Description

Advantest (TSE: 6857)

The semiconductor industry is volatile and cyclical, but certain sectors are predictable. There was a supply and logistics constraint due to covid, but the economy slowly recovered.

Semiconductor fabrication, assembly, testing and packaging, which costed billions are now commonly outsourced (OSAT). This allows chip manufacturers to focus on design and IP, so they can be less capital intensive and more resilient in downturns.   

Teradyne and Advantest are a duopoly (combined 60-75% of the market) for outsourced testing equipment with Cohu driving third wheel (5-15%). They all provide equipment to conduct tests for chips to reduce defects after the wafer is diced to ensure shorter design periods and higher production yields.

 

In my experience, Japanese companies have been known to keep sales domestic, hoard a crazy amount of cash for “safety”, and have a polite IR reception that lacks transparency. Advantest fits the stereotype with a strong net cash position with little debt, but they do deploy the cash for strategic acquisitions. Advantest also was once listed on the New York Stock Exchange and they have 88-90% of their total sales are international.

Advantest has an ROE of over 20%, with a goal of sustaining￥400bn (approx. USD 3B) in sales and would like to pay out dividends on a continuous basis.

 

 

Automatic Testing Equipment- operation and parts

Historically, manufacturers made their own semiconductor testing equipment, but it didn’t work out— if a semiconductor operated at 1 gigahertz, that operating speed for the testing equipment must be 30 to 50% faster at 1.3 or 1.5 gigahertz and developed in advance. In addition, tests for memory require hundreds of systems working simultaneously, which seems simple in principle, but the actual difficulty is insane.  

For example, a medium sized jet airliner has about 1 to 2 million parts, where as a semiconductor test system has 2-3 million parts. In comparison, a car only has 30,000 to 50,000 parts. With more computing power packed in smaller chips at an exponential rate— test systems will only require more parts— which is not feasible for chip manufacturers unless they outsource.

In a test, a signal which checks for disconnections and shorts is input to the semiconductor, and the signal output in return is compared with expected values— this is called a “parametric test.” Devices are then classified according to the test results in a “function test.” Testing classifies performance characteristics such as input and output voltages, current values, and timing. After the performance of the chip is assessed, the handler places each chip into a performance bin.

 

 

The illustration below shows the steps of semiconductor test.

 

 

 

 

Intro to Advantest and competitors Teradyne and Cohu

Teradyne and Advantest have a semiconductor ATE market share with over 30% each. Both companies have incredibly long histories— Teradyne was founded in the 1960s and Advantest in 1954.  In the 1990s, the industry was still fragmented, but has since consolidated with high barriers to entry.

 

Advantest has a broad portfolio of semiconductor equipment testing solutions, spanning a wide range of semiconductor product categories including memory, logic, discrete, analogue, and optical semiconductors. Most of the growth is driven by servers and mobile phone electronics.

 

 

 

 

 

 

 

Semiconductor test systems are roughly divided into testers for SoC “System on Chip” and for memory.

 

Memory Testing (mass production, vertically integrated)

Memory (split into dynamic random access memory DRAM and NAND flash) is dominated by Advantest, with a lion’s share of 40-57% the testing market, as they have one of the fastest devices in production today with parallel testing capabilities.

Both Japanese companies Advantest and Cohu started in the memory space as testing providers for both DRAM (dynamic random-access memory typically used for the data needed by a computer processor to function) and NVM, or non-volatile memory (does not require a continuous power to retain data in the device).

Testers are optimized for mass production of memory semiconductors, such as DRAM and NAND flash. In the memory sector, device types are less diverse, and production volumes are huge; customers seek testers capable of hundreds of devices at once.

One of Advantest’s biggest customers is Samsung, while Cohu is close with Intel. Every investment from Samsung provides Advantest with an opportunity to open another factory or increase R&D. Teradyne has TSMC and Apple as its largest customers and entered the logics testing market earlier than Advantest. Apple is a big customer for Teradyne— Apple is developing their 3nm chip for their next line of iPhone through TSMC.

 

 

Silicon on Chip (diverse supply chain and products)

Advantest's rapid test systems for DRAM memory provide high margins, but not as high as some products for Silicon on Chip (SoC). While Silicon on Chip was originally dominated by Teradyne, Advantest was able to enter and penetrate the SoC market thanks to its acquisition of Verigy (ranked 3^rd by size in the field at the time) in 2011, which is three to four times as large as the memory tester market.

With the acquisition of Verigy, the model V93000 tester became Advantest’s flagship SoC testing machine. Advantest's test equipment sales are now 80% SoC Testers and 20% Memory Testers (see below).

 

 

Advantest’s Silicon on Chip product— V93000 EXA scale test platform for high end devices

Advantest’s V93000 EXA Scale SoC Test Systems are designed with the latest generation of Advantest's test processors with 8 cores per chip, which targets advanced digital ICs up to the EXA scale with 256 channels on the Pin Scale 5000 digital card and XPS256 power supply cards which can be configurated to smaller physical systems, reducing infrastructure cost and floor-space requirements. The systems' new test heads incorporate Xtreme Link technology, and the EXA Scale that enable new test methodologies, lower cost-of-test and faster time-to-market.

 

SoC testers have brought up Advantest’s operating margins due to higher pricing. Sales of SoC testers have a higher gross and operating margin than memory products in general. As of 2018, Advantest’s operating margin (11%) was lower than Teradyne (23%) due to lower margin products in memory. By 2022, Advantest caught up due to SoC, with Advantest having operating margins reaching 30% while Teradyne was near 27% (see below).

Source: Annual Report

 

In the SoC business, the flexibility to apply test systems to a wide range of devices –logic (APU & GPUs), analogue, and RF devices— is a major factor to business success.

The SoC tester market will be driven by 5G base station-related business and investment; both Advantest and Teradyne now control around 75-85% of the market.

 

 

 

 

 

Advantest sticks to its knitting, Teradyne branches into industrial automation— Time will tell which strategy works

While competitor Teradyne is more diversified than Advantest, especially in terms of industrial automation, its multiple is higher, and their acquisition of the Danish robotic arm company Universal Robots in 2015 for USD 285 million net of cash was not for profitability, but for integration in all tiers of manufacturing— they are using their robotic arm to automate PCBA, and other automated testing (see below).

Source: Teradyne Company Presentation

Advantest has decided to stick to its core business, so its returns have continued to expand at an accelerated pace. Whether this strategy is short sighted or wise, only time will tell.

 

 

 

Total Addressable Market and Valuation (how much more can Advantest grow?)

 

SEMI, an association which provides market data, predicts semiconductor market will grow from 600B to 1 trillion by 2030. In 2020 the semiconductor testing equipment market is valued around 6-7 billion. This represents around 8-10% of the entire semiconductor manufacturing equipment market, which is forecasted to grow at a compounded annual growth rate of around 6-9% driven by 5G and data centres until 2030 (Global Market Insights).

 

There are roughly 1.5B smartphones manufactured per year which require complicated microprocessors that go obsolete every year, and 40-50% (600-700 million units) comes from 5G. Most of the growth in sales and price hikes for Advantest’s SoC tester sales outlook was due to 5G initial mass production of cutting-edge process nodes.

 

A smartphone has one main system-on-chip (SoC) module with one central microprocessor embedded into the SoC, which acts as the system's CPU. In phones, this is usually made up of ARM cores. ARM is a microchip architecture firm currently owned by Softbank (Nvidia tried to buy it) which is licensed for manufacturers to use. Its monopoly and dominance is similar to x86’s chip architecture which helped grant Microsoft and Intel a holy trinity of— software, chip, and chip architecture— a unsurmountable fortress. However, new open-source architectures such as RISC V has emerged which may challenge ARM in the future.

 

About 300 million of those smartphones comes from Samsung. Of the 300 million units in total, Samsung made about 240 million of them. The rest were made by ODMs such as Wingtech and Huaquin. Newer and better chips will process data in a world that transitions from 5G to 6G.

 

SoC, system on chip, will have many additional processing elements for things like power management, picture, and sound processing, and even for accelerating AI. The board also has a couple DSPs and microcontrollers to handle power related applications (dynamic voltage scaling) and contains several individual CPU cores (or application cores), a graphics processing unit (GPU), digital signal processors (DSPs), image signal processors (ISPs) and several fixed function decoding blocks which handle things like video/audio decoding, security, and encryption.

 

Advantest has the business of major FPGA (Field Programmable Gate Array) manufacturers. The major manufacturers of RF devices for base stations, transceivers, and baseband chips use Advantest and Teradyne’s products. At present, sharing data at the local level requires 10B transistors but in the future, 100B transistors will be the standard. Qualcomm, Qorvo (another great VIC article), Skyworks will be in phones for the front-end RF receivers. A typical smartphone has 6 to 10 RF components. The design requirements for testing are complex, and there is an extremely narrow go-to-market window with an incredibly fast volume ramp-up.

Assume 600B of semiconductor sales in 2022. Around 145B is expected to come from smartphones, with image sensors for phone receiving the greatest demand. 115B comes from personal computing and 100B comes from servers and data centers (see below).

 

Source: Statista

 

The greatest growth (see above) in this decade should come from servers, data centers (250B by 2030), and smartphones (213B by 2030).

 

 

SoC market size outpaces Memory Tester Market

 

The SoC semiconductor tester market, at approximately 4.3B, is three times larger than the memory semiconductor tester market (1.3B), with hundreds of customers, including fabless and OSAT companies. The reason why Advantest estimates a 4.3B market (see chart below) for SoC testers is mainly due to APUs (accelerated processor units), and high-performance computing (HPC) chips, which are poised to migrate to more advanced process nodes.

 

To give you a sense of that, here's some data (based on Advantest filings) for 2016-2021 comparing the market sizes of SoC and memory:

Source: Annual Report

 

From 2016 to 2021, SoC’s market size roughly doubled, while memory tester has almost tripled. This growth doesn't seem to explain the success SoC (MCUs, MPUs, GPUs, etc) chipmakers experienced.

 

It needs to be understood in the context of testing equipment instead of end market revenues. Each piece of the supply chain is different, and revenues might not always flow down— adoption and ramping up of advanced technology such as Extreme Ultraviolet lithography (EUV) starting 2019 with 5nm and 7nm nodes results in lower yields/deterioration in yields, which means that customers will need more SoC testers/volume to achieve the same output. This directly affects testing costs and demand for display driver ICs and Chip on Film.

However, tester demand is determined in large part by changes in the volumes of semiconductors produced, not just by yields. 5G semiconductors have decent volumes, which drives volume growth despite lower yields on certain nodes.

 

Regarding DDIs (Display Driver ICs), while the CoF (Chip on Film) type of DDI requires doubled or tripled test times, parallelism is limited to up to 4 devices due to technological factors. Thus, advancing CoF conversion will increase test demand but volume is limited to parallelism. Advantest expects the annual size of TDDI and CoF business to remain at the 10-billion-yen level for now despite greater usage.

 

 

Advantest in the Silicon on Chip Tester Market

Growth for SoC comes from advance process nodes for smartphone and servers. While CIS (CMOS image sensor— a device that converts the appearance of a subject into an electrical image signal) from phones and other applications may not move the needle for growth now, it will in the future, including DDIs (Display driver IC). Advantest benefits more from CIS and DDIs than Teradyne, since it already commands a large share of the DDI testing market and is still experiencing sharp growth for non-DDI products such as application processors.

Source: Advantest Company Presentation

 

 

For Advantest, high end SoC testers, used in high performance computing and communications, CPUs, processors, accounts for 70% for SoC tester sales; with just over 20% for automotive, industrial machinery, and consumer products; and almost 10% for DDICs (see above). SoC growth ranges from 8-20% yoy.

 

In the mid-end smartphone space, customers are always asking Advantest to lower test costs, so there’s a huge pressure to improve the efficiency of testing. To reduce the CoT (Cost of testing), reducing testing time and increased utilization is required. For Advantest’s smartphone silicon on chip customers, the test cost of semiconductors is about 1% for low end products and 8% for high value devices.

 

 

 

Advantest’s position in for memory testing devices– DDR5/ LPDDR5

 

Memory testing will play less of a role to Advantest, since SoC is 80% of its business, but despite memory being 20% of Advantest’s sales, it still holds the greatest market share in this sector at 55-65%. Samsung is a major Korean DRAM customer and they have invested continuously in memory testers.

 

Keep in mind that while Advantest holds about half or more of the memory market, Teradyne is right behind, it has about 40% or more.

 

In a 2021 earnings call, CEO of Teradyne Mark Jagiela mentioned—

 

“The memory market in 2021 was about flat with 2020 at approximately $1 billion. Our Magnum family continues to shine in the NAND segment, and we're reinforcing our position in the DRAM segment, as the industry prepares for LPDDR5 and DDR5 ramps in 2022 and beyond. Our share remains at about 40%.”

 

The market has an increased demand for high bandwidth memory and is transitioning to DDR5 which leads to higher layer counts, further miniaturization (density gains driven by DRAM node shrinkage and by multilayering in NAND flash memory). This results in quality assurance requirements increasing, leading to DRAM testers to grow in 2023-2024. For the memory market, high-end smartphone manufacturers will initially be the primary customers driving growth for DDR5/LPDDR5.

 

DDR5 is short for “DDR-SDRAM 5” which means “Double Data Rate Synchronous Dynamic Random Access Memory generation 5”.  Let me simplify it for you— it’s the fifth iteration of a RAM (random access memory) technology that uses both rising and descending cycles in memory refresh rates, with a data-rate up to 6400 Mbps, as opposed to only the rising cycle of the original SDRAM tech.

Mobile DRAM bit share has grown more than 500% since 2009, according to market research firm IHS Markit. To meet this demand, chip makers are developing advanced SDRAM technologies such as LPDDR5 and DDR5 memories with data-transfer speeds up to 6.4Gb/s.

Despite such growth, a conservative estimate for the memory tester market should be high single-digit yoy growth. Let me explain why.

 

Advantest has fierce memory testing competitors in South Korea and China for the low-end tester models, with some existing customers switching to in-house testers. With regards to DRAM testing in particular, efforts have been made to increase parallelism in wafer testing, including probe card vendors. However, increase efficiency per piece of equipment is soon approaching its limits. It will be more difficult to maintain a market share in the memory space, but SoC is overshadowing memory testing for Advantest’s sales, so the harm will be minimal.

 

Advantest flagship DRAM memory tester is T5503HS2, CEO Masuhiro Yamada claims—

 “The T5503HS2 delivers best-in-class performance and precision for next-generation memory ICs; the system’s unmatched scalability and productivity enable it to evaluate LPDDR5 and DDR5 devices as quickly, accurately, and cost effectively as possible.”

Advantest’s T5503HS2 is designed to provide test solutions for this new wave of memories (LPDDR5 and DDR5 devices with DQS-DQ timing differences with real time tracking and ALPG) as well as existing devices. It is capable of testing at speeds up to 8 Gb/s with overall timing accuracy of ±45 ps. Leveraging its 16,256 channels, it is a versatile system that achieves parallelism with cost efficiency in testing next-generation devices.

 

 

 

 

Cashflow and Valuation  

 

As of March 2022, Advantest had roughly USD 7 to 8 billion (1T yen) of revenue. Advantest is expected to grow their revenues by a further 15% in 2023. R&D is around 13% of sales and operating margins are over 30%.

Advantest expects operating cashflow will accumulate to a range around ¥220-290B (USD1.6-2.2B) for the next 3 years, about ¥140B will be in growth investments which include capital expenditures (about ¥40B) and strategic acquisitions and investments (about ¥100B), while about ¥150B (previously ¥65B) will be used for dividends and buybacks to bring shareholder returns.

Advantest has mentioned that additional debt may be incurred during expansion as they may increase dividends to 50% in conjunction with their expansion plans. Under normal working business conditions, minimum cash reserves will be around ¥80B yen (USD590M), with management aiming for a return on equity of 20% or more.

With a conservative 10x multiple, that’s roughly a market capitalization of ¥1.5T or USD11B. As of this writing, the market capitalization for Advantest is 2.2T yen or USD16.3B. Advantest is currently trading at a cheaper multiple relative to other automatic testing equipment manufacturers (well, I got in earlier… so it was even cheaper then). If you don’t like crazy high valuations, Advantest would be a suitable candidate for you.

 

 

Source: Annual Report

 

 

 

 

Track record differences between Advantest and Teradyne in the past 5 years

 

With major growth in the total addressable market and effectively operating in a duopoly, Advantest has been making strides in gaining market share. Advantest had a 36% market share in 2017; by  2021 it was estimated to be at 47%. This major advancement was made possible via a combination of two key factors:

 

1. A strategic shift to expand into the SoC (System-on-a-Chip) testing market in 2010, with the successful acquisition of US company Verigy with its V93000 EXA Scale. 

 

For memory testing, Advantest has a lion’s share of 57% the memory testing market and 30-40% of the SoC tester market. Advantest was able to enter and penetrate the SoC market thanks to its acquisition of the U.S company Verigy (ranked 3^rd by size in the field at the time) in 2011, which is three to four times as large as the memory tester market.

In contrast to memory, where the same company handles everything from design to manufacturing, SoC semiconductor production is increasingly spread across separate countries that handle design, wafer manufacturing, assembly, and chip packaging testing, so SoC requires understanding the supply chain and catering to high performance logic semi-conductors such as APUs and GPUs.

 

2.  Continued bolt-on M&A to strengthen Advantest's product offering, building up its expertise over the complicated global SoC supply chain.

Advantest's track record speaks for itself. The 5-year CAGR for sales and earnings as well as book value is far superior to Teradyne, with sales growth at nearly three times and EBIT growth five times faster.

 

 

Acquisitions

Following the acquisition of the system level test (SLT) business of Astronics in 2018, Advantest strengthened their recurring businesses with the acquisition of two more US companies, Essai and R&D Altanova in 2021. Altanova is a leading supplier of consumable test interface boards, substrates and interconnects for high-end applications.

Advantest will expand its high-end test board manufacturing footprint in Asia by combining Altanova's high-density PCB design technology with Xingpu Technology's manufacturing capabilities.

In addition, Advantest has recently reinforced their core businesses with the acquisition of Italian company CREA in August 2022 in preparation for  the test market for high-power analogue ICs such as SiC/GaN, which contribute to improved energy efficiency since there is a need for lower power consumption in data centers and a wider adoption of electric vehicles.

Apart from these acquisitions for Advantest’s hardware, Advantest is also building for cloud services and data analytics.

 

Source: Karreta Advisors

 

For Advantest, both free cash flow margins and returns have vastly improved over 5 years, with ROE rising over 20%. Teradyne has seen free cash flow margins drop.

 

Source: Annual Report

 

 

 

 

 

 

Teradyne vs Advantest’s Products and Installation Base Figures

 

Source: Teradyne, Advantest, Cohu (Khaveen Investment research)

 

Cohu ASP, Install Base and Revenue

Cohu has an average revenue in the past 2 years (2021-2022) of approximately USD850M with operating earnings of approximately 125M. Cohu has the unique plus of having the world's largest installed owner base of installed test handlers (now exceeding 37,500 installed owner base). This brings us to an ASP for Cohu of approximately USD22,600.

 

Advantest ASP, Install Base and Revenue

Advantest, on the other hand, has less units sold than Cohu, but my guess is with brand recognition and IP’s, they can sell with more pricing power/higher ASP. Advantest has an average annual revenue of about USD3.5-4B and operating earnings of about USD 950M-1.2B. With Advantest, approximately 28,000 test systems and handlers have been sold since 2020. 6,818 units out of the 28,000 systems have been sold in the last five years. This means a quarter of a total unit sales were sold in the since 2015. This demonstrates the exponential growth in semiconductors.

Advantest sells less units than Cohu, but has an ASP of approximately USD135,000. This means that Advantest has certain models in SoC that are priced way higher than Cohu.

 

Teradyne ASP, Install Base and Revenue

Teradyne has been more transparent with their install base. In the past few years, approximately 23,000-25,000 test systems and handlers have been sold by Teradyne. Teradyne has an annual revenue average of approximately USD 3.1 to 3.7 billion, with operating earnings around 850M to 1 billion. This roughly translates to an ASP of USD132,000.

 

Install base x ASP = roughly translates to revenue

Source: Author's Estimates

 

Sectors that may not move the needle, but increase the ASP

Advantest is actively making new inroads into SSD test systems, E-beam lithography, and terahertz spectroscopic imaging systems. These products don’t move sales but brings up ASP.

 

 

 

 

Teradyne and Advantest have superior pricing power

Assuming my figures are all correct, the average sales price for Teradyne and Advantest are about triple of Cohu. Cohu sells more units but does not sell extremely high-end testers.

Unfortunately, Advantest has not been transparent with their install base regarding memory, microcontrollers, FPGA, and other IC products. It would be good to know the install unit base figures for each category, so we can have a direct comparison to see how much stronger Advantest is in the DRAM & NAND (memory) space in contrast to Teradyne.

Here is Advantest’ product line up compared to its competitors Teradyne and Cohu. Teradyne and Advantest caters for a broader range of semiconductors ranging from microcontrollers to memory such as DRAM and NAND as shown in the table below. Cohu has the most limited options without equipment for GPU/CPU and NAND & DRAM.

 

 

 

 

Barriers to Entry: Development costs, broad testing equipment portfolio, IP, and special staff

 

Advantest’s broad portfolio range and high product development costs associated with testing equipment provides a significant advantage and barriers to entry. Designing and developing new semiconductor testing equipment takes between 400 to 600 engineers and a quoted one billion or more in development costs and four to five years to build out.

 

Teradyne and Advantest both have low employee turnover— only 2% or less of its engineers which enables both companies to remain ahead of the competition. As the pace of change in the semiconductor industry is blisteringly fast, developing new testing products which are ready to ship requires the best and brightest.

 

Customer have been sticky despite a competitive landscape. Even though Teradyne has tried entering the memory segment— backend DRAM, there won’t be major shifts in market share for memory testers since there always has been dual-sourcing policies that pit Advantest against South Korean firms.

 

 

 

Risk and other concerns

 

Teradyne, Advantest, and Cohu all have surplus cash reserves, and minimal debt. The cyclicality industry requires solid balance sheets to ride out the waves.

 

Honestly, I think Advantest’s biggest risks are:

 

1. Procurement and supply chain of vendors
2. Delays in investment to necessary infrastructure/capital expenditure
3. Getting complacent, or an inability to catch the next paradigm shift
4. Offering a ridiculous price for acquisitions.
5. An unfavourable policy stipulated by China or USA

 

Advantest refuses to disclose order data (they did in previous years) but mentioned that they have a substantial order backlog for the next few quarters.

 

Advantest’s supply has been unable to keep up with customers' demand over the past years. The main risk or concern is procuring components. Advantest has the production capacity in place to support roughly ¥120 billion (USD880M) in quarterly sales and nearly ¥500 billion(USD3.6B) on an annual basis. However, Advantest may struggle to keep up with the sharp rise in orders. To ensure that Advantest has enough components, suppliers of components for semiconductor testers must be prioritized and customers have to be categorized.

 

 

 

Orders may weaken further in the second quarter of 2023, but considering factors such as Advantest’s order backlog, service revenue, and consumables revenue, the odds of any year-on-year decline in Advantest’s sales exceeding 15% and reaching the level of 20-30% to be extremely small but not impossible.

 

 

Restrictions on exports to China

 

China accounts for a quarter of Advantest’s sales. China accounts for some of the volume production of NAND flash memory which Advantest benefits from. The launch of a 64-layer device saw increase in orders. In addition, in DRAM, because the ramp-up time of LPDDR5 and DDR5S has been pulled in, Advantest is seeing traction in high-speed testers, including for GDDR6.

 

Tester sales in China comes from indigenous fabless companies, and outsourced semiconductor assembly and test (OSAT) vendors. A blockade or any policies from the U.S can and will adversely affect memory testers sales into China. This is a real threat/risk and should not be taken lightly. 

 

Both the Taiwanese foundries and Taiwanese OSATs are busy at present, but the OSATs expect less business going forward because of the tensions between the US and China. If that proves to be the case, Advantest believes that it will take a bit of time before the resulting excess tester capacity at the OSATs is filled.

 

Is the Arizona plant feasible? Are fabrication plants in China possible? It does not matter to Advantest… Heads I win. Tails I win. Government spending only brings in more business

 

U.S and China are starting to bring industrial production back onshore (domestic fabrication) as a national security imperative to be less reliant on overseas supply chains.

As part of Biden’s 2T infrastructure plan, 50B will be invested in semiconductors. The first TSMC plant in Arizona is expected to produce chips by 2024 despite problems with sourcing water in the desert, higher expenses than Taiwan and China, and a lack of technical and competent workers which they are trying to resolve by sending people from Taiwan. TSMC was initially aimed to produce 5nm chips, but now will create 4 nm chips. The second plant will open in 2026 and produce 3 nm chips, the most cutting-edge chips currently available. Morris Chang, the man who brought prowess to Taiwan as semiconductor powerhouse, believes it is an exercise in futility, as they tried establishing plants in the U.S decades ago.

 

China is about 6-8 years behind TSMC and is trying to catch up by spending hundreds of billions to trying to build domestic fabrication and expertise.

To meet the surge in demand and close the supply gap, semiconductor manufacturers are set to spend 190B on capex in 2022, a 25% increase on 2021 levels. TSMC is in the process of building a new fabrication plant across 22 football fields of land in southern Taiwan. This new plant will produce the latest generation 3nm chips which will be 15% faster.

It does not matter if governments pointlessly spend billions to build fabrication plants—the increased investment from manufacturers will benefit equipment providers like Advantest.

 

 

 

 

Financials

Management had revised their shareholder return policy so that the dividend pay-out model was changed from the performance linked semi-annual dividend pay-out ratio of 30% to a minimum stable and continuous dividend of around ¥50 per share semi-annually and ¥100 per share annually.

Although there are no buybacks, with issued share remaining at 199M shares (see below), they have also not split shares, which is a good sign.

 

Source: Annual Report

 

Sales during Covid has been better than expected, and Advantest is raising their forecast from 350-¥380B(usd2.7B) to ¥480-520B (usd3.6B). Advantest employs more leverage than Teradyne.

Advantest has a debt to equity of 14 to 20% while Teradyne has a debt-to-equity ratio of 2 to 5%.

Advantest is benefiting from their larger scale as SG&A relative to sales has shrunk from about 48% to 29%, while R&D as a % of net sales has shrunk from 20% to 11%.

While capital expenditures for Advantest have more than tripled in 5 years, ROE has also doubled, with both operating and net margins improving by a sizable amount.

                                                                                                                   

Source: Annual Report

 

 

Conclusion

Advantest expects tester demand to be solid in 2024 given that chips using leading-edge 3nm processes are likely to get off the ground in a meaningful way in 2H 2023. Predicting where the macro-economy is headed is extremely difficult, but despite technological change, Advantest will prosper.

 

From competitor Teradyne’s earning’s call, CEO Greg Smith’s tone shows there’s a bit of a slump in testing equipment—

 “As a result, we don't expect any 10% customers in these end markets in 2023. In semiconductor test, lower end market demand and high channel inventory is persisting. And our measures of tester utilization in Q1 of 2023 are at their lowest level in over 10 years.”

This should be temporary; over the long term, I believe testing equipment is a very resilient and good industry.

 

The exponential growth in chips according to Moore’s law brings opportunity but is also fraught with risks. An incumbent can knock off the current king of the mountain; even startups that pass an inflection point with positive cashflow may eventually fall off a cliff. Just like the gold rush in California, not everyone who mines for gold succeeds. However, people who sell shovels, denim jeans, and mining equipment prospered with a minimal amount of risk.

VIC member Motherlode has suggested Micro Super Computer to take advantage of the capital expenditure increase for servers from Facebook.

Li Lu bought into Micron because memory has consolidated into a space of just 3 players— SK Hynix, Samsung, and Micron. 

I read a book called Chip Wars which really painted a broad picture of the semiconductor industry and its dominant, entrenched players. ASML, a spinoff from Philips, has a stranglehold on the most advanced lithography and in semiconductor manufacturing TSMC is the leader.

With chip testing, Advantest is well positioned in a duopoly. While testing equipment has risks, Advantest can be seen as analogous to an oil refinery, there’s always business for testing chips, and consistent profits is more important to me than fabrication plants that operate like vertically integrated oil conglomerates.  

Catalyst

- Growth from servers, data center roll out, smartphones

- Increased functionality in IC devices with less power consumption

- Consolidation of SoC market

- Stranglehold on the Memory market

- CREA (2022) and Altanova (2021) acquisitions prove complementary to business