FormFactor’s HBM mix is pulling margins forward before Farmers Branch can dilute them

FormFactor printed the kind of quarter that should change the debate from “can HBM offset a soft test backdrop?” to “how much of the HBM margin curve is already embedded before the company spends through Farmers Branch?” What was priced in was a modest top-line beat: revenue came in at $215.2 million versus the street’s $210.6 million, a +2.2% surprise. What actually surprised was operating conversion: EPS was $0.46 versus the street’s $0.35, a +31.4% surprise, despite Q4 being the quarter where investors were already watching for the cost of capacity expansion. The variant perception is that the print is not just a cyclical DRAM recovery marker. It is evidence that FormFactor’s product mix, especially HBM probe cards, is pulling the company toward its target economics before the Farmers Branch investment shows up in revenue capacity. That matters because the market can discount capex years too mechanically: the company is committing $140 million to $170 million of 2026 cash expenditures for Farmers Branch and $20 million to $25 million of 2026 preproduction operating expenses, but the current margin trajectory says the HBM cycle is funding more of that burden than feared.

The surprise is cleaner when separated from the street setup. The revenue bar at $210.6 million already allowed for Q4 to be up from $202.7 million in Q3 FY2025 and for the HBM narrative to remain intact, so the $215.2 million outcome was helpful but not thesis-changing by itself. EPS was different. The $0.46 actual was not a rounding-error beat against $0.35, and it followed diluted EPS of $0.20 in Q3 FY2025 on the quarterly history basis and management’s own non-GAAP net income of $36.6 million or $0.46 per fully diluted share versus $25.7 million or $0.33 per fully diluted share in Q3. The reason this matters for a PM is that the stock debate should move from incremental demand to incremental margin dollars. Revenue rose +6.2% QoQ in Q4 FY2025, but gross margin moved to 42.8% from 39.7%, and the Q1 FY2026 history line shows revenue of $226.1 million with gross margin of 47.9%. Whether one uses the street-comparison print for the beat or management’s own reporting basis for company accounts, the direction is the same: the upside is disproportionately below revenue.

That financial trajectory is the core of the call, because Q1 guidance implies the company is running near the capacity conversation investors expected to matter only after Farmers Branch. CFO Stan Finkelstein said, “We expect Q1 revenues of $225 million plus or minus $5 million,” and the point is not just that the midpoint is above Q4. The call itself framed the investor tension when Charles Shi said the Q1 guide “implies basically $900 million annual run rate already” versus prior existing capacity “around maybe it can be slightly above $850 million.” The data pack does not provide management’s answer to that capacity question, so it should not be over-read. But the existence of the question is the tell: investors are already forcing Q1 guidance into the capacity framework, and the company is guiding non-GAAP EPS of $0.45 plus or minus $0.04 while startup expenses are rising. If the current footprint is already supporting a revenue level investors associate with constraint, then the incremental Farmers Branch risk is less about near-term demand validation and more about execution cost.

The margin guide is what makes that capacity question investable rather than academic. Finkelstein guided “higher non-GAAP gross margins of 45% plus or minus 150 basis points” for Q1, while also saying Q1 non-GAAP operating expenses should be $62 million plus or minus $2 million, approximately $4.5 million higher than Q4, mainly from Farmers Branch startup costs. That combination is the bull case in one sentence: higher gross margin is expected to more than coexist with the first visible step-up in expansion expense. The data pack gives the bridge. On a company non-GAAP basis, Q4 gross margin was 43.9%, up 290 basis points from 41% in Q3 and above the high end of the range; probe card segment gross margins were up 364 basis points to 44.5%, while Systems declined 50 basis points. The qualitative conclusion has to be tied to those numbers: the mix is not broadly improving across every business line, it is concentrated in probe cards, and that is exactly where HBM demand shows up. If a PM is skeptical of the multiple because a factory build can flatten earnings, the counter is that the probe-card gross margin expansion was already 364 basis points in Q4 before the Q1 revenue step.

The HBM language on the call narrowed the demand driver enough to assign second-order implications. CEO Mike Slessor said, “In the first quarter, we again expect to post an all-time DRAM record, this time on HBM strength, with contributions from both sustained demand in HBM3E and the early stages of the HBM4 ramp.” The words that matter are not “record” alone; they are “sustained demand in HBM3E” and “early stages of the HBM4 ramp,” because that means the Q1 lift is not just one product generation finishing a qualification cycle. For SK Hynix, a customer for DRAM/HBM wafer-test probe cards, the read-through is direct: FormFactor is seeing enough HBM3E demand and enough HBM4 ramp activity to guide Q1 revenue to $225 million plus or minus $5 million after Q4 revenue of $215.2 million. For Samsung, a customer for DRAM wafer test using ATRE technology, the second-order implication is less direct but still measurable: DRAM probe-card demand is strong enough that management expects another all-time DRAM record in Q1, and probe card segment gross margins rose 364 basis points to 44.5% in Q4. For Intel and TSMC, both customers for wafer-test probe cards in logic, the caution is that the data pack ties the upside language to DRAM/HBM, not logic; the Q4 company revenue beat was +2.2%, so the read-through to logic customers should be smaller than the read-through to memory customers unless Q1 shows broader mix evidence.

That customer read-through also explains why this print should be judged differently from the broader test and assembly peer set. FormFactor’s latest reported revenue YoY was +13.6% in Q4 FY2025 and gross margin was 42.8%, which places it close to 7729.T at +13.7% revenue YoY and 42.4% gross margin, rather than the higher-margin Japanese equipment names such as DSCSY at 70.8% gross margin or ATEYY at 67.4%. The comparison is useful because FormFactor is not winning the screen on absolute gross margin versus those high-margin peers; it is winning the inflection argument inside its own model. Peers such as 6871.T showed +48.3% revenue YoY and 47.3% gross margin, and ATEYY showed +43.8% revenue YoY, so FormFactor’s +13.6% does not make it the broadest cyclical acceleration story in the group. The investable distinction is that FormFactor’s margin moved from 39.7% in Q3 FY2025 to 42.8% in Q4 FY2025 and then to 47.9% in Q1 FY2026 in the quarterly history, while its own call included a target-model reference of 47% margins at $850 million a year run rate for revenue. The market may be using peer growth to frame the stock, but the better frame is how quickly FormFactor is approaching its own target economics.

The cash-flow line reduces the risk that the Farmers Branch build consumes the story before the HBM cycle pays for it. Finkelstein disclosed free cash flows of $34.7 million in Q4 versus $19.7 million in Q3, and operating cash flows of $46 million versus $27 million in Q3 2025. Total cash and investments were up $9.1 million to $275 million at quarter-end, while the remaining authorization under the $75 million two-year buyback program was $70.9 million. The buyback number matters because it shows management has barely used the authorization, and that is rational when 2026 cash expenditures for Farmers Branch are expected between $140 million and $170 million. But the more important conclusion is that the company entered the heavier spending year with $275 million in cash and investments and a Q4 free-cash-flow run rate that improved by $15 million sequentially. The bearish version is that preproduction operating expenses rise from $1.7 million in Q4 2025 to roughly $6 million in Q1 2026 and then to $20 million to $25 million over 2026. The bullish rebuttal is numerical, not narrative: the Q1 guide still calls for $0.45 plus or minus $0.04 of non-GAAP EPS.

The call delivery supports the thesis, but with one important warning embedded in the tone data. The tone history shows Q4 FY2025 sentiment at 0.55 and guidance_tone at 0.63, both higher than Q3 FY2025 sentiment of 0.52 and guidance_tone of 0.61, while prepared_sentiment stayed at 0.81. That is the language pattern one would expect after a quarter where Q4 revenue came in at the high end of the company outlook range of $205 million to $215 million and non-GAAP gross margin beat the high end. But the Q1 FY2026 call tone in the table is less clean: sentiment fell to 0.35 from 0.55, guidance_tone slipped to 0.56 from 0.63, and uncertainty rose to 62.2 from 48.9. That conflict matters. Management’s numeric guide is constructive, but the language model says the Q1 call was more uncertain even as ai_optimism rose to 0.59 from 0.41 and qa_evasiveness fell to 7.9 from 27.1. In plain terms, the company is giving better numbers while fielding a more complex set of investor concerns, chiefly capacity, startup costs, and sustainability of HBM demand.

That tonal conflict should not weaken the thesis, but it should define the risk. The prepared portion was less positive in Q1 FY2026, with prepared_sentiment down to 0.41 from 0.81, and qa_sentiment fell to 0.25 from 0.37, so there is no evidence that management was simply promotional. At the same time, tone_confidence rose to 0.47 from 0.30 and qa_evasiveness fell by -19.2 on the call-over-call delta. That combination is unusual: less euphoric prepared language, more uncertainty, but less evasive answers and higher confidence. The interpretation is that management is not hiding the cost and capacity complexity, but the numeric outlook is still firm. For a PM, this is preferable to a smooth call with soft guidance. The investable issue is whether the company can keep Q1 non-GAAP gross margin near 45% plus or minus 150 basis points while absorbing $62 million plus or minus $2 million of non-GAAP operating expenses and roughly $6 million of Farmers Branch preproduction operating expenses. If it can, the market has to underwrite a higher through-cycle earnings base.

The main bear case is not demand in Q1, because the guide already includes Q1 revenues of $225 million plus or minus $5 million and management expects another all-time DRAM record. The bear case is that HBM-driven probe-card margins are peaking just as Farmers Branch costs accelerate, turning Q4’s $0.46 EPS beat into a one-quarter event rather than a reset. The evidence against that bear case is the sequence: Q2 FY2025 gross margin was 37.2%, Q3 FY2025 was 39.7%, Q4 FY2025 was 42.8%, and Q1 FY2026 was 47.9% in the quarterly history, while company guidance for Q1 non-GAAP gross margin was 45% plus or minus 150 basis points. The evidence supporting the bear case is equally concrete: GAAP operating expenses were $67.3 million in Q4, Q1 non-GAAP operating expenses are guided to $62 million plus or minus $2 million, Farmers Branch preproduction operating expenses are rising to roughly $6 million in Q1 from $1.7 million in Q4, and full-year 2026 project cash expenditures are $140 million to $170 million. The conclusion is not that costs do not matter. It is that the margin expansion is showing up before those costs become a full-year drag, which gives management room to prove the factory is accretive to strategic capacity rather than dilutive to returns.

What to watch next quarter is therefore narrow and numerically testable. The thesis is confirmed if Q1 revenue lands within or above $225 million plus or minus $5 million, non-GAAP gross margin holds within or above 45% plus or minus 150 basis points, and non-GAAP EPS tracks the $0.45 plus or minus $0.04 guide despite non-GAAP operating expenses of $62 million plus or minus $2 million and roughly $6 million of Farmers Branch preproduction operating expenses. It is strengthened if management again ties DRAM upside to HBM3E demand and the HBM4 ramp, because that would extend the SK Hynix and Samsung read-through beyond a single quarter; it is weakened if the company stops using that HBM framing while revenue remains near the $225 million midpoint. The break point is margin: if Q1 revenue is near the guide but non-GAAP gross margin falls below the 45% plus or minus 150 basis points range, the market will be right to treat Q4’s +31.4% EPS surprise as mix timing rather than structural earnings power. If margin holds and the Farmers Branch 2026 ranges remain $140 million to $170 million of cash expenditures and $20 million to $25 million of preproduction operating expenses, the print argues for owning the capacity cycle rather than avoiding it.