Management said they can't grow HPC this year, then guided full-year growth from 25% to 35% anyway.
Thesis: Chu owns the thermal test bottleneck at 1000W+ processors with one credible competitor, Hon Precision, and its booked orders prove it. Capacity is doubling into a $190M qualified and $250M in-qualification pipeline while the core business runs at 80% utilization, two quarters into a historical six-quarter capex cycle. Near-term margin compression from memory pre-buys is a cost timing issue, not a structural pricing failure.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Investor day November 10 launching a long-term financial framework, plus the fifth qualification green light management expects within a month and the $26M Eclipse order shipping in Q4 against a 'flattish to Q3' core.
Key Risk: Hyperscaler custom ASIC and GPU capex inflects while $445M of the pipeline sits in early-stage engagement with no qualification timelines, and memory input costs stay elevated with pass-through still unagreed. Core cyclical decline hits the same quarter.
The Tell: Unprompted, Luis moved past guidance: 'if we're delivering $100 million to $110 million this year, we should have the capacity to do more than $200 million, probably closer to $250 million by the beginning of next year.' He then added 'or more' and a path to triple output. He set 2027 expectations without an analyst asking.
Friction Level: MODERATE_FRICTION — Both sides accept the beat and the $26M booked order. Disagreement is whether the $850M pipeline, 52% early-stage, converts against a hyperscaler capex cycle while Cohu has zero pricing power on memory pre-buys.
2026Q2 · Jul 30, 2026
3 of 5 readings agreed3 LONG · 2 HOLDRead the full call
Q4 Output Maxed Out, Pipeline Still Adding $100M
Chu owns the thermal test bottleneck at 1000W+ processors with one credible competitor, Hon Precision, and its booked orders prove it. Capacity is doubling into a $190M qualified and $250M in-qualification pipeline while the core business runs at 80% utilization, two quarters into a historical six-quarter capex cycle. Near-term margin compression from memory pre-buys is a cost timing issue, not a structural pricing failure.
Unprompted, Luis moved past guidance: 'if we're delivering $100 million to $110 million this year, we should have the capacity to do more than $200 million, probably closer to $250 million by the beginning of next year.' He then added 'or more' and a path to triple output. He set 2027 expectations without an analyst asking.
Revenue $149M above guidance midpoint, gross margin 45.5% above guide, full-year growth raised from 25% to 35%, tight new HPC range of $100-$110M versus prior $80-$100M. Consistent conservative guidance being marked up.
Luis names a single competitor, Hon Precision, and quotes a Bay Area peer calling Cohu's thermal technology 'the best thermal in the market, period.' One credible supplier against a constraint customers cannot solve elsewhere is the setup for ASP increases.
Malaysia floor fitting out now, new building in Q1 to free production space, Philippines thermal head capacity expanding in parallel, output doubling by year-end and 100%+ by mid-2027. Capex only 2% of revenue, so the build is self-funded.
Luis gave exact capacity math, named the competitor, quantified per-customer pipeline at $30-$60M, and Jeff quantified Q3 growth as 50% HPC and 50% core, roughly $10M each. No hedging on the parts that matter.
Luis: 'I don't think we have much more room to grow in the HPC side this year' and Q4 output is 'maxed out.' The fastest-growing segment is supply-constrained exactly when the $26M order lands.
Jeff is pre-buying memory at elevated cost with no agreed recovery: 'We've just started conversations with customers... stay tuned.' Q3 gross margin guides down to 45% on 14% sequential revenue growth.
80% serialization is framed as a test capex turning point for IDMs, but it also means limited core growth headroom. The six-quarter historical pattern Jeff cites came from 2021-2022 and ended in normalization.
$445M of the $850M pipeline is early-stage engagement across 10 customers with no qualification timelines, and the $100M pipeline increase came on 'better visibility in forecast' rather than new qualifications.
+78
CONFIDENT"Luis volunteered specifics nobody asked for: capacity to 'more than double output,' '$200-$250 million HPC revenue next year or more,' a path to triple output, and named Hon Precision as the only competitor. Direct answers throughout, no deflection."
Booked $26M single order, per-segment order data, explicit capacity expansion timelines, pipeline staged into qualified/in-qual/early buckets, and named competitor. Only softness is memory pass-through timing.
Investor day November 10 launching a long-term financial framework, plus the fifth qualification green light management expects within a month and the $26M Eclipse order shipping in Q4 against a 'flattish to Q3' core.
Hyperscaler custom ASIC and GPU capex inflects while $445M of the pipeline sits in early-stage engagement with no qualification timelines, and memory input costs stay elevated with pass-through still unagreed. Core cyclical decline hits the same quarter.
The verdict is public. The case behind it is Pro: the forensic reading of guidance and language, the bull and bear cases, the red flags, and the play. It opens to everyone on Oct 28, 2026.