The Hatch Fund is not a registered investment, legal or tax advisor or a broker/dealer. All investment/financial opinions expressed by The Hatch Fund are from personal research, analysis and experience of the owner of the site, and are intended for educational purpose. Please conduct your own due diligence before relying on the opinions presented on the site for your investment use. If in doubt, please seek advice from your legal, financial, tax, or other professional adviser(s). In the event that you choose not to seek advice from such advisers, you should consider whether the product in question, or any course of action recommended, is suitable for you. While best efforts are made to ensure the quality of our content, The Hatch Fund will not be liable for any gains/losses arising out of any recommendations made. All opinions/statement are made based on public information only.
If you enjoy reading the article and find it helpful, please consider subscribing to The Hatch Fund’s newsletter to get the latest notifications and insights! Your viewership makes putting hours into research and writing justifiable. Thank you!
For the uninitiated, AEM Holdings is a SGX-listed company that provides back-end test equipment to a variety of customers. Intel Corporation happens to be a key customer contributing 80% – 90% of its revenue. Essentially, for the CPUs sold in retail shops to reach end consumers, they must first be designed, manufactured, tested and packaged before being shipped off for sale. AEM’s products come into the picture for the “testing” phase.
In this post, I look at the latest developments for AEM and its wider industry (integrated chip (IC) / personal computing (PC) / data center (DC)), before revisiting its financials/valuation to determine if it is still worthwhile to hold on to it.
Semiconductor Market Segmentation (FYI):
AEM Holdings Share Price Chart
Since the August-high of S$4.29 (closing price), AEM’s share price has fallen by about 19% to S$3.47. In this post, I hope to take a closer look at what has transpired over the last two months and to re-evaluate if AEM is still a sound investment. On the macro front, a couple of concerns are at the top of my mind:
- Vaccine sparking rotation out of tech sector: When a viable vaccine passed the final stage trials and becomes widely distributed/accessible, will the inevitable rotation out of the tech sector drag down AEM?
- Investors rotating into bonds: Investment managers are expecting 10Y treasury yields to rise following the U.S. presidential elections. When money starts to flow out from equity back into bonds, will AEM have enough price support?
- As I am not close to the details of the U.S. Presidential Elections, there is likely a blind spot (or several) that I maybe missing here (So feel free to point them out in the comment section below!).
1. Overview of Commercial Analyses
It is hard to predict the extent to which a rotation out of the tech sector, or equity asset class, will impact AEM’s valuation/share price, so we can only take comfort in AEM’s fundamental performance. While markets can be irrational and sentiment-driven in the short-term, they rarely deviate from a company’s true value in the long-term. In this first section, I take a commercial review of AEM Holdings based on the latest developments relating to the Company and its broader market. As this turned out to be a very lengthy article, these developments are summarized below:
- AEM insiders have sold a combined 2.7 million shares, which sparked two sell-offs in the last two months.
- AEM revised revenue guidance from S$460m – S$480m to S$480 – S$500m will likely be ignored by the market.
- Intel’s PC-centric business comes under threat from AMD and Nvidia as competition intensifies.
- Intel shifts focus to data-centric business but delivers disappointing Q3 2020 results for its data-centric business.
- Data-center market competition intensifies with AMD’s acquisition of Xilinx and Nvidia’s acquisition of Mellanox and ARM
- Intel is planning to sell its NAND memory business to SK Hynix.
- Intensifying competition in the PC-centric space (particularly CPU) is likely to push Intel to accelerate the development of 7-nm chips as well as other top-tier and mid-range chips to rival AMD. Additional capital expenditure is likely to be incurred for chip testing, which will benefit AEM in the short-term.
- Intel is pivoting to its data-centric business as the main growth driver going forward. However, data center competition is starting to intensify with AMD’s acquisition of Xilinx and Nvidia’s acquisitions of Mellanox and ARM.
- Intel’s sale of NAND memory business is unlikely to materially affect AEM’s performance as Intel will continue to operate its smaller Optane memory business, which has just reached break-even.
- Cloud Service Providers (CSPs), which are major data center customers, are increasing pivoting towards the use of ARM-based processors to reduce reliance on Intel and AMD.
- While it is irrefutable that Intel will lose data center market share, it is losing share in a rapidly expanding market. Rapid total addressable market expansion, coupled with new revenue streams from AEM’s recent acquisition is likely to prop up AEM’s performance in at least the next 2-3 years.
i. Company Update: Insider selling a combined 2.7 million shares
In early September 2020, we caught a glimpse of a rotation out of the tech sector when news of a working vaccine were published. Tech stocks closed out a ‘nightmare’ September with the tech-heavy NASDAQ shedding ~5% as Big Tech stocks took a nose dive.
Being a company with one of the strongest financial performance in SGX and being the darling of the SGX (Bloomberg, BusinessTimes, StraitsTimes), AEM needed a jolt to help in its dip. And that came in the form of Executive Chairman, Loke Wai San, selling a combined c.1.2 million shares of AEM (announcement 1, announcement 2, announcement 3). This was swiftly accompanied by a 1.5 million share disposal (1 million indirectly held) by James Toh Ban Leng, who is the single largest shareholder and a NED/ID of AEM Holdings, and also the Founder and NED of Novo Tellus.
According to the 2019 Annual Report, Loke Wai San had approximately 3.4 million options outstanding as at 31 December 2019; James Toh had none. The exact exercise price of the share options are undisclosed, but they range from S$0.196 to S$1.142, which represents a c.220% to c.1700% discount to the current market price.
In essence, options granted to Loke Wai San gives him the right to buy additional shares of AEM Holdings at a discounted price. While I want to be convinced that these insider transactions may be nothing more than some astute investors timing the market, it is important to remain objective in evaluating if the insiders are exiting because of AEM’s prospects. Furthermore, LWS is planning to step down from his Executive Chairman role to become a Non-executive Chairman in 2021. This represents another key event that the market will likely react negatively to in 2021.
ii. Company Update: AEM raises revenue guidance
Following the sell-offs sparked by insider sales and global tech selloffs, AEM announced that it is raising revenue guidance from S$460m – S$480m to S$480m – S$500m for FY2020. This comes as no surprise as the firm is known to “time” its announcements over the years to lend support to the price / momentum. The market reacted to its revision in a rather muted fashion.
iii. Key Customer Update: Intel PC-centric business comes under threat
a. AMD catches up to Intel’s lead in the CPU market
Intel’s long-standing lead in the CPU market is increasingly coming under threat from AMD (selling more value-for-money and faster high-end CPUs) and Nvidia (following the announced acquisition of ARM). For context, Intel has been dominating the personal computing (PC) segment for years, specifically in the central processing unit (CPU) category. However, Intel has had several hiccups in recent years — including delays on its 10-nm and 7-nm chips, prioritizing share buybacks over R&D, and the restructuring of its technology team which caused its Chief Engineering Officer to leave the firm — all of which allowed Intel’s long-time rival, AMD, to play catch up.
In 2020, AMD released its 7-nm CPUs which performed better than Intel’s chips for a fraction of the price. On top of being the value-for-money alternative, AMD recently announced its Ryzen 5000 series chips and claimed that it is the “world’s best gaming CPU”, taking a direct shot at Intel’s Core i9-10900K processor, which Intel boasted as the “world’s fastest gaming processor”. Meanwhile, Intel prepares for a response in Q1 2021 with its 11th Gen Rocket Lake desktop processors, further intensifying competition in the PC space.
Source: The Verge
b. Nvidia acquiring Arm for US$40B: Market is moving away from x86 to ARM architecture, posing a huge threat to Intel’s long-term viability in PC business
Nvidia has long been the leader in the graphics processing unit (GPU) market (FYI: GPU vs. CPU). Nvidia’s growth had been largely fueled by the demand for GPUs, the main computing unit used to power accelerated computing systems, such as AIs, autonomous vehicles, super computers, etc.. In August 2020, Nvidia acquired ARM to extend its product line into the CPU segment.
i. Shift to ARM-based processors to benefit Nvidia at Intel’s expense
While ARM plays in the CPU market, it is largely different from Intel and AMD — Intel and AMD design and manufacture their chips for direct installation onto a motherboard; On the contrary, ARM designs the silicon and architecture upon which allow chip manufacturers to have more customizations, giving extra flexibility to have chips designed around a product rather than the other way round. As such, ARM has been the preferred CPU for mobile devices due to its efficiency and design flexibility. This allowed ARM’s global market share in core design architecture for mobile processors grow to more than 90%.
Recently, the use of ARM-based CPUs in computers has also been catching on. Microsoft produced its first ARM-based Surface laptop, while Apple ditches Intel’s x86 chips for its own ARM-based processors (Apple Silicon) in its Mac computers. The benefits of an ARM-based chips include better battery life, greater ease of adding LTE options to hardware than x86 processors, being able to run cooler and, consequently, allowing for thinner/lighter form factors. Thus, these developments bear testament to Nvidia’s ability to compete directly with Intel in the PC segment.
ii. Synergy arising from ARM acquisition
Furthermore, the acquisition of ARM will likely bring revenue synergies as the combination of CPU and GPU businesses will allow Nvidia to create better integrated designs. For example, in June 2020, an Arm-powered supercomputer, coupled with Nvidia’s Telsa V100 GPUs, took the Supercomputing Top500 performance crown. Beside supercomputing, this acquisition will allow Nvidia to compete in the data center space by allowing it to offer complete platform solutions alongside its GPUs offerings for ML, AI, cloud and big data applications.
iv. Key Customer Update: Intel shifts focus to data-centric business but delivers disappointing 3Q results
As competition in the PC segment heats up, Intel is looking to shift its focus to the more profitable data-centric business (Data Center Group (DCG) and other data-centric business segments), which collectively accounted for ~50% of Intel’s total revenue in 2Q20. However, on 20 October 2020, Intel announced revenue from DCG fell from US$7.1b in 2Q20 to US$5.9b in 3Q20. Other key updates on earnings:
- PC group (Client Computing Group) earned US$9.8b revenue, beating analyst estimates of US$9.09b;
- Operating margin fell from 44% last year to 36% in 3Q20 as higher volume of the less profitable chips in PC business were sold.
v. AMD's acquisition of Xilinx and Nvidia's acquisition of ARM and Mellanox heats up the Data Center market
a. Nvidia acquiring ARM to challenge Intel in DC market
In 2019, cloud service provider (CSP) giant, Amazon, was reportedly working on a second Arm-based chip for data centers as it looks to break free from reliance on Intel and AMD for server chips. Furthermore, Nvidia plans to create products for the server CPU market that combines Nvidia’s AI and GPU offerings with ARM’s offerings. With CSPs being the largest client segment in the data center market and for Intel’s data-centric business (accounting for 47% of DCG’s revenue), this development is quite concerning for Intel.
b. AMD looking to acquire Xilinx for FPGA technology
Akin to Intel’s US$16.7b acquisition of Altera in 2015, AMD’s proposed acquisition of Xilinx for US$30b is aimed at moving AMD’s focus from desktops and consoles to FPGA technology, which are crucial to server chips used in data centers. While I have not followed this market for long, it seems clear that engaging in acquisitions to break new ground is the name of the game. In the longer term, Intel’s competitiveness in the DC space will likely hinge on whether ARM-based chips can replace x86-based ones (Intel’s / AMD’s chips). (See below video to understand the purpose of FPGA in data centers)
Effectively, both AMD and Nvidia are closing in on Intel and starting to chip away at its market share in the DC market. In July 2020, AMD CEO Lisa Su announced that AMD has successfully surpassed 10% market share in the profitable server chip market. However, it is noteworthy that this market is rapidly expanding and Intel may still have sizeable growth potential.
vi. Intel to sell NAND memory business to SK Hynix for US$9b
a. Basic deal structure
On 20 October 2020, Intel announced the sale of its NAND memory and storage business to SK Hynix for US$9 billion. Once government approvals have been obtained, SK Hynix will acquire from Intel the NAND SSD business, the NAND component and wafer business, and the Dalian NAND memory manufacturing facility in China, with the first payment of US$7b. The remaining assets, including IP related to manufacture/design of NAND flash wafers, R&D staff and Dalian fab workplace will be acquired upon final closing (expected to occur in March 2025) for US$2b. Post-deal, Intel will retain its smaller Intel Optane memory business, which houses the 3D XPoint memory co-developed with Micron.
b. Why is Intel selling the NAND business?
The NAND business contributed to 7% of Intel’s total revenue and 93% of its Non-Volatile Memory Solutions Group (NSG) revenue in 1H20, generating US$600 million in operating profits in the same period. So why is Intel selling its NAND business? It was long thought that Intel’s Optane business was subsidized by the NAND business as its prices are half of the DRAM memory which it is trying to displace. Importantly, the chart below shows Intel was making losses in 4Q17, 3Q18, 2019 and 1H20 when most of its peers were enjoying a healthy operating margin. Forbes reported back calculations showed that the Optane unit was making loss of US$2b in 2017, US$2b in 2018, and US$1.5b in 2019. In Intel’s 2Q20 earnings announcement, the Company mentioned that its Optane memory business has finally reached break-even, which is an encouraging sign for Intel as Optane will be retained after the NAND memory business sale.
To my understanding, Optane performs the same functions as chips in flash-based thumb drives, but (1) runs faster than most flash memory and almost as fast as the DRAM chips used in PCs and servers, (2) has lower latency and (3) has better endurance than NAND flash memory.
According to various sources, the rationale for Intel to carry out this deal include the following:
- To divest a less profitable business unit — Intel CFO declared the NAND business was generating inadequate profits, suggesting that Intel lacked the scale and pricing power of larger memory chipmakers; and
- To reduce China exposure ahead of rising trade tensions between China and U.S.
What do these all mean for AEM Holdings?
Having caught up with the latest developments, here’s my take on how this is likely to affect AEM from a commercial point-of-view:
- AEM will benefit from competition in the PC segment in the short-term as Intel will ramp up chip production to fill the void created by its 7-nm delay and to keep up with AMD (similar game plan to its 10-nm delays). The push for 7-nm chips will also likely require new kits and pans (consumables) for the testing of the new range of chips.
- The sale of Intel’s NAND memory business might bog down AEM’s sales in the near term as there will be less chips to test post-deal, and hence, less consumables will be required (given the Optane is of smaller size).
- While 5G, AI, IoT, increased internet penetration, social media and content consumption (IPTV, OTT) trends will drive the need for data centers and edge data centers in the future. The adoption of ARM-based processors by cloud service providers (CSPs) like Amazon and Google is a worrying sign for Intel, which stands to lose its market leadership.
- In the coming months/years, it will be important to keep an eye on the success of these ARM-based processors in executing DC operations. Should they prove to be successful, servers due for refresh may replace existing Intel server chips with ARM-based ones and Intel faces the risk of being phased out.
Overall, I remain cautiously optimistic that AEM Holdings still has a lot of room to grow given that memory chips/SoC test equipment competitors like Teradyne (Intel’s other Automated Test Equipment (ATE) supplier) and Advantest (AMD’s ATE supplier) are about 10x bigger. However, I am of the opinion that investing in AEM carries oversized downside risks (i.e. downside is significantly more than upside) in the longer term for the following reasons:
- Intel’s market leadership in the PC and DC segments are not only under threat, but may also be rendered redundant by ARM-based processors;
- The release of COVID-19 vaccines and treatment will likely spark a rotation out of the Tech sector, which will adversely affect AEM’s share price;
- Post-pandemic, interest rates will likely be raised to curb inflation arising from increased employment levels, U.S.-Sino trade tensions. During the great inflation of the 1970s, stock market valuations collapsed as interest rates soared, leaving the Dow Jones Industrial Average trading relatively flat between 1965 and 1980 (Source: Reuters).
This is not to say that AEM’s performance will collapse as Intel loses market share, but I expect the market to take this news badly and oversell AEM. In all likelihood, the rapidly expanding DC market and AEM’s new revenue streams from its multiple acquisitions (MuTest, DB Design, UTAC) will likely prop up AEM’s performance. Furthermore, should Intel opt to exit its chip manufacturing business and outsource it, this will likely play to AEM’s advantage as more chips will be produced and, hence, tested.
Overall, from a commercial point-of-view, holding on to AEM in the long term may inadvertently turn into a bet on a dying horse if ARM-based chips become successful. An unclear path to restore market leadership and a seemingly inevitable fall in valuations across the sector/market will continue to keep me on my toes. Regardless of how rosy MEMS, 5G, IoT or AI trends look, I will prefer to err on the side of caution and continue to monitor developments in this space. As the saying goes, “a bird in the hand is worth two in the bush”.
Verdict: HOLD/Exit at a higher price and buy on the dip (to monitor ARM-based chip’s success)
Now, let’s move on to the financial analysis.
2. Overview of Financial Analysis
As this article is becoming quite lengthy, I will try to keep this section as succinct and informative as possible. The financial analysis performed here is split into two parts: First, a quick comparable company analysis to see how AEM is performing / valued vis-a-vis its peers. And second, a quick discounted cash flow (DCF) analysis to determine what AEM is “actually” worth.
a. Comparable Company Analysis: Selected Global SoC Test Handler Equipment Providers
This list is split into two groups: (1) global comparable SoC test handler equipment providers, and (2) APAC backend test equipment providers. In finance, this exercise (comparable company analysis) aims to understand how much a firm is worth by looking at companies that are similar to it in terms of business operations, geographic exposure, client base, and various risks faced by the company (e.g. political risks). However, I want to compare apple to apples in the first parse. As such, I first compared AEM to global System-on-Chips (SoC)/IC test handler equipment providers (with a PC/DC client base) before comparing them with backend test equipment provider that are closer to home.
i. Comparing Profitability Margins
Notably, AEM’s GP and EBITDA margins fall below larger peers like Teradyne and Advantest (although it is arguably close to the peer average). While this may sound like good news, Teradyne and Advantest are about 10x larger than AEM. And typically, as companies grow, it becomes harder to maintain a high level of profitability. Thus, I am of the opinion that AEM’s profitability lags its peers after considering the scale of its operations. For those who are not familiar with finance/accounting, I attempt to explain these jargons in the table below in a relatable manner:
ii. Valuation Multiples
While the 4 charts above seem to suggest that AEM is massively undervalued, this is not an apple-to-apple comparison for a couple of reasons:
- With the exception of Cohu, the global peers are much larger than AEM (in terms of revenue and market capitalization).
- Teradyne and Cohu are NASDAQ-listed while Advantest is listed on the Tokyo Stock Exchange (TSE), making them much more accessible to global investors (i.e. more money can flow into these shares to support a higher valuation).
Hence, we also have to take a look at semiconductor test handler equipment providers that are closer to home (Singapore) to get a better sense of the valuations on these types of companies in the region. Below are the selected comparable companies.
b. Comparable Company Analysis: Selected APAC Semiconductor Test Handler Equipment Providers
Here, the importance of finding comparable companies producing similar products for similar client base is exemplified as margins and growth rates for selected companies deviated greatly from AEM. However, the aim of looking at the APAC companies is to mainly look at the valuation multiples awarded to companies in the backend test equipment manufacturing space in APAC.
From the charts above, AEM seems undervalued vis-a-vis peers (AEM falls below the red dotted line). However, one has to take into account the inflated valuations in the Tech sector and tone down expectations of AEM achieving such valuation multiples in the future. Looking at AEM’s historical valuation multiples (chart below) shows us that the company has never traded near these multiples suggested by its peers and will likely remain in this range going forward.
b. Discounted Cash Flows (DCF) Analysis
For the benefit of all readers, DCF is a method of valuing a company based on projections of the future cash flows that it is expected to receive. This is akin to saying Apple Tree X is worth $10 today because it will likely bring in $1 every year for the next 15 years. But if Apple Tree X brings in $15 ($1 x 15), why is it only worth $10 today? Isn’t that free money?
Well, no. Here, we bring in the concepts of (1) time value of money / opportunity cost, and (2) risks. The concept of time value of money is such that $1 in the future is worth less than it is today because of inflation, or potential interest to be earned from $1 today if it was saved/invested. Also, we have to account for the risks associated with buying and growing Apple Tree X. For an apple tree, these risks include disease, possibility of drought /fire/war, and potentially wilting. For a company, this ranges from political risks (e.g. US-China trade tensions), disruptions to operations (e.g. social unrest / war / natural disasters), and perhaps even the possibility of bankruptcy / insolvency. Thus, we apply a “discount” on the projected future cash flows to account for these risks and opportunity cost of investing in an Apple Tree, as we would for a company.
In Valuation / DCF terms, this “discount” is what we call “Discount rate” or ‘Hurdle rate”. So in the case of Apple Tree X, we can back-solve to find that the implied discount rate is actually 5.6% (assuming it gives $1 for 15 years only. Run the numbers yourself!). Without running the risk of going into too much detail, this should give you a basic understanding to roughly follow the analysis below.
Below is the projected 3 financial statements used in the calculation of free cash flows to firm (FCFF). FCFF is basically a jargon that means cash that is readily distributable to all stakeholders of the firm (debt holders, preferred shareholders, equity shareholders). In essence, it is the money that is in excess of amounts needed to keep the company operational (after paying salary, rent, utilities, administrative expenses, tax, buying and maintaining machinery, etc.), and hence, readily distributable to said stakeholders.
Here, an exit multiple of 7.7x EV/EBITDA in FY2024F and WACC (discount rate / hurdle rate as discussed earlier) of 10.0% are assumed. What does this mean? Essentially, it is saying the value of the company TODAY is equals to the sum of discounted free cash flows to firm (FCFF) in the next 5 years + the value of the business based on existing public market valuations in 2024 (discounted to present day). With assumptions of growth rates, margins, exit multiple and discount rates made, we get an implied price per share of S$3.90.
To practise conservatism, I made the following assumptions:
- No multiple expansion — i.e. AEM continues to trade at 7.7x EV/EBTIDA, which is much lower than its comparable peers as seen from the comparable company analysis done earlier.
- AEM’s revenue and profitability margins are pegged to analyst consensus estimates for FY2020, FY2021 and FY2022. Thereafter, I pared down revenues in FY23F and FY24F to ~S$450m levels. This discounts AEM’s efforts to increase the number of revenue streams through various acquisitions (MEMS testing, network cable testing, Camera sensor testing, etc.).
As the target price is very sensitive to discount rate and exit multiple assumptions, I co-varied them to find out the intrinsic value in each scenario should my assumptions be inaccurate.
The below table shows us that there is still a good margin of safety at the current price of S$3.47 (as at 23 October 2020) and AEM will be a good buy if it dips to S$3.0 – S$3.3 range. [How to read the table: (1) match exit multiple column with desired discount rate row, (2) intersecting point is the implied share price given those two assumptions –> e.g. 6.5x EV/EBITDA + 8% WACC assumed will give us S$3.75 as the intrinsic equity value (what the stock is really worth)]
All things considered (commercial and financial aspects), I will recommend fellow investors with no exposure in the Tech space to BUY AEM near the S$3.2 – S$3.3 range (based on a potential upside of at least 20%). In the near term, I remain bullish on AEM’s fundamental performance but will actively monitor the success of the ARM-based processor in DC applications and its response to AMD in the PC space.
While it is irrefutable that Intel WILL LOSE DATA CENTRE MARKET SHARE, it is equally important to consider that it will be doing so in a rapidly expanding market. Furthermore, if rumors of Intel outsourcing its chip manufacturing crystallizes, it will also provide a boost to the demand for test handler equipment, further lifting AEM’s performance.
The main concerns that stop me from recommending a strong buy include the following:
- AEM’s valuation will likely be challenged due to the rotation out of the Tech sector and the Equity asset class;
- ARM-based processor’s success in displacing x86 chips will have significant adverse impact on Intel and AMD (x86-based chips);
- Intel’s ability to consistently roll in revenue and profits from its DC business (given the lacklustre third quarter performance); and
- Outsourcing of manufacturing may lead to testing being outsourced as well — although the probability is likely low as Intel has co-developed equipment with AEM for more than 10 years now, and AEM’s equipment has reduced Intel’s costs significantly.
If you care about timing the market (like me), I expect valuations to dip once vaccines finish their final stage of trials and become distributed. This will be a good time to buy into AEM. Any complications with the vaccines (hopefully not), their distribution/manufacturing, or negotiations between government bodies is expected to delay this. Any time from now till then, if prices rise to the S$4.00 range, it will probably be a good time to take profit and wait for the eventual dip.
If you do not care about timing the market, the huge margin of safety (based on my conservative estimates) should provide enough comfort to buy into AEM Holdings.
If you read till the end, thank you. This was an unexpectedly long article!