Advantest Competitors

[Neomi Bensch]

Ido want to point out that this universe has some of the best performers amongst all semiconductor stocks in 2020, especially in 2H 2020. In other words, I am worried that this initiation is poorly timed. There are embedded expectations for a better future in advanced packaging related companies, evidenced by multiple expansion. Specifically, note NVMI, CAMT, TER, and FORM in the 1-year returns below:

In 2020, Teradyne took massive share relative to competitor Advantest. As such, in the near term, the company looks fundamentally well-positioned with solid revenue guidance and strong broad-based demand for their products. However, I am somewhat worried about 2H2020 comps because of how well they performed in 2020. And guidance can really only be given guidance quarter to quarter.

Test, which for years was stuck somewhere in the middle of the pack when it came to hardware engineering, suddenly became much more challenging over the past few years. Test budgets were increased, up from a flat 2% of manufacturing costs, and analytics were added into the mix in order to make sure chips would be able to withstand environmental, electrical and even unplanned events and continue working according to spec.

System Test and Semiconductor Test are closely related to each other. The Wireless Test segment is just a modern iteration of telecommunication related testing, an old deprecated segment. Industrial Automation, on the other hand, is a real de-novo business. Here is a graph of the segment by revenue percentage over time.

In any case, it seems to me that there actually are two schools of test demand, the initial ramp to volume manufacturing, and then the maintenance volume manufacturing that comes after volume production. The second category is a much larger dollar amount.

You might think that we've failed over the last 15 years, but over the last 15 years we've been able to reduce the cost of test from 2.5 percent of IC revenue to 1 percent of IC revenue.

So the key drivers are mostly around a variety of form factors more so than the volume of semiconductors, as well as the criticality of the type of device. Medical and Automotive have much stricter testing requirements than PC and mobile. Recently each quarter, the testing intensity seems to be bucking its 2012 - 2018 trend and has been rising significantly. A great example is below and is likely the single most important driver for Teradyne:

The principal driver of mobility test demand continues to be increases in complexity of cell phone silicon. This is especially notable in 2020 when smartphone unit shipments are expected to decline about 10% to 1.2 billion, yet the collective test intensity of each unit continues to grow at a rate in excess of this unit decline. Within smartphones, the mid-to-high tier is the place to be in test, and that's where Teradyne is solidly positioned. These phones are seeing disproportionate growth and complexity related to multiple high-density camera arrays, and the associated processing power and storage to manage this data.

Wireless Test likely becomes more important as UWB (ultra-wideband) and Wi-Fi become larger and more varied parts of IoT devices. Every package with RF has to make sure the signals work together, and thus the results must be tested. This is good for semiconductor testing intensity.

Universal competes against the likes of Fanuc and other famous robot arm companies (as well as Teradyne). Incidentally, unlike Fanuc, Universal Robots trades at a huge discount, exacerbated by the last 3 years of sluggish industrial automation. In any case, I think this could turn very shortly.

Most of what Teradyne does applies to Advantest as well. Note that Advantest has a very good IR presentation. Their documents provide much more clarity and color compared to Teradyne, which makes them one of the few cases where a Japanese company has better disclosure than a US-listed one!

Advantest actually breaks out memory and SOC (system on chip) exposures - along with other services and support segment items. Mechatronics is really cool but very immaterial - a good wiki and podcast here.

Their acquisition spree in the last two years also is meaningfully above average (and might explain the Capex gap) as acquiring new technology could make up for investing it yourself. The two most recent acquisitions are Astronics and Essai - both American companies that fit into their portfolio nicely as expansions of their ATE line.

Cohu is the distant third player in test behind Teradyne and Advantest. Cohu actually was the original Fairchild semiconductor test company that was the arch-rival of Teradyne back in the day. It got bought by Schlumberger during the conglomerate spree and was subsequently spun off in the early 2000s into its own company. There are some great slides from their deck.

This slide on testing intensity is helpful for understanding all 3 of the semiconductor testing companies, especially with regards to wireless test. The same drivers that contribute to mobile phone testing intensity right now will eventually drive 5G and IoT testing intensity in the future, specifically in the form of electronics comingled with Radio-Frequency devices.

Cohu is a much smaller and higher beta play on semicap. Their last year was one of the most interesting in the space. They were the clear testing winner in terms of price performance in 2020, but that was partially driven by how beaten down Cohu was at the beginning of the year. Cohu went to a single-digit P/E in March, had their numbers cut, and then proceeded to have a solid year and multiple expansion, exceeding dismal expectations. Quite the round trip.

Probe cards are an electronic interface that serves as an electrical path between a piece of silicon and a testing machine (like Teradyne or Advantest). Probe cards are custom made for each piece of silicon and have the important task of verifying if the piece of silicon is working as it should.

Since probe cards touch the silicon, they tend to accumulate debris on the tip of the needles and must be replaced every so often. To sum, probes are highly custom and consumable products that are replaced over time and are being used at higher rates due to more wafer-level package testing. A youtube video showing the process is below:

The current shift to advanced packaging will result in a large number of varied semiconductor patterns and increasingly complex structures, which means that probe cards will see greater demand (volume-wise) and become even more mission-critical. Probe cards should and will win market share in the coming years for test products. They are a perfect play on the advanced packaging cycle and have more direct exposure than the traditional testing companies. Management is doing a very good job of expressing this:

The opportunity for them is winning share through a focus on the steps in new processes. I think it will be difficult to win against the likes of KLAC, but they will likely gain share from tail donors.

Yet another premium price, especially recently, but it seems somewhat reasonable given strong historical growth and buyout candidacy. It would be nice to see them buy out a company, especially Camtek (logic outlined here).

First I want to direct you to my recent write-up on Camtek. We will see if the merger between them and Nova is still on during the fourth-quarter results, but make no mistake: I still like Camtek as a standalone. There has been no deal announcement, but the stock has worked all the same. Think of a buyout as a put for this well-positioned company.

The company insists on paying dividends, which is odd to me given their size and express intention to do some kind of M&A. This likely is driven by a desire to pay their parent company tax-free dividends in Israel.

Atlas System - Atlas is a traditional optical metrology tool that likely leans memory heavy. Interestingly, they use pretty advanced non-X-Ray optical technology, which is very helpful on the sub-scale (X-ray has trouble). The III system is for Gen 5/6 NAND, while Atlas V is focused on Gen 7/8 NAND and 1z-1yy DRAM and 3nm / GAA metrology.

The Impulse system is for CMP (Chemical Mechanical Polishing) process steps and is used in conjunction with other systems to verify the CMP process. The Impulse V system is an extension that is used for deposition, etch, and litho inspection as well.

Metapulse is for measuring layer thickness, especially in metal films. This is likely an advanced packaging type exposure, as more complex metal packages are delivered. There are a lot of specific applications so it looks like a broad tool.

NovusEdge looks to be an interesting tool that focuses on the edge, notch, and back sides of the semiconductor wafer. This is mostly pre-processing but post-doping (refer to the doping part in the lithography writeup). At this point, there are no patterns etched into the wafer, but the wafer needs to be measured to see if the layer was correctly applied.

NSX 330 System is another advanced packaging focused system with an emphasis on wafer-level metrology, micro bumps, RDL, TSV, and other processes. These are all discussed in the packaging primer and are all secular growth segments of WFE (wafer fab equipment).

Last but not least is the software, which management takes the time to highlight. I like thinking of it as comparable to enterprise software. Often it is not sold as independent units, but in tight cointegration with the tools. These are never broken out and sometimes can be bundled in services.

So first off, this template is not smart enough to appreciate organic revenue versus acquired and teasing out organic revenue out of the deal post-merge is pretty hard to do without a lot more work. Anyway, the good news is that we get to keep up with organic growth going forward!

In the near term, I think the testing and inspection segment of the market is running a bit hot and would like it to cool off for a bit before getting comfortable again (why yes I understand that these are famous last words).

3a8082e126