Tower’s SiPho Ramp, Not the Penny Beat, Is the Mispriced Part of the Print

The right read on this print is not that Tower delivered a large top-line beat. It did not. The street-comparison basis shows Q4 FY2025 revenue of $440.2 million versus $439.8 million expected, a +0.1% surprise, while EPS of $0.78 versus $0.67 expected delivered a +16.4% surprise. What was priced in was already a Q4 revenue step-up after Q3 FY2025 revenue of $395.7 million, and the actual revenue result did little to change that near-term demand view. What surprised was the quality of the dollars: Q4 FY2025 gross margin reached 26.7%, up from 23.5% in Q3 FY2025 and 20.4% in Q1 FY2025, while diluted EPS on the company history basis moved to $0.70 from $0.47 in Q3 FY2025 and $0.35 in Q1 FY2025. The variant perception is that investors anchoring on the +0.1% revenue beat are missing the fact that Tower’s Q4 print converted a modest revenue surprise into a much larger earnings surprise because the incremental revenue is increasingly photonics-weighted rather than merely wafer-volume driven.

That distinction matters because the market’s likely setup before the print was straightforward: revenue had already recovered sequentially for three quarters, from $358.2 million in Q1 FY2025 to $372.1 million in Q2 FY2025, $395.7 million in Q3 FY2025, and $440.2 million in Q4 FY2025, so the debate should have been whether utilization and mix could repair margins. The print answered that more clearly than the top-line beat did. Gross margin expanded from 20.4% in Q1 FY2025 to 26.7% in Q4 FY2025, while revenue moved from $358.2 million to $440.2 million over the same span. Russell Ellwanger put the operating leverage in unusually explicit terms: “The revenue growth from Q1 to Q4 of 2025 was $82 million, of which there was a $40 million net profit drop down and almost 50%, to be exact 48.78% and this due to the high value of the incremental Photonics revenue.” That sentence earns attention because it ties the profit step-up to a named technology mix, not to generic factory absorption, and because 48.78% is a conversion rate that can re-rate a specialty foundry if repeatable.

The financial trajectory also explains why the Q1 guide should not be read as a thesis-breaker even though it steps down sequentially. Management guided Q1 FY2026 midrange revenue to $412 million plus/minus 5%, and the quarterly history shows Q1 FY2026 revenue at $413.6 million, gross margin at 26.8%, revenue QoQ of -6.0%, revenue YoY of +15.5%, and diluted EPS of $0.57. The sequential revenue decline from Q4 FY2025’s $440.2 million would normally revive concerns that Q4 was a pull-forward, but the gross margin staying at 26.8% versus 26.7% in Q4 FY2025 argues the mix benefit did not disappear with lower revenue. That is the key conflict to interpret: revenue QoQ was -6.0%, but gross margin was 26.8%, not a retracement toward Q1 FY2025’s 20.4%. The market may price Tower as if quarterly revenue volatility still dictates earnings power; the Q1 FY2026 gross margin print suggests mix is becoming a second, and potentially larger, driver.

The longer-range management model is aggressive enough that it should not be capitalized blindly, but it frames the option value created by the 2025 mix data. Tower reported full year 2025 revenue of $1.566 billion in Ellwanger’s prepared remarks, and Oren Shirazi described the model as “targeting $2.84 billion in annual revenue, which is $1.27 billion higher or 81% higher in revenue than our actual full year 2025 revenue.” The same model targets $900 million in annual operating profit, 32% operating margin, $750 million in net profit, and 26% net margin. Those are not quarterly guidance numbers, but they are useful because the bridge is not just more wafers; it assumes the company can scale the mix it just showed. The near-term proof point is that Q4 FY2025 net profit was $80 million or 18% net margin on the company’s reported basis, up from 11% in Q1 ’25, 13% in Q2 ’25, and 14% in Q3 ’25. The 2026 model’s 26% net margin is therefore not validated by Q4, but it is directionally consistent with a 2025 exit rate that already moved well beyond the first-half margin structure.

The photonics details make the thesis less abstract and more investable. Silicon germanium and silicon photonics represented 27% of corporate revenues in 2025 and were $421 million, up from $241 million or 17% in 2024. SiPho alone was $228 million in 2025, up from $106 million in 2024, and in Q4 RF infrastructure revenues were 32% of corporate revenue, with SiPho at $95 million or a $380 million annual run rate. Those numbers are the center of the call, not the $440.2 million revenue line. If SiPho can run near the Q4 annualized level while Tower adds capacity across Fab 3 Newport Beach, Fab 9 San Antonio, Fab 7 Uozu, Japan, and Fab 2 Migdal Haemek, then the 2025 profit conversion is not a one-quarter artifact. If it cannot, the thesis breaks quickly, because 2025 SiPho of $228 million is meaningfully below the Q4 $380 million annual run rate, and the burden shifts to proving that Q4 was not an unusually dense shipment quarter.

The capital plan is the clearest signal that management is underwriting demand beyond normal cyclical recovery, and it also introduces the main execution risk. Tower announced additional CapEx investment of $270 million on top of the previously announced $650 million capacity expansion plan, bringing the above-stated CapEx investments to $920 million. Shirazi said approximately 28% of the $920 million had already been paid, while the remaining 72% is expected to be paid in 2026 and 2027. That creates a binary setup for investors: if SiPho demand supports the capacity ramp, Tower’s model can move toward 39.4% gross margin and 31.7% operating margin; if demand slips, shareholders own a larger fixed-asset base at exactly the point the company’s assets already totaled over $3 billion, including $1.5 billion in fixed assets and $1.7 billion of current assets. The balance sheet can fund the expansion, with a current assets ratio at about 6.5x and shareholders’ equity of $2.9 billion at the end of December 2025, but funding capacity is not the same as utilization certainty.

The customer and supplier read-through is narrow but important because the revenue mix points to where Tower’s ecosystem gets leverage. For NXP, Tower’s Q4 FY2025 gross margin of 26.7% and RF infrastructure at 32% of corporate revenue suggest specialty analog/RF/power foundry capacity is tightening first in differentiated platforms rather than commodity nodes. For ON Semiconductor, the relevance is less SiPho-specific and more that Tower’s full year 2025 revenue rose 9% to $1.566 billion while gross profit and operating profit were $364 million and $194 million, respectively, indicating specialty foundry demand can improve profit even before all end markets recover uniformly. For STMicroelectronics, the TPSCo JV in Japan sits directly in the footprint discussion because Tower cited Fab 7 Uozu, Japan among the silicon photonics production ramps. On the supplier side, Soitec’s Photonics-SOI exposure through the PH18 SiPho platform gets the cleanest positive read-through: SiPho revenue of $228 million in 2025, up from $106 million in 2024, and Q4 SiPho at $95 million or a $380 million annual run rate imply substrate demand is being pulled by Tower’s highest-value incremental mix, not by a broad foundry recovery alone.

The peer comparison reinforces why Tower’s print should be judged on specialty mix rather than absolute scale. In the foundry peer table, TSEM’s latest reported quarter shows revenue YoY of +15.5% and gross margin of 26.8%. That is below UMC’s 29.2% gross margin and 5347.TWO’s 29.3% gross margin, but above 3105.TWO’s 26.3% gross margin and far above 6770.TW’s 8.1% gross margin; it also compares with GFS at 27.6% gross margin and +3.1% revenue YoY. Tower is not trying to look like TSM at 66.2% gross margin and +35.1% revenue YoY. The more relevant point is that Tower’s +15.5% revenue YoY with 26.8% gross margin puts it closer to differentiated specialty recovery than to flat mature-node digestion, and the SiPho mix gives it a specific route to close the margin gap versus peers sitting around 29.2% and 29.3%.

The call delivery supports the thesis, but with one warning flag that should keep position sizing disciplined. The tone history shows Q1 FY2026 sentiment at 0.35 versus Q4 FY2025 at 0.33, guidance_tone at 0.66 versus 0.59, prepared_sentiment at 0.72 versus 0.66, and qa_sentiment at 0.15 versus 0.09. More important, uncertainty fell to 46.6 from 55.0 and qa_evasiveness fell to 8.4 from 29.9, so management’s delivery became more direct after the Q4 event even as the company entered the year with a sequential revenue decline. The warning is tone_confidence: it fell to 0.27 from 0.41, which conflicts with the improved guidance_tone and lower qa_evasiveness. The best interpretation is not that management sounded less positive, since sentiment and guidance_tone improved, but that the transcript carried more model-dependent ambition than quarter-specific certainty.

That tone pattern matters because management is asking investors to bridge from a reported year to a much larger capacity-backed model. The prepared commentary included precise targets for $2.84 billion in annual revenue, 39.4% gross margin, 31.7% operating margin, $900 million in annual operating profit, $750 million in net profit, and 26% net margin. At the same time, Q1 FY2026 revenue was $413.6 million, well below Q4 FY2025 revenue of $440.2 million, and Q1 FY2026 diluted EPS was $0.57 versus $0.70 in Q4 FY2025. That is why the print is actionable but not riskless: the next leg in the stock should come from evidence that Tower can keep Q4-like margin structure as revenue normalizes, not from assuming a straight line to the long-term model. The numbers to emphasize are not only 2026 targets, but the already-observed gross margin stability from 26.7% in Q4 FY2025 to 26.8% in Q1 FY2026 despite revenue QoQ of -6.0%.

What would change the view is therefore concrete. For the next quarter, the first confirmation level is revenue around the Q1 FY2026 outcome of $413.6 million against the prior guide of $412 million plus/minus 5%, with gross margin staying near 26.8% rather than sliding back toward 23.5% from Q3 FY2025 or 20.4% from Q1 FY2025. The second confirmation is SiPho disclosure that keeps the business near the Q4 reference point of $95 million or a $380 million annual run rate, because full year 2025 SiPho of $228 million only supports the thesis if Q4 was the start of a higher run rate, not a spike. The third is CapEx execution: the remaining 72% of the $920 million expected to be paid in 2026 and 2027 must be tied to customer-backed ramps, especially across Fab 9 San Antonio, Fab 7 Uozu, Japan, and Fab 2 Migdal Haemek. The break points are equally clear: gross margin below 26.7%, SiPho tracking materially below the $95 million Q4 marker, or management walking back the $2.84 billion revenue and $900 million operating profit model would undermine the variant view that Tower’s earnings power has reset. Until those break points appear, the +16.4% EPS surprise against only +0.1% revenue surprise is the signal, not the noise.