|Shares Out. (in M):||32||P/E||22||15|
|Market Cap (in $M):||373||P/FCF||0||0|
|Net Debt (in $M):||0||EBIT||0||0|
PDF Solutions (PDF) was founded in 1991 as a spin-off of Carnegie Mellon’s Sematech Center for Rapid Yield Learning by John Kibarian (CEO), Kimon Michaels (VP), and Tom Cobourn and is headquartered in San Jose, CA. PDF Solutions has had an uneventful history since its IPO in 2001. Early investors in PDF have been disappointed by the lack of performance over the company’s lengthy public history. However, we believe that investor fatigue and selling have masked an underlying business which is actually setting up to perform quite well. With the stock down over 52% since the recent high in late 2016, now may be an opportune time to take advantage of a perceived mispricing. PDF Solutions trades at a high multiple due to what we believe are some temporary factors, with over $3+ per share in cash (almost all of which is domiciled in the U.S.) and a multi-year contracted royalty stream of $4.00+ per share (estimated & discounted contracted royalties) in cash flow. Backing out the cash and discounted royalties we think further downside is minimal as expectations have come down significantly.
While the poor stock performance and management missteps have dampened some of the enthusiasm behind PDF Solutions, we think there are three fundamental catalysts presenting themselves that will be a turning point in the operating metrics. First, the 14-nanometer node volumes should start increasing in production throughout 2018; This volume ramp should be mostly incremental to the existing 28-nanometer volumes, and should start to outweigh any declines from 28-nm nodes. The second catalyst is the Design-for-inspection (DFI) venture, which has the potential to revolutionize a large portion of optical inspection and process control market. PDF Solutions has been investing in DFI since 2012, and has only had commercial viability since early 2017. The final catalyst is PDF’s high growth Exensio big data platform which is seeing better traction at over 130 customers across the supply chain.
Corporate Overview and Business Segments:
PDF Solutions offers capabilities that are designed to enable clients to lower costs of integrated circuit design and manufacturing, enhance time to market, and improve profitability by addressing design and manufacturing interactions from product design to initial process ramps to mature manufacturing operations. Simply put, PDF Solutions helps clients increase integrated circuit yields while lowering time to market, which in turn lowers overall production costs. This has become a secular theme within the semiconductor industry, given the challenging yield dynamics at sub 28nm non-planar structures.
PDF Solutions is comprised of two segments-- one being the core business, DFM, and the other being Design for Inspection (DFI). DFM is comprised of Design-to-silicon-yield (DTS) solutions which includes the engineering consulting business, Exensio big data platform, and the royalty business, Gainshare. The margin profile and general direction of the traditional core business has been largely dependent on Gainshare, which has around 100% flow through margin. These contracts typically average around five years in length with foundry customers for specific nodes.
As noted on recent conference calls, one area of notable growth has been the Exensio big data platform, which, as of last quarter was growing rapidly with very high margins (gross ~ 80% and Op. ~40%). This is important because the revenue comprising DFM has typically been a lower margin consulting business which in turn drives revenues for the high margin Gainshare business. While management does not explicitly break-out detailed Exensio margins, they have suggested it should have 40% adjusted operating margins over time.
Gainshare is what PDF calls a “variable fee revenue model” and is usually dependent on achieving certain milestones (yield targets) by a deadline. When PDF enters into a contract to provide yield improvement services, the contract usually includes two components. The first component is a fixed fee for performance of services delivered (on-site engineers, consulting, & analysis) to the customer over a specified time-frame, called Design-to-Silicon-Solutions. The second component involves a performance incentive, where the customer pays a contingent variable fee, Gainshare. Revenues from Gainshare performance incentives represent profit sharing and performance is contingent upon customers reaching pre-defined operational levels with established contracts. Gainshare performance incentives revenue is largely determined by wafer volumes at contracted facilities, and therefore dependent on various supply and demand factors. Gainshare royalties are key to PDF’s overall business model, as these are effectively full margin. An essential part of the bull thesis hinges on increased incremental Gainshare volumes at 14nm (and lower) making up for slowing volumes at the 28nm nodes. Management addressed some of these concerns around 14nm cannibalizing 28nm volumes, stating that volumes would not be materially impacted by facilities switching over from 28nm to 14nm. There has been a history at PDF of initial low margin business with Design-to-silicon-yield solutions for long-term success in generating high margin Gainshare royalties.
“Our long-term success is largely dependent upon our ability to structure our future customer contracts to include a larger Gainshare performance incentives component relative to the fixed fee component. We typically recognize the fixed fee component earlier than the Gainshare component so if we are successful in increasing the Gainshare component of our customer contracts, we will experience an adverse impact on our operating results in the short term as we reduce the fixed fee component. Due to acquisitions and expanded business strategies, the mix of elements in some of our contracts has changed recently and the relative importance of the software component in some of our contracts has increased. We have experienced, and may in the future experience, delays in the expected recognition of revenue associated with generally accepted accounting principles regarding the timing of revenue recognition in multi-element software arrangements, including the effect of acceptance criteria as a result of the change in our contracts. If we fail to meet contractual acceptance criteria on time or at all, the total revenues we receive under a contract could be delayed or decline. Further, if we mix term-based licenses with perpetual licenses, it will impact the timing of the recognition of revenue from that customer. In addition, by increasing the Gainshare or the software component, we may increase the variability or timing of recognition of our revenue, and therefore increase the risk that our total future revenues will be lower than expected and fluctuate significantly from period to period.” - 10-K
We think the real catalyst is when Gainshare royalties start to stabilize and inflection after trending lower over the past few quarters. There is large potential fixed cost leverage in Gainshare revenue, and should there be an improvement in volumes at newer nodes this would support incremental margins as 14nm ramps. Although, as of last quarter it looks like management underplayed the declines at 28nm, with 28nm volumes dwarfing 14nm gains. It is clear that there will be at least a few quarters of headwinds from 28nm, enough to drop 2018 consensus topline growth from around 12.5% to 9%.
Exensio Big Data Platform
PDF has built up a robust big data analytics platform which is becoming increasingly scalable, thereby leading to significantly higher margins over time than the remaining core segment of DFM. Exensio is PDF’s Big Data analytics platform, which analyzes and reports on data generated at various points within semiconductor manufacturing. Exensio currently has over twenty one thousand process tools tied into the big data analytics platform. This large and growing data set is helping PDF use its proprietary algorithms to find small leakages and quantify yield losses as low as tool sensor traces. While this has been a small contributor over the past few years, Exensio is the fastest growing business segment within PDF. Management does not explicitly break down Exensio revenues, but on a recent call, it did imply that Exensio was growing rapidly from what we believe is around a $10-15m base. Based on our discussions with management, we believe Exensio could potentially generate a significant portion of total revenues by EOY 2019. PDF recently bulked up its software solutions business with recent bolt-on acquisitions in 2015 of Syntricity for $5.2 million and Salland Engineering. While PDF has not been particularly acquisitive, we continue to believe that it may utilize excess cash to make synergistic software acquisitions to bolster its portfolio. Exensio per management’s last investor day had a TAM of $110mn to $500mn.
“Exensio now has 130 customers including many fabs, IDMs, travels, system companies, equipment companies and OSATs. Customers tell us a select Exensio because it links data from wafer fab, wafer stored assembly in final test. It delivers machine learning capability customized semi-conductor manufacturing and because many in the industry use Exensio, it is easy to collaborate with your OSAT foundry design partner, if your team also uses Exensio.” –John Kibarian Q4 2017 CC
As noted earlier, we think Exensio is providing a margin uplift to the DFM segment which has historically been margin constrained by incremental headcount additions. While this is a slow transition, focusing more on SaaS solutions is fundamentally changing the business toward a more scalable, and diversified cash flow generative operation.
Management on the Q4 2017 call mentioned Exensio should have 40% adjusted operating margins, and continue to scale up over time. Management also noted that they are incrementally more bullish on Exensio as it is now becoming a marketing tool for fabless customers looking for big data and analytics.
Design for Inspection (DFI)
DFI is the patented on-chip inspection business that only recently had commercial viability. PDF Solutions started investing in DFI in 2012, with the annual investment consisting of approximately 5% of revenues. Management mentioned that investment in DFI (Eprobe 150 and 250) should continue to increase as a percentage of sales as they invest more aggressively to have commercial viability for the faster commercial use Eprobe 250. To date, this business segment has generated minimal revenues; however, we think there is a high degree of potential with DFI that is currently not being priced into the stock. As of last quarter there were over 100 chips taped out with DFI content on them, these include full production designs and R&D test chips. DFI is a patented technology which utilizes the dummy space (empty space) on chips which electronically tests for defects missed by optical inspection. These on-chip inspectors can electronically test billions of structures that optical inspection has difficulty with as complexity increases exponentially. This has become a recent trend in semiconductor manufacturing as fabs and foundries move from planar 28nm nodes to more complex 14nm and sub 14nm nodes.
“Now, in logic and foundry, with the introduction of the new 3D gate architectures, the yield issues our customers are grappling with today are proving to be the most challenging that the industry has ever faced and even the smallest variation in process margin can cause significant yield losses for these devices,” -Rick Wallace, KLA-Tencor Source: Lapedus M. (2014)
“Poor yields have been a major issues which have affected the production of chips with smaller process nodes. Intel Broadwell was the first chip that was affected due to poor yields since it was making use of a new 14nm process node and it came two years after the launch of Haswell (Haswell Desktop came in 2013 and Broadwell desktop came in 2015). Intel’s Skylake made use of a similar yet refined 14nm process node while featuring a better core architecture. The Skylake processors have been available in the market since 1st August and there are still several chips that have yet to make their way to the market in 2015 and 2016. With the process nodes becoming complex day by day and yields being a major issue, Intel issued their plans to move towards a totally new, 2.5 year ‘tick tock tock’ cadence.”
Source: Mujtaba H. (2015)
Consistent with the trend since 2010, we believe that the largest logic foundries will continue to increase their investment in leading edge nodes and capacity in 2015. Leading foundries continue to invest in new technologies such as multi-patterned lithography and 3-D transistor architecture, which has resulted in an increase in our business, and improved results of operations in the past few years.
Generally, the demand for consumer electronics and communications devices continues to drive technological innovation in the semiconductor industry as the need for products with greater performance, lower power consumption, reduced costs and smaller size continues to grow with each new product generation. In addition, advances in computing systems and mobile devices have fueled demand for higher capacity memory chips. To meet these demands, IC manufacturers and designers are constantly challenged to improve the overall performance of their ICs by designing and manufacturing ICs with more embedded applications to create greater functionality while lowering cost per transistor. As a result, both logic and memory manufacturers have migrated to more and more advanced manufacturing nodes, capable of integrating more devices with higher performance, higher density, and lower power. As this trend continues, companies will continually be challenged to improve process capabilities to optimally produce ICs with minimal random and systematic yield loss, which is driven by the lack of compatibility between the design and its respective manufacturing process. We believe that as volume production of deep submicron ICs continues to grow, the difficulties of integrating IC designs with their respective processes and ramping new manufacturing processes will create a greater need for products and services that address the yield loss and escalating cost issues the semiconductor industry is facing today and will face in the future. “ Source: 10-K
The complexity of integrated circuits continues to evolve over time with increased pressures on time to market, smaller sizes, lower costs, greater performance, and lower power consumption; we see a secular trend emerging within semiconductor testing and process control. PDF is well positioned to capitalize on yield-learning trends (Exensio) as well as electrical detection through on-chip inspectors (DFI). After discussions with sell-side analysts and PDF’s management team, we still believe the TAM for both of these emerging technologies is over $1bn.
Customer Concentration Risk
While the PDF story sounds great and holds much potential, there is one snag: large customer concentration at the major foundries. The massive customer concentration as presented in the 10-K comes off as a bit alarming. The top two customers for PDF were Global Foundries, and Samsung. GlobalFoundries, and Samsung represent 44%, and 11% of sales, respectively. Although over time this has been getting more diversified on the product level as other Exensio and DFI drive sales.
We also think that PDF has a good chance to further diversifying its customer base by securing additional business at TSMC, which has historically been a negligible customer for PDF. Based on our discussions with management, we felt that TSMC was potentially testing DFI (Eprobe 150à250). We were also encouraged when management mentioned that every tool at TSMC is connected to the Exensio big data platform. Therefore, it would only seem like a logical progression that if TSMC were to proceed with an on-chip inspector, PDF would be providing that solution.
Though much of PDF’s future intrinsic value is contingent on the success of forward looking developments with Exensio and DFI, we think the current fundamentals alone merit some consideration as the downside seems fairly protected. While PDF has recently been under earning as large investments in both Exensio and DFI have added considerably to the expense base, we think over time R&D and SG&A should mean revert. For instance, compared to 2013, LTM R&D expense (~30% of total sales) is up over 125%, and SG&A is up around 640 basis points during that time frame. While it may take a few quarters for costs to trend down, we believe there is likely an incremental $10m of associated annual costs due to just DFI and Exensio.
"By the second half of the year as we complete initial development of eProbe 250, we will reduce our R&D spending with third-party suppliers. By the middle of 2018 incremental spending on DFI will track more with revenues as the fundamental development should be complete." --John Kibarian Q4 '17
Bear Case- $7.20
We assume there is no top-line growth and revenues remain stagnant ~ $102m for 2018 and decline thereafter. We account for cash at the end of year of $101.3m, plus the discounted value of the royalty stream ~$128m. Being conservative, we attribute no value for the IP portfolio (71 U.S. patents), staff, and PP&E, which in theory would have substantial value from an acquirer’s perspective. We also assume DFI is a complete bust, and attributes no value to the overall business. Our bear case only values the cash and discounted royalty cash flow in run-off mode. Based on weighted average fully diluted share count of 32m, we arrive at a downside case of roughly $7.20 ($128+$102/32).
Base Case- $16.80-19.20
We think PDF is able to grow top-line at 9% (consensus) in 2018, and 16.4% in 2019, inclusive of moderate growth of the DFI platform. With Gainshare royalty revenues increasing (14nm & 28nm showing some stabalization) in 2018 we assume that EBITDA margins can improve back to 30% levels as more fixed cost leverage is realized and expenses come under control. PDF is expected to earn around $0.80 in EPS in 2019, and should do roughly $1.20 in EPS in 2020, assuming a reasonable multiple of 14-16x arrives at $16.80-19.20 in 2020. We think this valuation is conservative, especially when accounting for the excess cash and significant contracted cash flows.
Bull Case- $36
While it is hard to forecast how DFI and Exensio will pan out in the next few years, it appears likely that secular trends should keep the tailwinds at PDF’s back for the foreseeable future. We think there is the potential for the high margin royalty revenues to increase substantially from current levels. It seems plausible that PDF could earn $2.00+ per share (2020) in EPS due to higher royalty payments, fixed cost leverage enhancement, and DFI becoming a meaningful royalty contributor. We would attribute a higher multiple 18x EPS ‘20, dependent on the success and growth path of DFI and Exensio.
In our view, PDF will eventually be a margin expansion story with reasonable top-line growth after a multi-year investment phase. What we think investors are missing is that 2H 2018 should be an inflection point for the start of a multi-year growth phase. There are multiple catalysts over the next 18 months to support this claim such as continued Exensio growth, 14nm incremental royalties, possible 28nm stabilization, and the commercialization of DFI with the first Eprobe 250 release. Another potential catalyst is increased buybacks; management has done some marginal buybacks over the years but hasn’t fully utilized its overcapitalized balance sheet. We assume there is likely to be some EPS uplift (3-5%) due to management buying back stock more aggressively over the next few quarters. With the stock down over 52% from Q4 '16 due to temporary headwinds at the core business, we think investors are underpricing two big future catalysts, DFI and Exensio.
Disclaimer: This report is intended for informational purposes only and you, the reader, should not make any financial, investment, or trading decisions based upon the author's commentary. Although the information set forth above has been obtained or derived from sources believed to be reliable, the author does not make any representation or warranty, express or implied, as to the information's accuracy or completeness, nor does the author recommend that the above information serve as the basis of any investment decision. Before investing in a security, readers should carefully consider their financial positions and risk tolerances to determine if such a stock selection is appropriate.
At any time, the author of this report may trade in or out of any securities that are mentioned in the report as long or short positions in his own personal portfolio or in client portfolios that he manages without disclosing this information. At the time this report was published, the author had a long position in PDFS either in his personal account or in accounts that he managed for others.
· Eprobe 250 sees more commercial success and deployment in 2018.
· Continued platform growth within Exensio and industry wide adoption.
· China continues to be an area of growth and opportunity as the government mandated semiconductor build out continues.
· Stabilization is seen on 28nm volumes and 14nm volumes more than offset declines.