FormFactor (FORM): The AI Test Cycle Collides With an Operating Reset

The semiconductor test market is fragmented but the advanced end is consolidating around a handful of suppliers with real scale in probe cards. FormFactor's primary probe card competitors include Japan Electronic Materials Corporation, Korea Instrument Co., Micronics Japan, MPI Corporation, STAr Technologies, and Technoprobe, and several of them, most notably JEM and Technoprobe, build MEMS probe cards with the same underlying architecture as FormFactor's. What separates the field is not the technology, which is shared, but the breadth of the customer base across advanced packaging nodes, the speed of custom design turnaround, and the manufacturing footprint. The relevant peer set for the Systems segment, which covers probe stations and thermal and cryogenic systems, runs against MPI, Tokyo Electron, and Wentworth Laboratories in the probe station market, and against smaller specialists in thermal and cryogenics.

The strategic repositioning of the last two years has been a deliberate shift away from the cyclical foundry and logic base that carried the company through the client PC and server downcycle of 2024, and toward the two growth streams that are now driving the print. High-bandwidth memory, the stacked-die DRAM used in AI accelerators, has become the dominant DRAM demand driver, and it is the single most important variable in the Probe Cards segment. Co-packaged optics, the technology that embeds optical interconnect directly into the processor package to cut data center power consumption, is the load-bearing variable in the Systems segment. The December 2025 acquisition of Keystone Photonics for total consideration of $20.6 million was a strategic bet that the CPO transition from lab to volume production creates a new test equipment category in which FormFactor could be first mover, not a late entrant.

The mix is shifting, and the geographic mix tells the same story. In the first half, Foundry and Logic was 48.1 percent of revenue and DRAM was 34.7 percent, while the Flash line that had been a small but steady part of the portfolio began to contract as the Baldwin Park plant that served it closed under the restructuring. South Korea, home to the two major HBM manufacturers, was 31.8 percent of first-half revenue, and Taiwan, home to the leading foundry and now also the home of the Triton CPO test line, was 34.1 percent. China, which had been a meaningful revenue contributor in the mid-2020s, had fallen to 5.8 percent of first-half revenue, a direct consequence of export controls on advanced U.S. semiconductor technology to China.

The restructuring and expansion are the two strategic decisions that define the current operating environment, and they are linked. The January 2026 restructuring plan consolidated manufacturing from Carlsbad and Baldwin Park, California into existing sites, with Baldwin Park producing through the close of that month and Carlsbad through December 2026, to align the cost structure with the new target financial model. The Farmers Branch, Texas facility, purchased in June 2025 with 50,000 square feet of existing clean room space, is the long-term answer to the capacity constraint. It is expected to ramp to initial target production levels over the course of fiscal 2027. The principal structural risk is customer concentration, because SK hynix was 19.2 percent of fiscal 2025 revenue and the two largest customers combined for a majority share in individual quarters of the last two years, a natural consequence of a business model in which probe cards are design-specific and the customer base for advanced packaging is small. A single customer's production schedule, a design change, or a sourcing decision at one of the top two customers can move quarterly revenue by a meaningful percentage in either direction, and the thesis depends on the depth of the HBM build-out continuing to expand, not just holding.

The probe card is the product at the center of the business, and the technical barrier is real even though it is not a patent moat in the conventional sense. FormFactor's MEMS probe cards use micro-electromechanical systems technology to manufacture composite contact elements with characteristic length scales of a few microns, and the manufacturing process spans clean room environments up to ISO Class 5. The cards are custom-designed for each customer's wafer and chip layout, with over 150,000 contact elements possible at spacings as small as 40 microns across a full 300mm wafer. The customization requirement is itself the moat, because each card is specific to a customer's chip design and test requirements, and the design-to-manufacturing loop requires intimate knowledge of the customer's test methodology.

The high-frequency capability is the specific technical edge that has made FormFactor the preferred supplier in HBM testing. HBM devices are stacks of eight, twelve, or even sixteen individual DRAM die assembled with through-silicon-vias and thermo-compression bonding, and they require probe cards that maintain signal integrity at high frequencies while making reliable contact to micro-bumps and copper pillars. FormFactor's millimeter-wave capability, which extends to frequencies exceeding 80 GHz in the current product line, is the basis for the company's favorable competitive position in HBM testing. The 422 research and development employees at year-end 2025 represent the standing investment in keeping that edge.

The Keystone Photonics acquisition adds a second technical capability that the company did not previously hold in an in-house form. Keystone provides optical probing technology for the testing of silicon photonics and co-packaged optics devices, and the acquisition extends FormFactor's integrated test system leadership into the optical domain. The strategic logic is that as silicon photonics matures into high-volume production, the test requirements span both electrical and optical domains, and a supplier that can address both in a single test solution holds a structural advantage over suppliers that can only address one. The Triton, the company's high-volume CPO testing solution, is the product that embodies this combined capability, and it is the driver of the Systems segment's growth in the first half of fiscal 2026, with legacy Systems product revenue declining as capacity and customer demand shifted toward the Triton platform.

The manufacturing footprint is part of the moat and the part most exposed to execution risk, with 2,153 employees at year-end 2025 distributed across North America, Asia, and Europe, and nearly all manufacturing employees located in California, Oregon, and Germany. The 50,000 square feet of existing clean room space at Farmers Branch, acquired rather than built from the ground up, is a practical advantage in a market where clean room construction is slow and expensive, and the $18 million California Competes Grant on the balance sheet reflects the strategic value the state places on the company's manufacturing presence. The services and application engineering business, embedded in the product line rather than reported as a separate segment, is the relationship layer that makes the design-specific model work, and the on-site probe card maintenance and service training create a switching cost that is not captured in any financial statement. The consequence for a customer switching probe card suppliers is not just a new card order, it is a re-qualification of the test process, which for a production wafer line is a multi-week disruption, and this is the practical moat that no competitor can replicate with a better product, because it is earned through the service relationship rather than the technology.

The second quarter was the strongest in the company's history on every load-bearing metric. Revenue of $258.2 million was up 31.9 percent year over year. GAAP net income of $56.2 million, or $0.71 per diluted share. That was a multiple of the prior-year figure. The non-GAAP figures tell a cleaner story of the underlying trajectory: non-GAAP net income of $65.0 million, or $0.82 per share. That compared with $21.2 million, or roughly a third of the current level, in the prior-year quarter. CEO Mike Slessor summarized the four-quarter trajectory in the announcement: revenue up substantially, non-GAAP gross margin up by a wide margin, and earnings per share tripled. The magnitude of the margin expansion is the defining feature of the print, and it requires decomposition to separate the structural from the cyclical.

The gross margin walk from 37.3 percent a year ago to 50.7 percent in the quarter is composed of three distinct drivers. The first is mix: DRAM revenue, which is lower margin than Foundry and Logic in general, rose from 29.1 percent of quarterly revenue a year ago to 32.9 percent, which would normally compress margin, so the mix effect here is actually negative and the reported expansion is stronger than the mix shift would suggest. The second is utilization: higher factory loading spreads fixed manufacturing overhead across more output, and the disclosure attributes the margin improvement explicitly to favorable product mix and higher factory utilization. The third is the restructuring, which is the most important driver for the forward thesis, because the restructuring plans generated $4.3 million of restructuring costs in the second quarter and $25.8 million in the first half, concentrated in cost of revenues as the California plants consolidate. The restructuring is a one-time cost event, but its benefit is a permanent step-down in the manufacturing cost base, which means the 50.7 percent GAAP and 53.3 percent non-GAAP gross margin levels are partially a restructuring benefit, not purely a demand-driven level. The falsification test is whether margin holds above the high-40s in the quarters after the restructuring charges roll off, because that is the point at which the underlying margin structure is revealed.

The segment breakdown shows where the growth is coming from and where it is not. Probe Cards revenue of $209.7 million was up 29.4 percent year over year. Foundry and Logic was up 22.4 percent on stronger demand for networking and high-performance compute microprocessor designs. DRAM was up 49.0 percent, with HBM accounting for the large majority of the year-over-year increase. Flash revenue of $2.9 million was down 48.4 percent, a decline that is partly structural because the Baldwin Park facility that served a portion of the Flash line closed under the restructuring, and management has explicitly stated that Flash is expected to represent a smaller percentage of the overall mix in future periods. Systems revenue of $48.5 million was up 44.1 percent, driven primarily by Triton sales, and the disclosure notes that legacy Systems product revenue was declining as capacity and customer demand shifted toward the Triton platform, a favorable signal for the CPO test line's share of the segment.

Cash flow is the part of the financial picture that has improved most dramatically relative to the prior year, and it matters most for the capital allocation decisions ahead. GAAP operating cash flow in the second quarter was $61.8 million, a sharp increase over the prior-year quarter. Free cash flow of $52.6 million was a sharp reversal from a prior-year deficit in the same quarter. First-half operating cash flow reflects both the net income improvement and the non-cash restructuring charges, which are added back in the cash flow statement. Cash and marketable securities of $345.6 million at quarter end, up meaningfully from the start of the fiscal year, leaves the company with a substantial liquidity cushion even while it funds the Farmers Branch ramp. The annual report estimates the full ramp at a substantial capital expenditure range. The company did not repurchase shares in the first half, prioritizing the Farmers Branch investment instead, and a substantial portion of the current buyback authorization remains available, a meaningful share return option that the company has explicitly deferred to fund the Texas expansion. The tax picture is also favorable and moving in the company's favor. The effective tax rate in the second quarter was low single digits, down from a year earlier, driven by the foreign-derived deduction eligible income deduction and stock-based compensation benefits. The prior fiscal year effective rate was mid-teens, so the tax rate is compressing toward the low-teens as U.S. taxable income grows and the FDDEI deduction becomes more material, and a lower tax rate on a growing pre-tax income base accelerates the EPS compounder. The balance sheet is clean, with the revolving credit facility undrawn, the building term loan carrying a low fixed effective rate under the interest rate swap, and total liabilities modest against the cash and marketable securities position. The gross liability for unrecognized tax benefits of $55.5 million is a standing item that has been on the balance sheet for years, and the timing of any payment is not estimable, and the company indefinitely reinvests a portion of foreign earnings, with repatriation a decision point only if U.S. capital needs outpace domestic cash generation.

The third quarter guide is the single most important forward signal, and it implies acceleration. Revenue is guided to $270 million with a modest variance band, which at the midpoint is a 4.7 percent sequential increase. GAAP gross margin is guided to 52.0 percent plus or minus 1.5 percent, implying a further sequential expansion from the already-elevated second quarter level. Non-GAAP gross margin is guided to 54.0 percent. Non-GAAP EPS is guided to $0.86 with a modest variance band. The guidance assumes consistent foreign currency rates, and the Q2 result already beat the high end of the prior outlook, a signal that management is guiding with conservatism rather than stretching. The risk embedded in the guide is that it assumes the HBM and Foundry and Logic demand environment holds at the elevated level, and that the Farmers Branch start-up costs, which were modest in the second quarter and larger in the first half, continue to be absorbed without compressing the guided margin.

The Farmers Branch ramp is the largest execution risk in the forward window. Production is expected to begin late in the fourth quarter of fiscal 2026, with a ramp to initial target production levels over the course of fiscal 2027. The factory start-up costs, which include consulting, compensation, utilities, taxes, and facility maintenance, are expected to be between $20.0 million and $25.0 million for the full year. That compares with $3.0 million in the prior fiscal year. The ramp is not just a timing question, it is a cost structure question, because the facility is in a lower-operating-cost region and is expected to support a more favorable cost structure overall once at target production levels. If the ramp slips, the start-up costs extend, the cost structure benefit is delayed, and the margin story that is now priced into the stock is pushed back by the same amount. The annual report estimates the total capital expenditure for the ramp at $140.0 million to $170.0 million, a significant deployment against the company's cash position. Management has explicitly stated that it paused share repurchases to fund the buildout.

The Carlsbad plant, which is expected to manufacture through December 2026, is the second transition risk. The consolidation of Carlsbad production into other sites happens at the same time as the Farmers Branch ramp begins, which means the company is running two manufacturing transitions simultaneously in the fourth quarter of fiscal 2026 and into the first half of fiscal 2027. The Baldwin Park closure in January 2026 set the pattern, and the Flash revenue decline of 48.4 percent year over year in the second quarter partly reflects the transition friction from that closure. The Carlsbad transition is likely to generate a similar, if not larger, amount of transition friction, and the start-up costs and any associated production disruption are the risk factors that the guide does not explicitly isolate. The Keystone Photonics integration is the smaller but strategically more important execution risk. The acquisition closed in December 2025 with purchase accounting that is still preliminary, with the developed technology intangible assigned a preliminary useful life of ten years and the goodwill allocated to the Systems reporting unit. The Triton line is the commercial expression of the combined electrical and optical test capability, and the 44.1 percent year-over-year growth in Systems revenue in the second quarter is the first real evidence that the CPO test line is converting design wins into volume orders. The risk is that the CPO transition in the data center is a technology adoption curve, and the pace at which co-packaged optics displaces traditional pluggable optics is a variable that the major hyperscaler customers control, not FormFactor. The multi-year partnership expansion with Keystone Microtech announced in the second quarter, which reinforces the company's presence in the Taiwan semiconductor ecosystem, is a positive signal for the Triton pipeline, but it is a partnership, not a long-term supply contract, and it does not lock in the CPO revenue stream.

The counterargument to the bullish reading is that the company is now priced for the HBM cycle to continue accelerating, and the forward window is exactly when that assumption is most exposed to a change in the demand environment. The HBM build-out is driven by a handful of AI accelerator customers, and the production schedules of those customers are the demand. If the AI infrastructure capex cycle, which has been the macro driver of HBM demand since 2024, moderates even modestly, the HBM revenue line that is up 49 percent year over year is the most exposed to a sequential decline. The margin story that depends on the volume level to support the 50-plus percent gross margin would compress alongside it. The company's own risk factors acknowledge that its revenues are difficult to forecast because it does not carry a sufficient backlog of unfilled orders to meet quarterly revenue targets at the beginning of a quarter, and that a substantial percentage of quarterly revenue depends on orders received and fulfilled within that quarter. This is a structural characteristic of the design-specific probe card business, and it means the forward visibility is shorter than the guide suggests.

The customer concentration risk is the most immediate and the least diversifiable. SK hynix at 19.2 percent of fiscal 2025 revenue creates a direct exposure to the production schedule of the largest HBM manufacturer. The combined top-two customer share has also exceeded 40 percent in individual quarters, extending the concentration risk across the top tier. A design change at one of those customers, a shift in the HBM stack configuration, or a sourcing decision that favors a competing probe card supplier on a specific generation would reduce revenue by a meaningful percentage in a single quarter, and the design-specific nature of the product means there is no inventory buffer to absorb the shock. The consequence for shareholders is that the earnings multiple is exposed to a binary event outside the company's control. The stock's drawdown from its 52-week high of $160.27 to the current level reflects some of this concentration risk being priced in. The 52-week low of $29.58 shows the full range of the cycle the stock has experienced.

The Farmers Branch execution risk is the second major downside scenario, and it is a risk that the company itself controls. If the ramp slips by two or three quarters, the start-up costs extend beyond the $20.0 million to $25.0 million full-year estimate, the cost structure benefit is delayed, and the margin trajectory that the guide implies is pushed back. The capital expenditure of $140.0 million to $170.0 million for the full ramp is a substantial deployment. While the $345.6 million cash position can absorb it, the extended start-up period compresses free cash flow and delays the point at which the Texas facility begins contributing to the margin story. The bear case is that the ramp is slow, the cost structure benefit is delayed by a full year, and the 50-plus percent gross margin level that the company reached in the second quarter proves to be a transitional level rather than a structural floor.

The HBM demand cycle risk is the macro-level scenario, and it is the one that would most significantly alter the valuation framework. The HBM build-out is driven by AI accelerator capex, which has been in an expansion phase since 2024, and the HBM revenue line that is up 49 percent year over year in the second quarter is the most exposed to a change in that cycle. The disclosure notes that HBM products accounted for approximately 75 percent of the year-over-year increase in DRAM revenue in the second quarter. That means the DRAM growth, which is 49 percent of the Probe Cards segment, is itself heavily concentrated in a single technology. If the AI infrastructure capex cycle moderates, the HBM line is the first to feel it, and the margin story that depends on volume levels to support the cost structure compresses. The bear case scenario is a 20 percent to 30 percent sequential decline in HBM revenue in a single quarter. Such a move would reduce quarterly revenue by roughly 8 percent to 12 percent and compress gross margin by a meaningful amount.

The competitive risk in the probe card market is real but has been partially mitigated by the HBM qualification advantage. The primary competitors, Japan Electronic Materials, Technoprobe, and Micronics Japan, all build MEMS probe cards, and the annual report identifies the high capital investment and customer evaluation time as the barriers to entry in MEMS probe cards, a true statement but one that understates the depth of the qualification advantage that FormFactor has built with the top two HBM customers. The consequence of a competitor winning a design win on a next-generation HBM stack is that the qualification cycle, measured in quarters, means the revenue loss is delayed but not eliminated. The CPO test line, by contrast, is a newer competitive field, and the Keystone Photonics acquisition was a defensive and offensive move to secure a first-mover position, but the field is not yet mature enough to establish a durable competitive moat, and the major CPO customers have not yet committed to a long-term supplier relationship. The geographic and trade risk is a standing factor that has been moving in the company's favor in the near term but is a structural uncertainty. China revenue has declined from 13.5 percent of prior-fiscal-year revenue to 5.8 percent of the first half, a direct consequence of export controls on advanced U.S. semiconductor technology to China. The annual report identifies this as a direct consequence of the trade restrictions. The geographic mix is now heavily concentrated in South Korea and Taiwan, which is a concentration in the two most geopolitically exposed regions for advanced semiconductor manufacturing. The tariff refund event is a small but notable positive: the U.S. Supreme Court ruled in February 2026 that the International Emergency Economic Powers Act tariffs were not authorized by statute. The company is in the process of receiving refunds through the Consolidated Administration and Processing of Entries system, with approximately $0.8 million received to date. An additional $7.0 million to $9.0 million is anticipated in the third quarter. The tariff refund is a one-time cash benefit that reduces the effective tax and tariff cost of the period, and it is not a recurring item that should be built into the forward earnings framework. The balance sheet risk is minimal in the current environment, with the revolving credit facility undrawn, the building term loan fixed at 2.75 percent effective, and the cash position substantial relative to the debt balance, and the risk materializes only if the Farmers Branch ramp requires more capital than the estimate, or if a competitive or demand shock requires a defensive capital deployment, which the company's own disclosures note it may fund through additional debt or a stock issuance, a dilution or leverage event that is not currently priced.

The current market capitalization of approximately $8.9 billion, at a share price of $113.53, sits at a multiple that reflects the company's transformation. The multiple marks the full repricing from the low-margin probe card supplier of 2024 to the high-margin AI test supplier of 2026. The multiple embeds a full repricing of the earnings power of the post-restructuring cost structure, which the prior-year financials did not support. Trailing twelve-month revenue of approximately $902 million implies a revenue multiple of roughly 9.8 times. That is at the high end of the company's historical range and above the norm for the broader semiconductor test equipment peer group, which typically trades at a lower multiple for mature, low-growth suppliers. The trailing twelve-month GAAP net income of approximately $115 million implies a price to earnings multiple of approximately 77 times. That trailing figure is a function of the prior-year base being depressed by the low-margin, low-volume quarter, and it overstates the true earnings multiple on a forward basis.

The forward framework is more useful. The third quarter guide implies a full-year fiscal 2026 revenue of approximately $1.04 billion on a linear extrapolation of the guide. That puts the market capitalization at a lower forward multiple than the trailing figure, a meaningful step down that reflects the earnings power the post-restructuring cost structure should deliver over the full cycle. The forward price to earnings multiple, using the sell-side consensus of approximately 33 times, is a more meaningful anchor because it reflects the expected earnings power of the post-restructuring cost structure. The non-GAAP earnings per share trajectory, from $0.27 a year earlier to $0.82 in the current quarter, implies that the company is compounding non-GAAP EPS at a rate well above the broader semiconductor capital equipment group. The forward multiple of approximately 33 times non-GAAP earnings is a premium to the peer group that reflects the AI-driven demand cycle.

The bear case assumes a 20 percent to 25 percent decline in HBM revenue over the next two quarters. That reduces full-year revenue to approximately $950 million and compresses non-GAAP earnings per share to the low $3 range. At a 25 times multiple, that implies a market capitalization of approximately $5.5 billion, or a share price in the low $70s. The bear case assumes the HBM cycle moderates, the Farmers Branch ramp is on schedule but the cost structure benefit is delayed, and the CPO test line does not accelerate as quickly as the current trajectory suggests. The share price in the low $70s represents a substantial decline from the current level, which is the magnitude of the downside scenario that the concentration risk and the demand cycle risk together imply. The base case assumes the HBM cycle continues at the current pace through fiscal 2026, the Farmers Branch ramp begins in the fourth quarter as scheduled, and the Keystone Photonics CPO line continues to grow at the current rate. Under that path, full-year non-GAAP earnings per share lands in a range in the low $3s, and the market capitalization at a moderate multiple is approximately $7.8 billion, roughly in line with the current level.

The bull case assumes the HBM cycle accelerates in the second half of fiscal 2026 on new AI accelerator designs, the Farmers Branch ramp is ahead of schedule, and the CPO test line wins a second major customer design in the data center. Under this scenario, fiscal 2026 non-GAAP earnings per share exceeds $3.60. At a 35 times multiple, the market capitalization is approximately $10.5 billion, or a share price in the mid $130s. The bull case requires the HBM cycle to extend beyond the current build-out and the CPO transition to accelerate faster than the current adoption curve, which is the scenario that the stock's 52-week high of $160.27 was pricing in before the recent pullback. The peer comparison for the valuation is the broader semiconductor capital equipment group, which includes companies with more diversified end-market exposure and a more mature earnings base, and FormFactor's current multiple is a premium to that group, a premium that is justified only if the HBM and CPO demand cycles extend through the fiscal 2027 ramp period. The valuation is not supported by the probe card business alone, which on a standalone basis trades at a lower multiple, and the premium is entirely a function of the AI-driven growth trajectory. The investment question is whether the demand cycle is long enough to support the multiple through the Farmers Branch ramp, and the next two to three quarters of HBM revenue and the Triton design win pipeline are the data that answer it. The base case is the one that the current share price is most closely aligned with, and it is the scenario that the third quarter guide is implicitly assuming, which means the stock is neither stretched nor cheap on the current evidence, but is pricing in a continued HBM build-out that has not yet been challenged by a sequential decline.

The second quarter revealed that FormFactor has completed a genuine operating transformation, and the margin structure that was built during the low-volume period of fiscal 2025 has now been validated at high volume. The 53.3 percent non-GAAP gross margin and the $61.8 million of quarterly operating cash flow are not transitional levels. They are the output of a cost structure that was deliberately rebuilt through the restructuring and the Farmers Branch investment, a transformation that took two years of deliberate capital deployment and workforce realignment. They represent the floor for the next phase of the HBM cycle, not a peak that the next downcycle would erode.

The central strategic initiatives are the HBM qualification advantage, the Farmers Branch capacity expansion, and the Keystone Photonics CPO test line. The HBM qualification advantage is the current earnings driver and the one most exposed to the demand cycle. The Farmers Branch expansion is the cost structure story that makes the margin level sustainable, and its ramp timing in the fourth quarter of fiscal 2026 is the single most important execution variable for the next two quarters. The CPO test line is the next growth engine, and the 44.1 percent Systems revenue growth and the Triton design wins are the first evidence that the silicon photonics transition is creating a real revenue stream, but the adoption curve is controlled by the hyperscaler customers and is the most uncertain of the three initiatives.

The variables that the next six to twelve months resolve are the HBM revenue trajectory against the current 49 percent year-over-year growth rate and the Farmers Branch ramp timing against the disclosed full-year start-up cost estimate. The CPO design win pipeline, the gross margin level after the restructuring charges roll off, and the customer concentration mix as the top two HBM customers' combined share moves with their respective production schedules complete the set. The fiscal 2026 outcome on these five variables determines whether the current multiple is supported or whether the stock re-rates toward the base case level, and the third quarter result, which the guide implies is another record quarter, is the first data point that confirms or challenges the bullish reading of the transformation.