impacts, the 2026 non-GAAP expense base, including cost of goods and operating expenses, is expected to increase in the range of 1% to 3%, while non-U.S. dollar-based expenses, mainly research and development teams in Europe and Israel, are expected to increase by approximately 10% year-over-year, with an incremental impact of around $5 million. Total non-GAAP expenses for 2026 are expected to be $104.4 million to $108.4 million, with non-GAAP cost of goods sold increasing by approximately $0.5 million and non-GAAP operating expenses increasing by approximately $6.1 million. The market can look at that and see FX dilution; the variant read is that the core cost base is controlled enough for management to guide non-GAAP operating income and non-GAAP net income up approximately 35% to 40% year-over-year anyway. This is not a top-line acceleration thesis alone. It is a conversion thesis: if revenue grows above that 1% to 3% organic expense base and the FX drag does not worsen, the incremental dollar should matter.

The Q1 guide is the cleanest test of whether investors are over-penalizing seasonality. Management forecast Q1 revenue of $24 million to $28 million, and the quarterly history datapoint for Q1 FY2026 is $27.0 million, down -13.6% QoQ from the Q4 record but up +11.5% YoY. Arieli framed the decline as seasonal, saying, “Specifically for the first quarter of '26, with traditional seasonality in shipments of consumer IoT and mobile products post the holiday season, revenue is forecasted to be between $24 million to $28 million, sequentially lower than the record fourth quarter we just reported, but still significantly higher than the first quarter of 2025 at the midpoint.” The wording matters because it commits to the cause of the sequential decline, consumer IoT and mobile post-holiday seasonality, rather than blaming deal timing or a customer pushout. Gross margin guidance of approximately 86% on a GAAP basis and 87% on a non-GAAP basis also tells PMs what to expect mechanically: lower seasonal royalties pressure margin, but the model should stay in the mid-to-high 80s if the thesis is intact. A miss below the $24 million low end would break the seasonality framing; holding near the $27.0 million datapoint would support it.

The shipment data sharpen the second-order read-through for CEVA’s customers. Samsung’s LTE/5G modem exposure is tied to the 108 million units, or 18%, reported for mobile handset modems in the quarter, making handset modem royalties a meaningful but not dominant slice of the shipment base. MediaTek, which uses CEVA-X DSP core and subsystem, should read the 42% year-over-year growth in cellular IoT shipments as a positive signal for attach opportunities in connected endpoint road maps, though CEVA’s total industrial IoT-related annual shipments were down 31% year-over-year, so that read-through is strongest for cellular IoT rather than broad industrial end markets. NXP, with Bluetooth LE IP and SensPro AI DSP exposure, and Renesas, with Wi-Fi 6 and Bluetooth IP plus DSP for ITS, are the clearest named beneficiaries of record quarterly IoT shipments of 60 million units, up 30% year-over-year, and record Wi-Fi shipments of 86 million units, up 30% year-over-year. Broadcom, listed as a DSP-core customer, gets a narrower read-through from the DSP licensing and modem connectivity base, but the $17.5 million of Q4 licensing and related revenue, up 11% year-over-year, suggests CEVA’s customer pipeline is still being replenished even before royalties mature.

That customer read-through is not uniformly bullish, and the split is important because it argues against buying CEVA as a generic IoT recovery. In Q4, 479 million units were consumer IoT products, up from 459 million in Q4 FY2024, while 19 million units were industrial IoT products, down from 35 million units in Q4 FY2024. On an annual basis, consumer IoT-related shipments were 1.7 billion units, up 14% year-over-year, while industrial IoT-related shipments were 87 million units, down 31% year-over-year. The numbers say CEVA is getting paid on consumer and connectivity proliferation, not an across-the-board industrial rebound. That distinction matters for PMs comparing the print with other semiconductor exposures: a weak industrial channel at customers should not automatically invalidate the CEVA thesis if Wi-Fi and cellular IoT royalties keep scaling, but it does cap the breadth of the read-through for NXP and Renesas until industrial IoT stops declining from the 87 million annual-unit base.

The peer comparison reinforces why CEVA should be judged as IP leverage rather than as a conventional fabless growth stock. The peers table includes much larger businesses with revenue YoY growth rates ranging from -2.7% for 2454.TW to +85.2% for NVDA, and gross margins ranging from 40.9% for 6526.T to 81.9% for META. CEVA’s Q4 gross margin of 88.1% sits above those listed peer gross margins, but its Q4 revenue YoY growth of +7.1% is not trying to compete with +85.2% or +33.1% growth profiles. The equity question is therefore different: can CEVA turn low-double-digit or high-single-digit revenue growth into much faster profit growth because IP gross margin stays near the high 80s and expense growth is controlled? Management’s 2026 non-GAAP operating income and non-GAAP net income guide of approximately 35% to 40% year-over-year is the number that should anchor the debate. If the stock trades only on the $31.3 million versus $31.0 million revenue beat, it is using the wrong comp set and the wrong variable.

The call tone helps explain why the print may not have been fully capitalized immediately. The tone history shows Q1 FY2026 sentiment at 0.46, up only +0.03 from Q4 FY2025, while guidance_tone improved to 0.57, up +0.04, and prepared_sentiment rose to 0.60, up +0.10. That is supportive but not unqualified. Tone confidence fell to 0.42 from 0.44, uncertainty rose to 60.2 from 52.0, and qa_evasiveness moved to -2.0 from -22.8, a +20.8 change. The delivery therefore matched the thesis: management was more constructive in guidance language and prepared remarks, but the uncertainty index increased at the same time the company laid out FX pressure and Q1 seasonality. Investors who wanted a simple demand-acceleration call did not get one. Investors willing to underwrite a royalty conversion path got enough, but with clear numerical tripwires.

The reason to still lean positive is that the uncertain pieces are visible and measurable rather than hidden. Q1 GAAP OpEx is expected to be $27.6 million to $28.6 million, with $5.2 million of equity-based compensation expense, $0.1 million for amortization of acquired intangibles, and $0.1 million of costs associated with business acquisitions; non-GAAP OpEx is expected to be $22.2 million to $23.2 million. Taxes are expected to be approximately $1.3 million, down from the Q4 level of approximately $2.2 million, and share count is expected to be approximately 27.7 million shares on a GAAP basis and 29.4 million shares on a non-GAAP basis. Those numbers leave little mystery about what has to happen. If revenue lands in the upper half of $24 million to $28 million, gross margin stays near approximately 86% GAAP and 87% non-GAAP, and non-GAAP OpEx remains within $22.2 million to $23.2 million, the model should preserve the 2026 profit-growth narrative. If revenue is closer to $24 million while OpEx is near $23.2 million and FX-related costs push total 2026 non-GAAP expenses toward $108.4 million, the 35% to 40% non-GAAP income growth guide becomes more vulnerable.

What to watch next quarter is therefore specific. For Q1 FY2026, confirmation requires revenue at or above the $27.0 million datapoint within the $24 million to $28 million guide, gross margin near approximately 86% GAAP and 87% non-GAAP, and non-GAAP OpEx inside $22.2 million to $23.2 million. The royalty thesis needs Q1 commentary showing Wi-Fi and cellular IoT still tracking off the 2025 records of 48% year-over-year Wi-Fi shipment growth and 42% year-over-year cellular IoT shipment growth, while the customer read-through needs consumer IoT to remain on the 1.7 billion annual-unit trajectory and industrial IoT to stabilize from 87 million units after a 31% year-over-year decline. The bear case wins if management walks back the approximately 35% to 40% 2026 non-GAAP operating income and non-GAAP net income growth guide, if total non-GAAP expenses trend above $108.4 million, or if the $125 million aggregated lifetime royalty potential attached to signed agreements starts to look like licensing backlog without royalty conversion. Until one of those breaks, the small Q4 beat is the wrong conclusion from the event; the right conclusion is that CEVA has enough unit and licensing evidence for operating leverage to matter more than the headline surprise.