FormFactor’s HBM mix is real, but the print exposed a margin bridge the Street was not paid to underwrite

The investable message from this print is not that demand collapsed. It did not: revenue of $195.8 million missed the Street’s $199.4 million by only -1.8%, and management’s own Q3 outlook still points to $200 million plus or minus $5 million. The problem is that the earnings power attached to that revenue is weaker than the market likely assumed, with EPS of $0.27 versus the Street’s $0.30 and a -10.0% surprise. What was priced in was a cleaner HBM-led recovery: the stock needed proof that advanced memory probe card demand could pull the model toward the long-term margin framework. What actually surprised was the cost and mix drag under that growth: Q2 gross margin in the financial history was 37.2%, while management is adding Farmers Branch expense into Q3 rather than letting volume drop cleanly through the P&L. The thesis is therefore deliberately narrow: HBM is becoming large enough to matter, but not yet large enough to rescue margins from foundry and logic volatility plus a higher fixed-cost base.

That distinction matters because the company did deliver the kind of end-market rotation bulls wanted to see, just not with the operating leverage they needed. Shai Shahar’s most important line was not the broad revenue guide; it was the mix disclosure that “DRAM revenues were $57.1 million in Q2, $8.2 million or 16.8% higher than the first quarter, an increase to 29.1% of total quarterly revenues as compared to 28.5% in the first quarter.” The wording matters because it pins HBM inside a DRAM base that is expanding, not merely taking share inside a flat bucket. Within that, HBM revenues increased from $29.5 million to $37 million, which gives the AI memory exposure enough scale to affect the quarter. The catch is that foundry and logic were still larger at $100 million, so even a visible HBM step cannot dominate company economics when the next guide explicitly calls for foundry and logic to fall.

The capacity story explains why the revenue guide is less bullish than it looks, because FormFactor is choosing to carry more infrastructure before the mix model has fully arrived. Management’s Q3 outlook calls for higher systems and DRAM, including HBM, but lower foundry and logic. That is a mixed signal, not a demand acceleration signal. Shahar put the cost commitment plainly: “At the midpoint of these outlook ranges, we expect Q3 operating expenses to be $55 million, plus or minus $2 million, approximately $2.5 million higher than Q2, mainly due to additional headcount and a full quarter of expenses related to operating our new manufacturing facility in Farmers Branch.” This quote earns attention because it commits to a near-term expense increase while the largest probe card category is guided down. If the Street was underwriting HBM as incremental leverage, the guide instead says FormFactor is absorbing factory readiness costs into an uneven demand mix.

The financial trajectory reinforces that interpretation: revenue has recovered from the Q1 FY2025 trough, but gross margin has not followed the shape investors would want. Q2 FY2025 revenue was $195.8 million after a +14.3% sequential rebound, yet gross margin was 37.2%, well below the 44.0% reached in Q2 FY2024. That is the core contradiction in the print. The top line is no longer the bear case, but the margin line still behaves like a company whose mix and absorption are not normalized. The company’s own long-term model sharpens the issue rather than resolving it: Michael D. Slessor said the target financial model “delivers 47% gross margin on $850 million of annual revenue.” On the current quarter, the business is near the revenue run-rate discussion investors care about, but Craig Andrew Ellis framed the gap as “700 basis points away on gross margin,” which makes margin the gating item for multiple expansion.

The Street comparison makes the surprise cleaner. Priced in was a quarter close enough to expectations to preserve confidence in the recovery narrative; actual revenue was only $195.8 million against $199.4 million, so the top-line miss alone should not have changed the debate. The EPS miss did. At $0.27 versus $0.30, the market learned that mix, tax, spending, and absorption are consuming more of the HBM benefit than a revenue-only read suggested. Management also raised the expected effective tax rate for the full year to 19% to 23% from 14% to 18%, which helps explain why the EPS line was more fragile than the revenue line. The market may be mispricing the print if it treats Q3 revenue of $200 million plus or minus $5 million as confirmation of a smooth recovery, because the cost base and tax line now require a higher quality of revenue to produce the earnings the Street expected.

This is why the Farmers Branch purchase is strategically defensible but tactically expensive for the equity. Free cash flow was pressured by CapEx that was $47.7 million higher than in Q1 due to the $55 million facility purchase, and the company lifted expected annual CapEx for 2025 to $110 million to $130 million from $35 million to $45 million. Those numbers do not invalidate the long-term HBM thesis; they make the payback window the central stock debate. A higher factory footprint can make sense if FormFactor is preparing for larger, more complex probe card demand in DRAM and advanced packaging-adjacent applications. But until gross margin moves toward the target framework, the spending reads as a call option funded by current shareholders rather than as proven operating leverage. The buyback is not a meaningful offset yet: $72.6 million remained available under the $75 million 2-year program, which tells us capital return has not absorbed the cash-flow shock.

The customer read-through is specific and uneven. SK Hynix is the cleanest positive read-through because DRAM rose to $57.1 million and HBM increased to $37 million, implying wafer-test demand for HBM probe cards is not merely a future talking point. Samsung also gets a constructive memory-test signal through DRAM probe card demand, though the data pack only identifies Samsung’s DRAM wafer test exposure and does not assign a customer-specific dollar amount. The negative offset is for TSMC and Intel, because management guided Q3 foundry and logic down after Q2 foundry and logic revenue reached $100 million. The magnitude matters: foundry and logic was still larger than DRAM by the company’s own disclosure, so a decline there can absorb a good portion of the HBM benefit. For portfolio managers holding across the semiconductor chain, this print supports HBM test intensity at SK Hynix more than it supports a broad logic wafer-test recovery at TSMC or Intel.

The competitive context also argues against paying for FormFactor as though it has already crossed into a higher-margin test franchise. The peer table shows several test and assembly comparables with gross margins far above FormFactor’s 37.2%, including 67.4% and 70.8% at the high end, while one closer peer sits at 38.0%. That spread is not a valuation footnote; it is the operating proof FormFactor still owes investors. If HBM probe cards carried the full margin power bulls assume, the company would not be discussing a bridge back to 47% while actual gross margin sits in the high-30s. The peer comparison does not say FormFactor lacks attractive exposure. It says the current print did not prove that exposure is translating into the margin profile investors associate with the better parts of semiconductor test.

The call delivery was more constructive than the numbers, and that gap is worth tracking rather than dismissing. In the tone history, Q2 FY2025 sentiment improved to 0.34 from 0.09, and guidance_tone rose to 0.21 from 0.02. Prepared_sentiment also moved to 0.41, while qa_sentiment was 0.24. That tells us management came prepared to sell a recovery narrative, but the Q&A did not become materially more enthusiastic than the script. The nuance is important: uncertainty stayed elevated at 52.8, and qa_evasiveness was 22.3, so the language model reads the call as more positive but not cleaner. In other words, management’s tone improved before the margin bridge was demonstrated.

That tone-versus-results gap is the best explanation for why this print is tradable but not de-risked. Management is pointing investors toward Q3 revenue of $200 million plus or minus $5 million, higher systems and DRAM including HBM, and a longer-term path to 47% gross margin. The actual quarter showed Q2 revenue of $195.8 million, gross margin of 37.2%, and EPS below the Street by -10.0%. Those facts can coexist, but they do not deserve the same multiple. The constructive case is that HBM demand is now visible enough to fund utilization and mix improvement as Farmers Branch ramps. The skeptical case is that foundry and logic cyclicality plus new fixed costs keep gross margin trapped below the target model for longer than investors expect. The print pushes us toward the skeptical side for now because the company has given firm cost numbers and directional mix caveats, while the margin recovery remains an aspiration.

The next quarter gives a clean scorecard. Confirmation requires Q3 revenue at or above the $200 million midpoint, with DRAM and systems increases large enough to offset the guided foundry and logic decline, and gross margin moving meaningfully above the 37.2% level reported in Q2 FY2025. The thesis breaks to the upside if operating expenses stay near $55 million while revenue lands at the high end of the $200 million plus or minus $5 million range and HBM exceeds the Q2 level of $37 million. It breaks to the downside if revenue falls below the low end of the guide, if gross margin fails to improve despite higher DRAM, or if the 2025 CapEx range of $110 million to $130 million rises again. The dates are straightforward: the next fiscal checkpoint is Q3 FY2025, and the debate will be decided less by the headline revenue print than by whether HBM finally produces margin leverage after the Farmers Branch cost step.