$FORM: A Full Factory Behind a Flat Guide

The 60 second version

• FormFactor ($FORM) makes probe cards, the custom consumable tools that test every chip while it is still on the wafer, and it leads in the advanced MEMS cards AI chips need.
• The shift: GPUs, custom ASICs and HBM stacks need more wafer test, finer probing and a fresh card for every design, and that demand has outrun FormFactor’s own factory.
• On our curve the AI probe card business is in Flight. The constraint easing is FormFactor’s capacity, as its Farmers Branch, Texas plant qualified in October 2026 and should add about $400 million a year of output by the end of 2027.
• The evidence: a fourth straight record quarter at $280.5 million, up 38.4%, GPU cards shipped in multimillion dollar volumes to the leading foundry, and $104.3 million of operating cash in one quarter.
• The next test is Q4 2026 results and the Q1 2027 outlook in early February 2027, where we want $290 million or more of revenue at an underlying gross margin of 53% or better.
• The thesis breaks if any of these happen: Q4 revenue lands below $275 million, the Q1 midpoint comes in below $265 million, margin is guided under 51%, or Farmers Branch volume production slips past Q1 2027.

On October 28 FormFactor reported the best quarter in its history, and it was the fourth record quarter running. Revenue was $280.5 million, up 38.4% on a year earlier. Gross margin before one off items reached 56.5%. The stock fell about 11% after hours anyway, and it is now down 41% in a month.

One number did the damage. Management guided Q4 to $285 million, only 2% above Q3, and a market already dumping chip stocks read that as the AI test cycle topping out.

We read the same call and reached a different conclusion, one the company stated plainly. The old factory is full. A few days before the report, a new plant in Farmers Branch, Texas passed qualification and shipped its first production probe card to a leading customer.

Our thesis, in one sentence: the flat Q4 guide reflects a capacity ceiling rather than a demand peak, Farmers Branch lifts that ceiling through 2027, and at about 21 times our central FY2027 earnings the price asks for very little of the growth the new capacity makes possible.

The shift

Wafer test used to be a quiet line item. A chipmaker probed each die, binned the bad ones and moved on, and the probe card was an ordinary purchase. AI changed the arithmetic. An accelerator now packages a huge logic die with stacks of HBM and sometimes several chiplets, so one bad die found after assembly can take everything bolted to it into the scrap bin.

That pushes test earlier and makes it denser. Cards need more contacts at finer pitch, run at higher speeds and cover more test steps per wafer, and every new design needs its own card. We put semiconductor test for AI parts well into its growth phase, with spending set by how many new designs reach production and how many wafers each one consumes.

For 2026 the binding constraint was the supply of advanced MEMS cards that can do this work. At FormFactor specifically, the limit was factory floor, and the company said so when it guided Q4 to 2% growth. The fix is concrete and dated. Farmers Branch passed qualification in October 2026 and shipped its first production card days before the Q3 report.

A second, smaller shift runs alongside. Co packaged optics, where the optical engine sits next to the switch or accelerator inside the package, needs its own test equipment. FormFactor’s TRITON system started 2026 with a $10 million to $20 million outlook. It now expects more than $30 million.

The company

FormFactor sells two things. Probe cards are the business, about $221 million of Q3’s $280.5 million. Each one is a precision instrument carrying thousands of microscopic MEMS probes, built for one chip design and one test step and replaced as the tips wear. The Systems segment sells probe stations and test systems, TRITON included, to chipmakers and research labs, and it posted a record $59.5 million in Q3, up 23% on the quarter.

The customers sort into two camps. Memory makers buy cards for HBM and ordinary DRAM. Foundries and chip designers buy them for GPUs, custom ASICs, data centre CPUs and mobile chips. Mike Slessor has been chief executive since 2016 and has taken the company from a memory heavy supplier to one holding a qualified GPU position at the leading foundry.

The detail we keep coming back to sits on the balance sheet. Deferred revenue rose from $24.0 million to $36.6 million in a single quarter, a jump of more than half. When customers pay ahead of delivery in a market that has supposedly just peaked, we want to know why.

A second oddity is a reason to read past the headline EPS. About $0.09 of Q3 earnings came from one off tariff refunds. Management stripped that out on its own and put the underlying gross margin at about 53% at Q3 volumes. We trust a company more when it tells you its own number is flattered. (A tariff refund lifting a chip tool maker’s margin is a very 2026 sentence, and we doubt it will be the last one we write.)

Capex is the third quirk. The 2026 plan fell to $100 million to $110 million from $140 million to $170 million, partly on timing and partly because suppliers offered better payment terms. That money moves into 2027. It has not been saved.

Where it sits on our curve

We place advanced MEMS probe cards for AI chips in Flight. Customers buy them in volume, buy them again, and in some cases pay ahead. Whether the product works is settled. What remains open is how fast it can be made, and we judge that on the product’s own evidence, separate from the broader test sector.

The buyer’s threshold is qualification, and it works like a switch. A foundry or memory maker will not put a card on a production prober until that card has passed its own qualification for that design, and until then orders are zero. FormFactor cleared the bar for a single GPU design at the leading foundry and then shipped cards in multimillion dollar volumes. On the memory side the threshold was crossed long ago, and DRAM revenue stayed near its record in Q3.

Our evidence sequence runs from demonstration to paid deployment, renewal, expansion and cash. The AI product sits at expansion. Foundry and Logic grew 10% quarter on quarter across several uses, the TRITON outlook rose to triple its original low end, and the commitment behind it shows up as $12.6 million of new deferred revenue. At company level the cash step is already arriving, with $104.3 million of operating cash in the quarter.

Is the change repeatable? Mostly, yes, because the card is a consumable tied to designs. Every new GPU generation, every custom ASIC tape out and every HBM node needs new cards, and volume wears out the old ones. The part that could be a temporary surge is HBM, which follows memory spending, and that is the part we watch most closely.

The numbers and what the street expects

Start with direction. FY2025 revenue was $785.0 million, up only 2.8%, according to the company’s Q4 2025 release. The first nine months of 2026 brought $764.9 million, and adding the Q4 guide puts the year near $1.05 billion, up about 34%. Growth has sped up through the year, and Q3 was up 38.4% year on year and 8.6% sequentially.

Margins tell the cleaner story. GAAP gross margin was 39.8% a year ago, 50.7% in Q2 and 54.7% in Q3. Before one off items the path ran 41.0%, 53.3% and 56.5%. Excluding the tariff refunds, management’s underlying figure is about 53%, and it expects to hold that level through 2027 after absorbing $25 million to $30 million of Farmers Branch startup costs in 2026.

FormFactor does not publish a backlog figure we would lean on. The best forward evidence is the deferred revenue line, which is cash already collected for product not yet shipped, along with the GPU qualification.

Cash against burn needs only a few sentences, because there is no burn. Q3 operating cash flow was $104.3 million and free cash flow was $64.8 million. Cash and marketable securities stood at $395.4 million on September 26, against about $11.4 million of debt. Diluted shares were 79.4 million, up from 77.7 million a year earlier. That 2.2% creep is modest for this sector.

Q4 guidance is $285 million of revenue, plus or minus $10 million, with EPS before one off items of $0.95, plus or minus $0.09, and a gross margin of 54%, plus or minus 1.5 points. Before the report, Zacks had consensus at $278.8 million and $0.80, so the guide came in about $6 million above on revenue and $0.15 above on EPS. Q3 itself beat consensus by $0.18 of EPS.

Revisions point up. Seven analysts raised estimates after the report, and MarketBeat recorded TD Cowen lifting its target to $110 from $100 while keeping a Hold rating. FY2026 consensus before the report was about $1.03 billion of revenue and $3.05 of EPS. The guide implies about $1.05 billion.

There are gaps in what we could confirm. We could not verify a clean analyst count or a high to low spread for FY2027. The aggregator figures we found predate the report and disagree with each other on EPS, so we left them out. Management gave no FY2027 revenue guide, only the 53% margin baseline, which means our own cases below carry the argument.

The capacity numbers need their own line. Probe cards brought in roughly $221 million in Q3, which by our arithmetic annualises to a little under $900 million. Farmers Branch should exit 2027 able to make about $400 million a year on top of that, roughly 45% more. That is how much the ceiling rises.

Why it wins

Follow the money. A chipmaker pays FormFactor per card, priced for one design and one test step, and pays again when the card wears out or the design changes. Revenue therefore tracks design count and wafer volume, and AI is pushing both up.

Each extra card brings costs: MEMS fabrication, assembly labour, and the yield loss that comes with probes this fine. Farmers Branch adds depreciation and the usual clumsiness of a new line, which management expects volume to absorb. The Q1 and Q2 2027 margins will show whether that claim holds.

For now the supplier keeps the benefit. Gross margin rose 15.5 points year on year. That is what pricing power looks like when capacity is short and the card is the cheapest way to avoid scrapping a very expensive package. Customers are paying for availability, and they are paying willingly.

Would a much better, much cheaper base AI model hurt? We think it would help. Cheaper models mean more inference, more chips and more custom silicon, and every one of those designs needs its own cards. FormFactor sits on the hardware side of that trade, where falling software costs turn into more wafers.

The moat has real strengths and one soft spot. The strengths start with the qualification at the leading foundry and the long held HBM and DRAM position. The expanded partnership with $TER Teradyne puts FormFactor alongside both of the big test equipment makers, and TRITON leads in co packaged optics test, helped by the Keystone Photonics acquisition.

The soft spot is that probe cards are made to order, and rivals in Asia and Europe make them too. Qualifications are won one design at a time and can be lost the same way. For 2027, at least, FormFactor has something we cannot see at the competitors we know of: about 45% more probe card capacity coming online in the United States.

What could go wrong

The strongest bear case runs like this. 2026 was a catch up year after a flat 2025, stacked on top of a peak in AI and HBM spending, and the flat Q4 guide is the first sign that demand is cresting. On this view, capacity is a convenient story for a management team that would rather not say orders are slowing.

We take that seriously, because for now the evidence fits both readings. A full factory and a peaking customer base produce the same Q4 number. The two separate in Q1 and Q2 2027. If revenue steps up as Farmers Branch ships, capacity was the limit. If revenue stays near $285 million or falls while the new plant is running, demand has turned.

Timing is the specific risk we fear. A pause in AI or HBM spending, or a turn in the memory cycle during the first half of 2027, would leave Farmers Branch underused just as its depreciation hits the income statement, and that is how gross margin falls below 51%. Memory has turned on investors faster than this before.

Concentration is the other honest worry. A single GPU design at a single foundry accounts for a meaningful share of the new growth, and a handful of memory makers dominate HBM. One delayed product could dent a quarter.

Financing is not the problem. With $395 million of cash, about $11 million of debt and $64.8 million of free cash flow in Q3 alone, FormFactor can pay for the capex rolling into 2027, reach the February test and collect the cash that follows without raising a dollar. If the bear case plays out, shareholders get hurt through margin and multiple. Dilution is not part of the risk.

A reader who believes 2026 was the peak has a coherent case, and we have priced it in our downside below. Where we differ is that the company has handed us a clean way to find out within two quarters.

The investment thesis

FormFactor is limited by supply rather than demand, Farmers Branch removes that limit through 2027 at a margin near 53%, and today’s price needs only about 5% revenue growth next year to hold.

The chain runs in order. First, the constraint is easing, since the plant qualified and shipped in October. Adoption is responding, with Foundry and Logic up 10% sequentially, GPU cards shipping in multimillion dollar volumes and TRITON heading past $30 million. The company is keeping the value, as a 15.5 point margin gain shows. Financing leaves that gain with shareholders, because the business pays for itself and the share count grew only 2.2%.

Now reverse the price. At $87.53 the market value is about $6.85 billion, and net of cash and debt the business is valued at roughly $6.46 billion. At 25 times forward earnings, which we treat as a fair multiple for chip equipment, the price needs about $3.50 of FY2027 EPS before one off items. With a 53% gross margin and about $270 million of operating costs, that takes about $1.10 billion of revenue, only 5% above 2026.

In our downside case, revenue stays flat at $1.05 billion, gross margin slips to 50% and EPS lands near $2.85, which at 18 times is worth about $51. Our central case has $1.23 billion of revenue, up 17%, at a 53% margin, for about $4.15 of EPS and a range of $87 to $104 at 21 to 25 times. In the upside, Farmers Branch fills, revenue reaches $1.40 billion at 54% and EPS comes to about $5.10, roughly $128 at 25 times.

The gap is plain. The evidence describes a company adding 45% more probe card capacity for customers who have already qualified its cards. The price assumes it barely grows. Our central case supports today’s price at the bottom of its multiple range, so the upside costs nothing at this price.

That is the asymmetry we look for.

The next test is the Q4 report and Q1 2027 outlook in early February 2027. The thesis strengthens if Q4 revenue reaches $290 million or more with an underlying gross margin of 53% or better excluding tariff refunds, if the Q1 midpoint is $290 million or more with management crediting Farmers Branch, and if Foundry and Logic grows again. Any one of these breaks it: Q4 revenue below $275 million, a Q1 midpoint below $265 million, gross margin guided below 51%, or Farmers Branch volume production slipping past Q1 2027.

What we are watching

• Early February 2027, Q4 2026 results: revenue of $290 million or more at an underlying gross margin of 53% or better strengthens the thesis. Revenue below $275 million breaks it.
• Early February 2027, Q1 2027 outlook: a revenue midpoint of $290 million or more, with Farmers Branch named as a contributor, strengthens the thesis. A midpoint below $265 million, or gross margin guided under 51%, breaks it.
• Early February 2027, full year TRITON result: anything above $30 million for 2026 confirms that co packaged optics test is a real line of business.
• Early February 2027, the 2027 capex plan: we expect the spending deferred from 2026 to appear. A plan far above the original $140 million to $170 million range without matching orders would worry us.
• By March 31, 2027, Farmers Branch volume production: confirmation that volume production is under way strengthens the thesis. Any slip past Q1 2027 breaks it.
• Late April or early May 2027, Q1 2027 results, if the usual timing holds: revenue above the Q4 level with Foundry and Logic up again would show capacity was the limit. Revenue at or below $285 million with the new plant running would tell us demand has peaked.

The bottom line

The shift is real. AI chips need more test, done earlier, on cards built for each design, and FormFactor leads in the cards that can do it. The market saw a flat Q4 guide during a brutal month for chip stocks and priced in a peak.

We think the flat guide comes from a full factory. The new one shipped its first production card in October and should add about 45% to probe card capacity by the end of 2027, while the price asks for 5% growth.

February will tell us which story is true.

The test that would change our mind is simple. If Farmers Branch is running and quarterly revenue still cannot climb above about $285 million, the factory was never the constraint, and we would treat the thesis as broken whatever the multiple says.

Not financial advice.

References

1. FormFactor, Inc. FormFactor, Inc. Reports 2026 Third Quarter Results. Exhibit 99.01 to current report. US Securities and Exchange Commission. Published October 28, 2026. Accessed October 31, 2026. https://www.sec.gov/Archives/edgar/data/0001039399/000103939926000040/ex9901-earningsreleasexq326.htm
2. Investing.com. Earnings call transcript: FormFactor beats Q3 2026 estimates, stock slips after hours. Investing.com. Published October 29, 2026. Accessed October 31, 2026. https://www.investing.com/news/transcripts/earnings-call-transcript-formfactor-beats-q3-2026-estimates-stock-slips-after-hours-93CH-4980992
3. FormFactor, Inc. FormFactor, Inc. Reports 2025 Fourth Quarter Results. Exhibit 99.01 to current report. US Securities and Exchange Commission. Published February 4, 2026. Accessed October 31, 2026. https://www.sec.gov/Archives/edgar/data/1039399/000103939926000004/ex9901-earningsreleasexq425.htm
4. Zacks Equity Research. FormFactor (FORM) Surpasses Q3 Earnings and Revenue Estimates. Zacks. Published October 28, 2026. Accessed October 31, 2026. https://www.zacks.com/stock/news/3017689/formfactor-form-surpasses-q3-earnings-and-revenue-estimates
5. MarketBeat. FormFactor (NASDAQ:FORM) Stock Price Expected to Rise, TD Cowen Analyst Says. MarketBeat. Published October 29, 2026. Accessed October 31, 2026. https://www.marketbeat.com/instant-alerts/analyst-formfactor-nasdaq-form-stock-price-expected-to-rise-td-cowen-analyst-says-2026-10-29/