Kulicke’s miss was a customer-timing reset, not a demand-cycle reset

The first read of this print is ugly in the place investors were primed to care about: revenue. Street expectations were for $169.8 million, and actual revenue came in at $148.4 million, a -12.6% surprise. That is not a rounding error, and it matters because the stock’s recovery case depended on evidence that the trough in core assembly tools was already behind the company. The offset, and the reason the print should not be treated as a wholesale thesis break, is that EPS beat anyway: actual EPS was $0.07 against $0.06, a +16.7% surprise. The market was priced for a revenue recovery with modest earnings, while the actual surprise was the inverse, a sharper top-line shortfall but better-than-modeled earnings conversion at the trough. That distinction is the whole investment debate after this call: if the revenue miss was broad-based end-demand deterioration, the beat is low-quality; if it was concentrated timing tied to one customer facility, the September guide is a credible bridge into a fiscal 2026 upcycle.

The company’s own language supports the concentrated-timing interpretation, but only narrowly. Fusen Ernie Chen framed the issue as something already disclosed and customer-specific: “We mentioned last quarter, this was a focus uniquely with a certain customer production facility in Southeast Asia, and we anticipate it will persist through the September quarters.” The word that matters is “uniquely,” because it tries to fence the weakness away from the entire installed base; the phrase that keeps the debate open is “persist through the September quarters,” because it says the drag is not done. Investors should not pretend this was a clean quarter, since Q3 FY2025 revenue of $148.4 million was down -18.3% year over year and -8.4% sequentially. But the defense is equally clear: gross margin came in at 46.7%, not at a level that would normally accompany an uncontrolled volume air pocket, and the company guided September revenue to $170 million with gross margins at 47%. The September setup therefore prices in a partial recovery before the offending facility issue has fully cleared, which is exactly why the next quarter is an unusually high-signal test.

That financial trajectory is easier to understand against the last several quarters: Kulicke’s revenue has been trapped below the prior cycle’s cleaner recovery levels, while gross margin has been distorted by discrete trough-quarter effects rather than sliding in a straight line. The recent path includes a Q2 FY2025 gross margin of 24.9% and a Q3 FY2025 rebound to 46.7%, which argues that the June quarter’s margin structure was not simply a function of revenue scale. In other words, the revenue miss does not mechanically imply a broken model. Lester A. Wong reinforced the margin floor on the call by saying, “Gross margin came in at 46.7%, and we delivered $0.07 of non-GAAP earnings per share.” That quote is useful not because it repeats the table, but because management paired margin and non-GAAP EPS as the proof point that the company could still defend earnings despite an underloaded revenue line.

The guide is the hinge between a one-customer disruption and a broader cycle problem. For September, management guided revenue to $170 million, gross margins at 47%, and non-GAAP earnings per share of $0.22. Those are company-basis guideposts, not the same street-comparison basis as the reported surprise, and the distinction matters because the actual Q3 print missed consensus revenue by -12.6% even though the company is guiding sequential improvement. What was priced in was likely a cleaner digestion of the Southeast Asia issue and a print closer to the street’s $169.8 million revenue estimate. What actually surprised was the reset of the June base to $148.4 million, paired with a September guide that tells investors not to extrapolate the miss linearly. The burden has now shifted: the stock no longer needs investors to underwrite a broad acceleration immediately, but it does need the September guide to prove that the facility drag is bounded.

The second-order read-through is most actionable for the OSAT and advanced-packaging customers listed in the supply chain, because Kulicke’s issue appears customer-production-specific rather than tool-category-wide. ASE Group, JCET, Samsung, and Amkor should not be read as receiving a blanket negative signal from this print; the company cited a “certain customer production facility in Southeast Asia,” not a collapse in wire bonding demand across all customers. The magnitude to attach to that caution is the $170 million September revenue guide after a $148.4 million June print, because management is still underwriting a rebound while saying the issue persists. For TSMC, the more relevant implication is not the June miss but Kulicke’s continued emphasis on thermocompression bonding and FTC tied to advanced packaging; management said it plans “initial higher volume production to begin in fiscal 2026” driven by a memory transition. That does not prove an immediate CoWoS/InFO order inflection for TSMC, but it preserves a fiscal 2026 equipment option in thermocompression bonding rather than pushing it out of the model.

The utilization commentary puts a ceiling on how bullish one should be about the near-term recovery. Wong said overall utilization was “about 81%,” while auto utilization was “below 70%.” Those numbers explain why the guide can recover without the company claiming a synchronized end-market rebound. An 81% overall utilization rate is not recessionary across the base, but auto below 70% is a visible drag that limits the breadth of the September recovery. The EV charging infrastructure commentary is therefore important only as a longer-cycle offset, not as a Q4 rescue. Chen said charging-related infrastructure is expected to exceed a “20% CAGR over the next 5 years,” but that sits outside the immediate September-quarter proof period. The more disciplined conclusion is that Kulicke has pockets of demand improvement and longer-cycle power opportunities, while auto remains too weak to justify paying for a broad-based industrial rebound today.

The peer context sharpens the variant perception because Kulicke’s weakness is not what the rest of test and assembly equipment is broadly reporting. In the latest peer set, several assembly-adjacent Japanese names are showing revenue YoY growth well above Kulicke’s Q3 FY2025 -18.3%, including 6871.T at +48.3% and ATEYY at +43.8%. Kulicke’s gross margin of 46.7% also sits closer to 6871.T at 47.3% than to lower-margin peers such as 6315.T at 36.2%, so the print is not a margin-collapse story relative to the group. The problem is revenue timing and exposure mix. That comparative point is why the miss should not be dismissed as “the cycle,” because the peer data says the cycle is not uniformly weak. It also means management does not get an easy macro excuse if September fails; the company needs to prove its specific customer issue is in fact specific.

The call delivery itself was better than the revenue miss, though not clean enough to call it a full confidence reset. The tone history shows Q3 FY2025 sentiment at 0.39, guidance_tone at 0.37, and tone_confidence at 0.62. That combination matters because tone confidence rose even as the revenue miss landed, suggesting management communicated the bridge with more specificity than in the prior call. But the split between prepared_sentiment at 0.66 and qa_sentiment at 0.11 is the red flag: scripted remarks carried the recovery narrative, while the Q&A sounded much less constructive. The saving grace is that qa_evasiveness fell to 37.2 from 83.8, which means the weaker Q&A tone was not simply management dodging. The better interpretation is that management was specific but not euphoric, and that is exactly the posture one would expect when a customer facility issue is still ongoing into September.

That tone pattern also helps interpret the capital-return message, which otherwise risks being overread. Wong said Kulicke repurchased approximately 668,000 shares, equivalent to 1.3% of diluted shares, during the June quarter. He also said that over a “7-quarter period,” the company deployed over $270 million in dividends and open-market repurchases. Those commitments matter because they indicate management is treating the trough as cyclical and balance-sheet-manageable, not as a structural impairment requiring cash preservation. But buybacks cannot substitute for orders, and the share count support should be viewed as a downside buffer rather than the central bull case. The central bull case is still September revenue delivery, margin retention near the guided level, and evidence that fiscal 2026 memory-related production is converting from evaluation to higher-volume demand.

The technology option is where the market may be most underpricing the print, precisely because the quarter itself was so underwhelming. Chen said the company is planning initial higher-volume production in fiscal 2026, “driven initially by an exciting technology transition within the memory market.” The useful part of that statement is the date and the customer adoption stage, not the adjective. Management also framed FTC as capable of outpacing overall TCV growth and later discussed a 2028 market reaching about $1 billion, with Kulicke targeting about $250 million to $300 million in 2028. Those are long-dated claims, but they give PMs a yardstick: the stock does not need to discount the full 2028 target after a $148.4 million quarter, but it also should not be valued as if Kulicke is only a legacy wire-bonder. The mispricing is that a June miss tied to one facility can obscure the option value of memory and FTC transitions that management still places in fiscal 2026.

The numbers do conflict in one place, and that conflict is why the thesis should be risk-managed rather than chased blindly. Street-basis revenue was $148.4 million, far below the $169.8 million estimate, while company-basis guidance calls for $170 million in September revenue and 47% gross margin. If the Southeast Asia facility issue is still persisting through September, a sequential guide to $170 million requires strength elsewhere or partial normalization within the same affected customer path. That is plausible given overall utilization at about 81%, but it is not proven while auto remains below 70%. The EPS beat helps, but it does not eliminate top-line risk, because operating expense is still material: Wong cited total operating expense of $75.3 million on a GAAP basis and $68 million on a non-GAAP basis. At this revenue level, the company needs the September sales recovery to carry the earnings bridge.

What to watch next quarter is therefore unusually concrete. The thesis is confirmed if September revenue is near the company’s $170 million guide, gross margin holds around 47%, and non-GAAP earnings per share tracks the $0.22 target despite the Southeast Asia issue persisting through the quarter. It is further confirmed if management repeats fiscal 2026 initial higher-volume production for memory-related technology without pushing the timing out, and if utilization commentary improves from about 81% overall or auto moves above the “below 70%” level. The thesis breaks if September revenue lands closer to the June base of $148.4 million than the $170 million guide, if gross margin falls materially below 47% without a one-time explanation, or if FTC and memory language loses the fiscal 2026 production marker. This print was not a clean beat, and the revenue miss was real; the actionable view is that the market may be overpricing the miss as cycle-wide while underpricing the evidence that the margin model and fiscal 2026 advanced-packaging option survived the worst quarter of the customer-timing disruption.