Q3 gross margin guides to 54% but 300 basis points of that is a one-time $79 million tariff refund the CFO already reset to a 51% baseline.
Thesis: FormFactor owns a real bottleneck: HBM4 stack final test, CPO insertion one, and MEMS probe qualification at every major customer, delivered into a footprint that is full. Four quarters of 30%+ growth with baseline gross margin up 430bps sequentially (51%, ex-tariff) proves pricing and cost leverage are structural, not tariff-flattered. The Street will anchor on the 54% headline and get marked down when the refund rolls off in Q4, which is the entry, not the exit, because Farmers Branch capacity lands into 2027 demand that is already backlogged behind their own build.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Q3 GPU production revenue first prints, CPO crosses $20 million a full quarter early, and Farmers Branch begins ramping in Q4 into 2027 hyperscaler ASIC MEMS content, the same inflection GPUs went through two generations ago.
Key Risk: Q4 guide-cut risk on margin optics when the $79 million IEEPA refund rolls off and Farmers Branch ramp inefficiency hits COGS before accretion arrives in 2028, on a stock where bulls are extrapolating 54% into 2027 exits.
The Tell: McKinnis volunteered the baseline reset unprompted in Q&A: 'The recurring or sustainable element of gross margins, we believe, is more like 51% at the current volumes and mix... That's the new baseline that you should be thinking of.' He corrected an analyst's read of the 53.3% beat before being asked, which means he knew the 54% Q3 headline was about to get flagged and moved to pre-empt the multiple on it.
Friction Level: MODERATE_FRICTION — Same facts, different read: the 54% Q3 guide is 51% baseline plus non-recurring tariff refunds and a DRAM mix shift toward lower-value DDR, so the fight is whether headline margin or baseline margin sets the multiple through 2027.
2026Q2 · Jul 29, 2026
5 of 5 readings agreed5 LONGRead the full call
Sold Out Into Their Own Capacity Ceiling, Funding The Build From Cash Flow
FormFactor owns a real bottleneck: HBM4 stack final test, CPO insertion one, and MEMS probe qualification at every major customer, delivered into a footprint that is full. Four quarters of 30%+ growth with baseline gross margin up 430bps sequentially (51%, ex-tariff) proves pricing and cost leverage are structural, not tariff-flattered. The Street will anchor on the 54% headline and get marked down when the refund rolls off in Q4, which is the entry, not the exit, because Farmers Branch capacity lands into 2027 demand that is already backlogged behind their own build.
McKinnis volunteered the baseline reset unprompted in Q&A: 'The recurring or sustainable element of gross margins, we believe, is more like 51% at the current volumes and mix... That's the new baseline that you should be thinking of.' He corrected an analyst's read of the 53.3% beat before being asked, which means he knew the 54% Q3 headline was about to get flagged and moved to pre-empt the multiple on it.
Q2 revenue $258.2M vs $240M midpoint, gross margin 53.3% vs 51% guide high, EPS $0.82 vs $0.56, then Q3 guided to a record $270M +/- $10M with $0.86 EPS. Third consecutive record quarter and four quarters of 30%+ revenue growth.
Baseline non-GAAP gross margin improved to ~51% at Q2 volume, up 430bps sequentially and ~1,500bps over four quarters. CFO attributes one third to durable baseline cost reductions and one third to volume leverage, not just the one-time tariff and precious-metal items.
$140-170M Farmers Branch build largely self-funded from operating cash (FCF $52.6M in Q2), plus a $24.2M Texas grant. Site roughly doubles California probe capacity and is the stated governor on fabless CPU share growth.
Content-per-test-dollar rising across HBM4 SmartMatrix, CPO insertion one, data center CPUs, and hyperscaler ASIC MEMS simultaneously. Probe cards are design-specific consumables that flow through as customers shift wafer starts.
Incumbent share at HBM final stack test with two of three makers, qualification plus multiple design wins at a new fabless XPU customer, and lab-to-fab decade lead on CPO. Switching means requalifying at the critical known-good insertion.
Q3 guide of 54% gross margin includes ~300bps from non-recurring $79M IEEPA tariff refunds. CFO states the sustainable level is 51%, yet prepared remarks sell the 54% milestone as path to the 55% 2030 target.
Q3 DRAM held flat to Q2's record but management concedes a 'significant underlying shift from HBM to DDR' and 'less favorable product mix in DRAM.' Flat revenue masks degradation in the highest-margin product line.
Fabless CPU share is 'low single-digit' with Farmers Branch named as the governor until Q4 2026 ramp. Output gains came from yield and cycle time, not new capacity. Demand exceeds supply at their own fab.
ProCard margin gains driven by 'higher factory utilization' while CFO says they 'see more constraints across the supply chain at current production levels.' Full utilization framed as execution is actually the growth ceiling.
Prepared remarks are all-time records with no hedging; in Q&A the CFO resets baseline margin to 51% and the CEO calls CPO quarterly forecasting 'a bridge too far.' Script sells $1.6B/55%/$5 by 2030 while the CFO quietly caps near-term margin optics.
+70
CONFIDENT"Prepared remarks carry zero hedging: all-time records, three-fold EPS, $1.6B/55%/$5 2030 target, and a $270M guide delivered flat. Caution appears only in Q&A, where McKinnis volunteers that the sustainable margin is 51% and Slessor concedes CPO quarterly forecasting is 'a bridge too far.'"
Dense quantitative disclosure: segment margins, tariff refund dollars, opex ratio, FCF, capacity timing, per-customer share, and specific product milestones. Even the weak spots were volunteered with numbers.
Q3 GPU production revenue first prints, CPO crosses $20 million a full quarter early, and Farmers Branch begins ramping in Q4 into 2027 hyperscaler ASIC MEMS content, the same inflection GPUs went through two generations ago.
Q4 guide-cut risk on margin optics when the $79 million IEEPA refund rolls off and Farmers Branch ramp inefficiency hits COGS before accretion arrives in 2028, on a stock where bulls are extrapolating 54% into 2027 exits.
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 27, 2026.