Description

Summary Thesis

Recommending shares of FormFactor, with a valuation of ~$16/share, vs. the current $12, or ~33% upside. The write-up below explains what they do, in addition to (obviously) the key points of the thesis.

In a nutshell, FORM provides products to semis manufacturers and are within the larger semicap universe, though FORM’s products are closer to consumables than capital equipment. FORM operates in a few end-markets: DRAM, SoC, RF, and NAND. All of these areas have solid long-term tailwinds and DRAM and NAND specifically have near-term call-options for FORM that aren’t being captured by consensus. Further, after acquiring CSCD in mid-2016, there seems to be an underestimation of what the new systems business can provide, in addition to good, old-fashioned M&A-based earnings accretion and cost synergies.

Categorically-speaking, this is the kind of idea where the call is that numbers are too low. If there is some multiple expansion that comes with that (there usually is), then that’s great, but its not really baked-in to the target. Also, I’m not approaching estimates from a super-conservative point of view, but rather what I think is reasonably within reach. In other words - there are call-options in each of the company’s primary businesses, which if only 1 or 2 hit, numbers will be too low (let alone if all hit).

Shares have had a nice run, which may put off some (in fairness, I too usually prefer companies that have sold off recently, vs. those at 52-week highs), though that in-and-of itself is not a gating factor, nor an argument against the thesis.

Company Overview

FormFactor (FORM) manufactures and sells probe cards and related probe systems to semiconductor manufacturers (both IDMs, like INTC and all the memory guys, and fabs, like TSMC). FORM specifically specializes in advanced probe cards.

Before providing a little more company background info, “what is a probe card”?

The semiconductor manufacturing process is generally divided between two areas: 1) front-end, which is process of fabricating chips from a blank wafer and includes lithography, etch, deposition, and other processes related to creating the chip; 2) back-end, which refers to all that happens to the wafer after the chips are created, such as dicing the wafer into individual chips, testing, assembly, and packaging. The last step of the “front-end” processes is to test the chips while still on the wafer. The testing equipment and the silicon wafer are connected via a probe card, which looks like a large, circular plate, roughly the size of a silicon wafer, with leads that attach to the chips on the wafer, essentially as the interface for the test equipment.

2 important things to understand about probe cards: 1) Any new/different/changed design of a chip requires a new probe card. This means not just with every node shrink, but also with any design tweak or new/changed chip architecture; and 2) commodity probe cards enable a tester to only test a few chips at a time, requiring several “touchdowns” (industry term) to test an entire wafer. Advanced probe cards - the ones FORM specializes in - are able to test many chips simultaneously, thus limiting the amount of touchdowns per wafer.

While probe cards aren’t pure consumables, they aren’t capital equipment either - semis manufacturing is all about speed/output. As a result, manufacturers typically maintain a target allocated time for testing. For example, if a single advanced probe card can test an entire wafer in 60 seconds, and the manufacturer is running at 100,000 wafer starts per month, a month’s-worth of wafers would take ~70 days to  test (100K minutes = 1,667 hours = 69.4 days). Of course, that would be preposterous. Rather, manufacturers budget a certain amount of time for probe-testing, say 2 days or a X amount of hours, and depending on the test time required per wafer (which itself depends on the complexity of a given chips design, which continues to rise), they run enough probe systems/probe cards to maintain that budgeted level of test time.

This makes probe cards similar to consumables, in that A) they are purchased/used in a fixed ratio relative to wafer starts and budgeted test time, so an X% increase in wafer starts drives a likewise X% increase in probe card purchases; and B) they change with each tweak in design, as opposed to capital equipment that can run several types of designs (either on its own or with a retro-fit, without ordering a new piece of equipment).

Last point I’ll make about probe cards - intuitively based on the above, probe card demand is mostly a function of 1) new chip design activity; and 2) high-volume manufacturing (HVM). Meaning, new designs require new probe cards, though an unramped design doesn’t require all that many cards, so demand comes from new designs that ramp to HVM.

Back to FORM - FORM is the market share leader in the overall probe card market, with its closest competitors being Micronics (public, Japanese), MPI (public, Taiwanese), and Technoprobe (private, Italian). FORM’s primary markets are SoC (mostly MPUs, MCUs, and RF chips), DRAM, and flash memory (NAND, NOR), in addition to probe systems (sold across all semis segments). Why those areas? Advanced probe cards are only needed in areas where the test times provided by commodity cards would be a material problem - i.e., complex chip designs with very high volume. That equates to high-end processors and DRAM. Up until recently, NAND wasn’t an area that required advanced probe cards, given the low-complexity of 2D NAND, as opposed to 3D NAND, which has provided FORM with a sizeable market opportunity (see more below).

FORM has a fairly concentrated customer base, with top-10 customers accounting for 62% of revenue, including INTC currently at ~30% of sales and Hynix, Samsung, and Micron collectively ~20% of sales.

FORM recently acquired Cascade Microtech (CSCD), which provided RF probe cards and the systems business. CSCD has significantly diversified the customer base and markets and increased FORM’s TAM. Insiders own ~1.2% of the company, with mostly mutual fund institutional ownership.

Investment Positives

• Market leader and purely high-end

• FORM is the leader in the probe card market, aided by the fact that advanced probe cards have much higher ASPs and margins than commodity cards

• Thus, FORM is both a technical leader and in the part of the market with the most margin protection

• Market share / competition varies by end-market

• DRAM is a duopoly with Micronics, together comprising >90% of the market; From talking to FORM’s mgmt, DRAM share goes from 40% to 60% depending on the quarter

• SoC doesn’t have a strong competitor to FORM. Technoprobe is the only one with a decent offering, though FORM doesn’t see any real competitors at INTC, which is the biggest player in the space. Competition from Technoprobe comes mostly on mobile application processors

• NAND doesn’t really have an advanced offering, given that there hasn’t been a need for one until 3D NAND

• Aligned with secular trends

• Simply put, chip designs continue to become more complex, both due to complicated node shrink, as well as architecture changes (e.g. 3D NAND)

• As chip design becomes more complex, the need for probe cards increases, in addition to related adjacencies for FORM to break into as a minimizer of testing times in complex environments

• Better-than-normal visibility re: INTC for next several quarters

• INTC is currently ramping its 14nm node, to be followed closely by 10nm

• The ramp of 14nm has necessitated significant probe card order by INTC, well above FORM’s normal capacity

• In fact, the demand spike blew up a quarter for FORM earlier in 2016, when they didn’t properly account for the demand spike and had to push rev into the following quarter due to capacity not being fully in-place

• In order to feel comfortable building out the excess capacity to provide INTC with the probe cards they are seeking, FORM received assurances from INTC for the level of demand they will be receiving from INTC for the next several quarters

• This provides FORM with better-than-normal visibility for 2017 in the SoC segment

• CSCD provides diversification and margin tailwind

• The biggest knock on FORM - like many in the semicap supplies space - has been customer and end-market concentration. DRAM (only 3 customers) was >40% of rev prior to 2016 and INTC was approaching 40% of rev in 2016 pre-acquisition of CSCD.

• CSCD provides diversification on several levels

• DRAM drops to ~20% of rev

• INTC becomes <30% of rev

• Probe systems becomes ~20% of rev

• Probe systems customers are fairly varied, selling to basically every fab in the market

• RF probe cards becomes ~15% of rev

• Probe systems runs at a higher GM% than legacy FORM, with GM% in the 50% range, vs. probe cards in the low-to-mid 30s

Why Numbers are Too Low

The real crux of the thesis is that consensus numbers appear too low and/or don’t account for several avenues of upside that, if any hit, will make current consensus too low.

• DRAM - As many following the industry for the last 2 years or so are aware, the DRAM market has gone through some tough times. There’s some pretty good threads on it in other posts on VIC (check out the many MU posts), but the short story is that there was a significant oversupply in DRAM caused by, among other things, the dramatic slowdown and decline in smartphones and PCs, respectively. As memory cycles go, oversupply led to a dramatic decline in both DRAM pricing and, consequently, DRAM production. While DRAM pricing doesn’t necessarily impact FORM directly, FORM got hit by 2 different issues in the DRAM market: 1) pricing slowed down production, which actually lengthened the design cycle, which is what drives FORM’s business; 2) DRAM manufacturers had trouble successfully ramping new DRAM designs on 20nm (both Hynix and Micron failed customer qualification tests and had issues achieving an HVM-ready design at 20nm). Implicitly, there is very little credit being given the FORM’s DRAM business for 2017 and beyond, which I think is just a lazy mistake from many who reflexively dismiss the idea of a strong DRAM business given how much of a headache its been for the last 18 months or so.

• DRAM has bottomed; this is not some bold statement, but rather based on pretty simple fundamental factors

• DRAM pricing has been steadily rising for months now

• DRAM suppliers and semicap vendors (AMAT, LRCX) have noted prep has begun for sub-20nm designs (1xnm, 1ynm, 1znm), to ramp over the next few years

• Both Hynix and Micron issues at 20nm are largely behind them

• 20nm to comprise >50% of Q4 DRAM production for both MU, Hynix

• Generally speaking, all the datapoints are pointing to a DRAM upswing over the next 24 months, maybe longer. There is an argument to be made for a longer bull cycle in DRAM than 24 months, but I’ll just save that has upside. For this to work, FORM just needs regular DRAM market dynamics to get back on track and everything from market pricing to manufacturer commentary to semicap commentary is pointing in that direction.

• In a more normalized environment, FORM’s annual DRAM opportunity is ~$100-150M (based on FORM’s DRAM biz ranging from $93M to $131M annually from From CY10-15), vs. the current $85-90M run-rate, which is roughly what sell-side guys are baking into their numbers.

• CSCD - consensus rev for the combined FORM-CSCD is materially lower than the consensus for each company individually before the acquisition. When compared to the S-4 estimates for CSCD, its WAY lower.

• While the S-4 estimates can be chalked up to an overly aggressive outlook not meant as a base-case view, its more difficult to explain why the combined company estimates are less than the sum of its parts.

• Asking mgmt didn’t help - they didn’t want to imply that numbers may be too low and only said that they feel consensus rev of $480M for 2017 is fair

• Speaking to sell-siders didn’t help much either. It seems an even split between guys just not giving enough credit to CSCD and guys effectively lowering legacy FORM numbers, due to the SoC miss from Q1 and just a generally negative outlook for DRAM. Either way, the sum of CSCD and FORM would yield rev above $500M, which is at least another $0.05 in a conservative scenario.

• NAND - for those who haven’t been following along with the NAND market, the inability to continue shrinking NAND chip sizes has brought a new chip architecture for NAND in order to keep scaling memory density (i.e., bits per chip), referred to as 3D NAND. Simply put, the chips are built with vertical layers (vs. the standard 2D/planar approach). This is a radically different approach to chip design.

• For FORM, this has brought NAND into their purview, given that 3D NAND is expected to materially hurt yields and test times for NAND suppliers, thus increasing the value prop of advanced probe cards in NAND.

• FORM mgmt believes rev from 3D NAND could reach $40-50M annually over next 2 years, based on 2 NAND customers (MU and Toshiba/WDC) reaching HVM in 3D NAND by the end of 2017

• Adding Samsung and/or Hynix could push that number to ~$100M annually

• This compares with FORM’s current ~$20M annual run-rate in NAND

• Skewed consensus - this isn’t an operational thing, but interesting to note that consensus is thrown off by the Citi analyst, who is so far below the rest of the Street that he’s moving the numbers materially. Consensus is at $0.82 for 2017 EPS, but without the Citi analyst at $0.52, EPS would be $0.87

Why This Opportunity Exists / Variant Perception

• DRAM blow-up of 2016 - the disastrous DRAM market of 2015-2016 has killed anything with any exposure to DRAM. Even though FORM’s DRAM exposure is only tangentially related to pricing (i.e., only insofar as it materially impacts production and design cycles), FORM’s DRAM business showed enough weakness to get rolled in with everything else DRAM-related. Many are still slow to come back to DRAM-related ideas, which I think is just lazy paranoia and provides an opportunity

• INTC mis-step - shares took a dive after an execution mis-step with INTC in Q1; FORM recovered all the business in the subsequent quarters (it was a capacity issue and the biz just got pushed out, not lost)

• Balance sheet - speaking in relative terms, FORM’s net debt position stands in contrast to most players in the semicap space. This also causes FORM’s EBITDA multiple to make it look more expensive. I view this as a red-herring - FORM is actively reducing its debt-load over the next 3-4 qtrs and should be in a net-cash position before the end of the year.

• Shares have already had a big run - I find this to be somewhat of an ad-hominem argument, though I can commiserate - I too prefer ideas that are at 52 week lows vs. 52 week highs. That said, the way I view it, the stock mostly took off following the close of the CSCD deal, which was extremely accretive (CSCD pre-synergy accretion estimated to be ~50-75%; in the first complete combined qtr, I estimate accretion to be ~50% based on disclosed data). Essentially, the stock has moved up commensurate to the immediate accretion from CSCD, which while excellent, is not the crux of the thesis (i.e., I don’t believe the run in shares is taking into account the multiple call options on numbers beyond the clear-as-day immediate pre-synergy accretion from CSCD).

• Bearish View

• Concentrated customer base

• Response - welcome to the land of semicap; everyone is concentrated

• Large DRAM exposure

• Response - now significantly less than pre-CSCD

• Poor balance sheet relative to most comps

• Response - not a long-term or structural issue; in the process of being rectified

• Short lead times / poor visibility

• Response - probably the most fair business criticism, but nothing new, plus the massive orders from INTC have provided better visibility than normal

Estimates / Valuation

Model/estimates below. Happy to answer questions re: my assumptions. A couple of easy ones though:

• Mostly run-rating Foundry/Logic from Q4 levels for 2017, inching up in 2018 (10nm and 7nm ramps)

• DRAM ramping up to normal and then peak-ish levels in 2017-2018

• Pretty conservative on NAND; if they add Samsung or Hynix, those numbers are too low

• OpEx I assume they can keep on the lower-end of the guided opex range ($35-37M/qtr), given the likelihood of outperforming on CSCD cost-outs

• Run-rating interest/other levels from Q4, stepping down slightly in 2018 (potential upside there if debt paydown has bigger impact than I’m modeling)

• Keeping tax rate flat from Q4 - with $300M NOLs, there shouldn’t be much change in taxes paid

• Modeling share creep

I value FORM at $16, which is 15x my EPS of $1.05, which is 30% upside from here. I’m not moving the current multiple. Comps trade in the mid-to-high teens P/E. The most direct comp - Micronics Japan (trades in Japan) trades at a much lower valuation, but for good reasons (not diversified, basically a DRAM pure-play, not as good margins, difficulty in taking out Japanese companies, etc.)

Breaking down the price target / upside, its all in the difference in estimates (i.e., my estimates are 30% higher than consensus). Of course, I’m not building in any real multiple expansion, which would be likely if they come anywhere close to my estimates, given there would be multiple beat-and-raise quarters along the way. In this sense, they could hit $16 even short of my EPS. With consensus in the $0.80-0.90 range, hitting a high-$0.90s EPS for the year should give them enough multiple lift from here to get there.

Risks

• Delay in INTC or TSMC 10nm ramps

• Any delays/mis-steps in new DRAM design ramps

• NAND story doesn’t play out (for any number of reasons)

• Hiccup in CSCD integration, mostly on the opex side (i.e., opex balloons or cost-outs stall)

Catalysts

• DRAM guys ramp 20nm and quickly move to 1xnm designs

• INTC and TSMC 10nm (and subsequently 7nm) ramps

• 3D NAND probe card demand takes off

• Gain Samsung and/or Hynix in NAND

• Outperform CSCD cost-out targets

• Consistent debt reduction

Accelerators

• Accretive M&A - mgmt has been open that they believe there are further M&A  opportunities out there and that’s the preferred usage of capital beyond debt reduction. Mgmt has ROI/accretion hurdles, so the hope would be any deal would be materially accretive

• Potential buyback - if no M&A comes about by the time target debt levels are reached, there’s the possibility of a buyback

Catalyst

• DRAM guys ramp 20nm and quickly move to 1xnm designs

• INTC and TSMC 10nm (and subsequently 7nm) ramps

• 3D NAND probe card demand takes off

• Gain Samsung and/or Hynix in NAND

• Outperform CSCD cost-out targets

• Consistent debt reduction