They booked 70% of next fiscal year before they've secured enough fab capacity to deliver it.
Thesis: Semtech owns the bottleneck position in the 1.6T optical transition: FiberEdge TIA share went from 18% to over 50% in 800G and is on track for 50%+ in 1.6T. Backlog covers 70%+ of FY2028 with pricing confirmed and no erosion on booked orders. The market is anchoring on CopperEdge when 1.6T already surpasses 50% of data center revenue next quarter. The capacity ceiling is real but a demand signal, not a decline signal: they can't supply enough to meet bookings, and they're spending CapEx to fix it at 2% of sales.
Verdict: LONG — Conviction: MEDIUM
Catalyst: October 15 investor day where management lays out TAM, share, and multi-year targets. Hong deferred all revenue mix questions to that event, indicating they plan to raise the framework. Q3 print proving 45% sequential data center growth with gross margin at high end of guide.
Key Risk: Fabless capacity ceiling: Hong said secured capacity 'may not be enough in supporting FY 2028, especially second half.' Prepayment and CapEx mitigation is unquantified. If hyperscaler 1.6T qualification timing slips back after pulling forward, the backlog shrinks and the 45% sequential guide misses.
The Tell: Hong told Rolland: 'We were just not very sure about the qualification timing, and that's why we were a little conservative in guiding for Q2 at that time. Now we have all the backlog, and the customers want parts tomorrow.' He's confirming both that they sandbagged (structural beat, not noise) and that the entire Q3 guide rests on hyperscaler qualification timing that could theoretically reverse.
Friction Level: MODERATE_FRICTION — Both sides agree on the numbers. Bull reads backlog and pricing power as structural; bear reads the 1.6T beat as a qualification timing pull-forward with no stated durability and flags the fabless capacity ceiling as unquantified balance sheet risk.
2027Q2 · Aug 25, 2026
3 of 5 readings agreed3 LONG · 2 HOLDRead the full call
CEO Warns Capacity May Run Out While Guiding 45% Sequential Data Center Growth
Semtech owns the bottleneck position in the 1.6T optical transition: FiberEdge TIA share went from 18% to over 50% in 800G and is on track for 50%+ in 1.6T. Backlog covers 70%+ of FY2028 with pricing confirmed and no erosion on booked orders. The market is anchoring on CopperEdge when 1.6T already surpasses 50% of data center revenue next quarter. The capacity ceiling is real but a demand signal, not a decline signal: they can't supply enough to meet bookings, and they're spending CapEx to fix it at 2% of sales.
Hong told Rolland: 'We were just not very sure about the qualification timing, and that's why we were a little conservative in guiding for Q2 at that time. Now we have all the backlog, and the customers want parts tomorrow.' He's confirming both that they sandbagged (structural beat, not noise) and that the entire Q3 guide rests on hyperscaler qualification timing that could theoretically reverse.
Tenth consecutive quarter of revenue growth. $342M beat above high end of guide, EPS $0.71 up 73% YoY. Q3 guide of $410M with data center up 45% sequentially. Management admitted they sandbagged Q2 because 1.6T qualification timing was unclear, confirming conservative guidance.
Hong confirmed no price erosion on 800G and ability to pass through cost increases to customers. 'Availability is more important to the customers than the pricing.' Semiconductor products margin hit 62.8%, up 210bp sequentially. Customers are ordering on availability, not negotiating on price.
Hong stated remaining fiscal 2027 is 'all booked' and FY2028 is 'over 70% there.' No cancellation terms disclosed but specific percentage of next fiscal year committed. Management is securing capacity ahead of demand based on this backlog, spending CapEx proactively.
Secured fab equipment deliveries, acquired clean room space, completed photonic acquisitions in under six months, acquired a fully facilitized fab to expand capacity 3-4x by year end. CapEx at 2% of sales going to <5%. Management is spending to remove the bottleneck before demand hits it.
Hong admitted: 'capacity we have secured may not be enough in supporting FY 2028, especially second half.' Fabless company depends on foundry and OSAT partners who supply everyone. Mitigation is prepayments, CapEx, and 'some other means' with no quantification. Balance sheet risk on capacity relationship.
Zero disclosure on cancellation terms, deposit structures, or whether surge orders reflect pull-forward. Management acknowledged upside came from 'a little bit earlier inflection for 1.6T FiberEdge' and qualification timing acceleration, raising durability questions on the booked book.
+80
CONFIDENT"Hong volunteered specific numbers on every question: 50%+ FiberEdge share, 1.6T surpassing 50% of DC revenue in Q3, $80-90 content per transceiver at 3.2T, 3-4x fab capacity expansion on track. He corrected himself unprompted to clarify content figures. No deflection on pricing, capacity, or backlog."
Specific revenue, margin, backlog, pricing, and capacity numbers throughout. Cash flow statements verified. Only gaps: no backlog cancellation terms disclosed, no quantified capacity prepayment commitments.
October 15 investor day where management lays out TAM, share, and multi-year targets. Hong deferred all revenue mix questions to that event, indicating they plan to raise the framework. Q3 print proving 45% sequential data center growth with gross margin at high end of guide.
Fabless capacity ceiling: Hong said secured capacity 'may not be enough in supporting FY 2028, especially second half.' Prepayment and CapEx mitigation is unquantified. If hyperscaler 1.6T qualification timing slips back after pulling forward, the backlog shrinks and the 45% sequential guide misses.
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 Nov 23, 2026.