Description

SUMMARY

AXT Inc. (ATXI-US) is an $150 million micro-cap company that develops advanced compound substrates for use in electronic devices, and increasingly, photonics applications (the use of light to transmit data) used in fiber networks, next generation data centers and components ("Big Data" thematic), autonomously driving vehicles and 5G wireless networks.

AXTI's indium phosphide (InP) substrate business, or the substrate key to photonics applications used in the latter high growth end markets, has grown in excess of 50% in each of the past two years and accounted for nearly a quarter of 2015 revenue, up from the mid-single-digits a few years ago. InP growth should continue to compound at similar rates moving forward driven by secular photonic tailwinds.

InP carries gross margins of ~50%, or ~2.5x total consolidated gross margins at AXTI; as InP becomes an ever-greater piece of mix it should drive material expansion in gross margins at AXTI to at least the historical "sweet spot" of previously strong growth cycles in the 30%-40%+ range and potentially higher thereafter. InP gross margins are likely defensible given there are only three other manufacturers of InP globally according to AXTI.

Ex the growth in its core InP business, AXTI owns a 300,000 square foot industrial production facility in Beijing. We estimate the building is worth ~$42 million, or ~25% of the company's current market cap and ~2x our balance sheet net carrying value estimate $20.8 million. The market is broadly unaware of this gem. Applying AXTI's current P/B multiple of ~1.2x to the ~$21 million step up to net assets on the balance sheet (i.e., $42-$20.8) and dividing by diluted shares outstanding of 32 million generates an immediate $0.80 per share in incremental upside to the stock, or +20% vs. the existing price of $4.40.

In addition to the Beijing-based real estate (~$42 million), summing cash on hand (~$44 million), A/R and inventory less a 50% discount (~$9+$19 million) and subtracting total liabilities including non-controlling interests (~$14+~$6 million) implies total liquidation value of ~$83 million, half the existing market cap of $150 million before assigning value to any other assets on AXTI's balance sheet (i.e., pre-paid expenses, PP&E other than Beijing building, full valuation allowance on DTAs, but $180 million in federal NOLs). This is fairly draconian for a business that generates operating cash and has been just shy of FCF neutral over the course of the past three years.

Assuming AXTI's InP business grows as expected and its other substrate businesses continue to degrade marginally through 2018, the company should be generating in excess of $100 million in revenue by then with gross margins just north of ~30%, or at the low end of its historical "sweet spot". Applying historical valuation multiples (EV/S, EV/EBITDA, P/B) afforded to AXTI when it has produced gross margins in the past similar to our 2018 estimates and discounting back yields an average fair value range of $5.40-$8.65 per share. We believe fair value to be the high end of the range given our expectations that our estimates for 2018 results are far too conservative given that AXTI is already printing gross margins of 30%, or where we thought they'd get in that out year.

For those uncomfortable using historical valuation multiples on out-year estimates, AXTI recently had investments in two Taiwanese-listed wafer companies, GCS Holdings, Inc. and IntelliEpi, Inc. Both are likely AXTI customers that it sells InP to and are also benefiting from InP-related growth, just one step down the supply chain, and both generate gross margins of ~40%. Thus, they offer convenient comparables to the company's InP business. For comparison, IntelliEpi has a market cap of $137 million on TTM revenue of $27 million, or what AXTI's InP business alone should generate in 2016 revenue despite a similar market cap. Thus, using IntelliEpi's valuation, AXTI's InP business alone should be worth the same, which, when divided by 32 million diluted shares outstanding, implies aworst-case fair value for the stock today of $4.28, assuming the company's non-InP businesses have absolutely no value.

BACKGROUND

AXTI Inc. (AXTI) designs, develops, manufactures and distributes high performance single-element and compound substrates used in various semiconductor electronic integrated circuit applications. While silicon is typically used in these applications, some have requirements that exceed the capabilities of silicon. For instance, silicon may overheat or conduct electrons too slowly in some situations thus requiring higher-performance single-element or compound substrates. In such instances, the advanced materials that AXTI makes are used in lieu of silicon to offset such disadvantages.

AXTI manufacturers its materials entirely in China through 10 different joint ventures (JVs), with ownership interests ranging from 20%-83%. Three such JVs are accounted for using the equity method while the other seven are consolidated. AXTI has board representation on all ten. These JVs not only produce compound substrates, but also the base raw materials that go into them. AXTI’s vertical integration strategy in China was developed with the belief that it provided both cost/labor advantages and reliability of supply as well.

Historically, the bulk of AXTI’s revenue was derived from gallium arsenide (GaAs) for use in power amplifiers and radio frequency circuits in wireless handsets, satellite communications, LEDs, and lasers or germanium (GE) for use in solar cells and various optical sensors. GaAs combines elements from Groups III and V in the periodic table while Ge is a Group IV element. In addition, revenue is also derived through the sale of the base raw materials that are used to develop substrate products, often times to competing substrate manufacturers.

AXTI sells its products to epitaxial companies. As semiconductor manufacturing goes, a substrate must be grown on a wafer before the electronic circuitry can be applied. AXTI manufactures the substrate and then sells it to customers, the epitaxial companies, who apply those substrates and then sell the modified wafers to fabricators, chip designers, LED manufactures, etc. During 2015 AXTI had one customer that accounted for greater than 10% of total revenue while its five largest accounted for 40%. This is less about customer concentration than it is industry structure.

For much of the late 1990s and the first decade of the 2000s AXTI’s substrate business benefitted from the explosion in growth and demand for new consumer electronics technologies; its raw material business also benefitted from the uninterrupted run in commodity prices. However, beginning in 2011 a confluence of events began to pressure results. Silicon-on-insulator (SOI) technology emerged as a less expensive alternative substrate for use in numerous applications where GaAs had historically been used, such as the radio frequency switching function in cell phones. Though SOI still does not perform as well as GaAs, it advanced to perform well-enough in the context of its cost advantages.

We estimate that share losses to SOI in wireless handsets resulted in GaAs generating less than $30 million in revenue at AXTI in 2015, or less than 40% of total revenue, down from in $68 million and excess of 70% in 2010, a remarkable headwind.

Growth rates in solar installations, especially in Germany, a major geographic market of AXTI’s, also began to slow noticeably beginning in 2011, pressuring the company’s Ge business. This can be seen in the chart below whereby Ge growth rates at AXTI have slowed considerably and gone negative in recent years alongside deceleration in German and global installed solar capacity growth rates (single year aberrations in this trend are due to AXTI customer-concentrations and order lumpiness).

Separately, AXTI’s raw materials business also began to weaken significantly exiting 2011, with items such as gallium (Ga) experiencing the same issues of oversupply as other commodities around the globe, resulting in a sustained period of price erosion and revenue declines since. Note the tight correlation in AXTI’s quarterly revenues since 2011 relative to a plot of the CRB Index.

Collectively, the latter factors transpired to produce a cumulative 25% decline in revenue to $77 million in 2015 relative to 2011’s total of $105 million.

INDIUM PHOSPHIDE OPPORTUNITY

Despite the challenges in its traditional GaAs, Ge and raw materials businesses since 2011, a silver lining has developed at AXTI more recently in the form of indium phosphide (InP), a new, rapid growth compound substrate. Like GaAs, InP is a group III and V compound substrate. Unlike GaAs which has lost share to SOI in established consumer electronic devices, InP-based substrates have recently emerged as a key material in various cutting edge technological applications revolving around “big data”, autonomously driving cars and 5G, among others.

As far as “big data” is concerned, global web traffic is now so large in scope and so fast in speed that it has overwhelmed traditional copper and cable-based networks which use electric impulses to transmit data. Fiber networks that transmit data as light are increasingly used instead as they are far faster and can be scaled on demand.1 In addition, the same scope and speed challenges facing data transmission over networks similarly confront data centers; the bottleneck here is no longer the computational power of CPUs in data center server racks, but the speed at which that data can be communicated between various components within these servers. Power consumption of the data centers is also a concern. Traditional electronic integrated circuits (EICs) which transmit data between components using electrical impulses over copper wires no longer suffice for required communication speeds. Instead, data centers, just like the networks over which the data originally travels, are increasingly relying on photonic-based components within their servers to transmit data at the necessary speeds. These components are referred to as photonic integrated circuits (PICs) 2,3. A PIC is an integrated circuit that contains no electronic circuitry. This description from Oclaro, a manufacture of network and data center photonic components, does a good job of explaining the requirements of modern day data-centers and PICS play a key role there.

The fundamental building blocks of any fiber-based network or PIC are lasers; they are the conduit through which light-based data travels. This video from Infinera does an excellent job of describing how fiber networks using PICs work to transmit data. According to page six of this presentation from Oclaro, InP is the “gold standard” compound substrate that underpins PIC-based lasers and detectors, not only due to performance but the ability to minimize the size and footprint of components within the PICs. This Photonics Spectra article confirms as much with the following affirmation of Indium-based materials in PICS:

“The primary functional sections of an optical transmitter PIC include light-generation and amplification sections, modulation and passive optical routing sections. Fortunately, fundamental material properties of indium phosphide (InP) and related quaternary indium gallium arsenide phosphide (InGaAsP) and indium aluminum gallium arsenide (InAlGaAs) compounds enable all required functionalities.”

Another major factor driving InP’s performance advantage in PICS comes down to the material’s relative band gap qualities. Band gaps refer to the wavelengths of light a material can emit. InP possesses what is referred to as a “direct” band gap –electrons can pass through it directly and emit photons at the required wavelengths for data transmission. By contrast, a material like silicon has an “indirect” band gap – electrons cannot directly emit photons at the required wavelengths without an intermediate process. 4

Indium-based chips will likely play key roles in other cutting edge technologies such as light detection and ranging (LIDAR) systems used in autonomous, self-driving cars. LIDAR systems fire lasers and detect returning photons, using the timing of those return trips to measure distance and to render 3-D images of surrounding objects and areas.5 As opposed to competing materials, indium-based components on chips allow such systems to operate at longer wavelengths of light, permitting the use of sensors that are small enough to fit behind bumpers, can detect objects at distances at least 100 meters away, see through rain or snow, provide good resolution over a wide field and be updated at 40-60 millisecond clips, all of which are perquisites to any well-functioning self-driving vehicle. 6 Cost of such systems is the currently the only limitation to their wide-spread use.

In addition to “big data” and autonomously driving car trends, the emergence of 5G networks and their technical requirements will also support a secular wave of future InP growth. Similar to the trend in wired networks, next generation wireless networks such as 5G must operate at much higher frequency bandwidths to accommodate exponential demands in data scope and speed. Here too light-based communication offers a less time and power-consuming alternative than electronics when converting radio frequency signals into light, and then back to radio frequency signals again. AXTI’s CEO, Morris Young, suggested as much in a recent Compound Semiconductor interview where he suggested devices using silicon or even GaAs substrates will struggle to support next generation wireless networks operating at higher than 25 GHz, as 5 G will. 7 For instance, Verizon is currently testing a 5G network in Texas operating at 28 GHz. 8



All of the latter secular drivers for InP are starting to manifest themselves in AXTI’s results. In each of the past two years InP-based revenue at AXTI grew by in excess of 50% year-over-year. According to the company’s 4Q15 conference call, InP was approximately 30% of $18.1 million in total revenue during the quarter, or $5.4 million. Assuming a ~60% growth rate for full-year 2015 implies InP revenue of ~$19 million for the year. Applying a similar growth rate of 50% for 2016, which, by all indications from the company, seems reasonable, implies the substrate could contribute nearly $30 million in sales at AXTI this year, or 36% of total company revenue using the 2016 consensus estimate $79 million. This would be a remarkable feat considering InP revenue at AXTI stood at a mere $7 million three years earlier in 2013, the last year the company actually broke out revenue by substrate.

In addition to the revenue growth opportunity of InP, the substrate will considerably improve mix and margins at AXTI. Consider another innocuous, but illustrative comment the company’s CEO made in the aforementioned Compound Semiconductor interview:

"During the nascent stage of GaAs, there were around twenty substrate makers worldwide although many have now dropped out," he says. "In comparison, in InP, we only have a few. The material is very challenging but I believe we can develop it."

That comment is noteworthy because despite all of that competition during GaAs’ “nascent” stages of the later 1990s and into the mid 2000s when it typically accounted for over 60% of total AXTI revenue, the company’s consolidated gross margins tended to run in the 30%-40+% area, well above last year’s 21.7% level, per the chart below.

Thus, if GaAs was capable of producing 30%-40%+ gross margins at AXTI with 7x the number of players (AXTI has said it is one of three players in InP), it is likely safe to assume that InP-related gross margins are well north of 30%-40%+. As secular forces drive sustainably high growth in InP and its GaAs, Ge and raw materials businesses begin to stabilize, overall gross margins at AXTI should begin to track back to at least their historical levels, if not eventually exceed them over time.

BEIJING REAL ESTATE

In addition to the emerging secular growth opportunity in AXTI’s InP business, the company has a hidden real estate gem of significant value on its balance sheet in the form of 300,000 square feet worth of industrial real estate in Beijing that it owns via its wholly-owned subsidiary Beijing Tongmei Xtal Technology Co., Ltd. This location is where AXTI conducts the majority of its China-based production and where over 675 of the company’s 702 employees as of year-end 2015 were employed. Though AXTI does not generally break-out accumulated D&A for “buildings”, or specifically, this Beijing-based property, on its balance sheet, with a little digging into current and historical results we can make a reasonably accurate estimate of the latter value.

Since 2008 this Beijing real estate is the only property that AXTI has owned, implying 100% of the gross carrying value of the “buildings” account on AXTI’s balance sheet dating back to 2008 has reflected it and it alone. In addition, AXTI depreciates buildings on a straight-line basis over the course of 27.5 years. Thus, we simply divide the gross carrying value of the “buildings” account on the balance sheet for each year 2008-2015 by 27.5 for an annual D&A estimate and sum the total to derive accumulated D&A for the property of $6.9 million over this period. Prior to 2008 and dating back to AXTI’s first available 10K in 1998 the company owned both Chinese and U.S.-based real estate. For this period we assume annual D&A for the Chinese real estate is equal to China’s proportional weighting as a percentage of total AXTI-owned real estate multiplied by the year-end total gross real estate value at the company, divided by 27.5 years. Summing these values amounts to $2.7 million in cumulative Chinese-based D&A from 1998-2007, which, when combined with the 2008-2015 total of $6.9 million, implies that AXTI’s balance sheet carries $9.6 million of cumulative D&A for its Beijing real estate. If we net this against the $30.4 million in gross buildings value on the balance sheet as of 2015 year-end, it implies that AXTI carries the facility on its balance sheet at $20.8 million on a net basis.

The next step is to derive a current, fair market value estimate of the real estate. Despite rummaging through Bloomberg for pricing data on Chinese-based commercial/industrial real-estate, we could find nothing. However, such data could be found on Statista.com. According to Statista, the average selling price per square meter of Beijing-based commercial real estate in 2014 totaled 9,817 RMB, or $1,510. AXTI’s 300,000 square foot Beijing facility converts into 27,900 square meters and at $1,510 per square meter this amounts to $42 million in fair value, or ~25% of AXTI’s ~$150 million market cap.

If we simply take AXTI’s existing price-to-book (P/B) multiple of ~1.20 and multiply it by the step up in fair market value for the property of $21.3 million vs. its carrying value of $20.8 million and divide that by the 32 million shares outstanding at AXTI, we get $0.80 in incremental upside per share to a stock that currently trades at $4.40, or +20%.

FAIR VALUE ESTIMATES

Before we begin assessing the future prospects of AXTI and the appropriate multiple to apply to those prospects, let’s begin with what we know about today. As it stands today, if we sum AXTI’s cash, cash equivalents and liquid investments, its accounts receivables and inventory discounted at 50%, the fair value of its Beijing R/E discounted at 25% and net the total against all of the company’s liabilities, the resulting $83.4 million, or $2.60 per share, amounts to half the existing market cap of $150 million.

In other words, without even including any of AXTI’s less liquid assets such as $4.1 million in pre-paid expenses or ~$11 million in net PP&E outside of its Beijing real estate, the company is currently trading for less than liquidation value. This seems fairly draconian considering the company’s InP business has compounded growth at over 50% a year over the past two years with what seems like a real possibility of maintaining similar growth rates into the foreseeable future. Further reinforcing the draconian current valuation of the stock is the fact that despite the noticeable degradation in AXTI’s top line since its business peaked in 2011 it has continued to generate cash from operations every year since, including $2.1 million in 2015, while its cumulative free cash flow over the past three years has been an immaterially negative -$3.1 million.

If we want to begin assessing fair value for the stock today as it relates to AXTI’s future prospects, we need to make a reasonable attempt to forecast its business a few years forward. Assuming InP-related growth can continue to compound at 40% through 2018 and that Non-InP revenue (GaAs, Ge and raw materials) continues to degrade by 5% annually over the same period, AXTI will likely generate in excess of $100 million in revenue in 2018 vs. < $80 million this year.

Using the latter growth assumptions InP should account for ~half of AXTI’s total revenue by 2018, up from less than 25% in 2015, creating the conditions for material gross margin expansion at the company. Assuming gross profit margins for InP and Non-InP businesses are similar then as they were in 2015, the company should be able to produce corporate gross margins of at least 30%, or within the lower end its historical sweet spot range of 30%-40%+.

Historically, AXTI has generally been afforded EV/Sales, EV/EBITDA and P/B ratios ranging between 2.5x-4.0x, 18x-25x and 1.7x-3.0x, respectively, when it has experienced previous structural growth cycles that helped to drive overall corporate gross margins of 30%-40%+.

As such, we believe the latter valuation ranges on various metrics are reasonable ones to apply on our 2018 assumptions and believe the higher end of those ranges is likely more applicable considering AXTI, in 2016, has already reached our imputed expectations of reaching at least 30% gross margins by 2018, with much more upside likely to come and potentially taking us well beyond the historical “sweet spot” range of 30%-40%+ gross margins.

Regardless, applying the latter valuation ranges to our 2018 assumptions and discounting back three years with a 15% WACC (16% cost of equity) yields an average fair value for the stock of $5.40-$8.65, or +~25%-95% from current levels.

WHY THE INEFFICIENCY (OPPORTUNITY) IN THE STOCK?

There are a number of factors that have contributed to the inefficiency in AXTI’s stock price that has left it trading for liquidation value.

InP/Beijing Real Estate not Well Understood: Given a relatively new CFO, AXTI has only recently begun attending conferences and broadening its interaction with the investment community beyond quarterly earnings calls. Further, AXTI has only begun discussing the growth rates of its InP business and the fact that it is nearing a critical-mass size over the past two quarters. Thus, the market has not fully absorbed the sustained, long-term growth impact InP will have on its business. We believe the market is also unaware of the mix shift benefits from InP as well. AXTI has also never discussed the imbedded value of its Beijing-based real estate on an earnings conference call.
Chinese stigma: Though its headquarters are based in Freemont, CA and the CEO who founded the company in the mid-80s remains at the helm, AXTI’s operations are based in China and carry with them the associated stigma of being a Chinese-based business.
Corporate structure: As noted, AXTI has 10 different JVs in China that it uses to conduct its operations; thus, not only does it have the Chinese-stigma, it has a labyrinth of entities with various ownership percentages that it operates there. The investment community does not like China, much less complicated China. However, as noted, AXTI has board representation on all of these JVs.
Whistleblower: In early 2015 an anonymous whistleblower complaint alleged that the former president of the company’s China operations was conducting third-party transactions with businesses owned or operated by his family members. AXTI launched an investigation into the matter but no intentional misconduct was found nor were these transactions revealed to have improperly benefited the former president. The company’s investigation did conclude that these related-party transactions were not properly disclosed in previous SEC filings and filed an 8-K on 2/23/15 detailing those historical transactions.
Management Change: The president of the company’s Chinese operations resigned in August 2014, likely as a precursor to the company’s investigation into the aforementioned related-party transactions. At the same time AXTI appointed a new CFO, Gary Fischer. Fisher has significant prior experience in the industry including in the public company arena.
Hazardous Product Issue: On 2/27/15 the China State Administration of Work Safety updated its list of hazardous products to include GaAs, a product that was not included in the last published list from 2002. We estimate GaAs accounted for ~35%-40% of 2015 revenue at AXTI. AXTI submitted an application to the Beijing municipal authority in May 2015 to continue manufacturing GaAs and that application remains under review. In a worst case scenario we believe AXTI would have to move production of GaAs to a new facility but more likely, the company will be granted a waiver to continue manufacturing the substance.
Commodity exposure: A plot of AXTI’s stock price with an overlay of the CRB Index shows the two are highly correlated. In other words, though commodities represent only ~20% of total revenue at AXTI (and peaked at only ~25% in 2012), the investment community has painted it with the commodity brush and left it for dead.
Sources

1 http://searchnetworking.techtarget.com/definition/photonic-network

2 http://www.myidst.com/home5/international-defence-security-and-technology/technology/photonics/silicon-photonics-for-next-generation-imaging-sensing-computing-and-communications/

3 http://www.computerweekly.com/feature/Will-Silicon-Photonics-replace-the-humble-copper-cabling-in-mainstream-datacentres

4 https://en.wikipedia.org/wiki/Direct_and_indirect_band_gaps

5 https://www.technologyreview.com/s/524166/the-worlds-most-powerful-3-d-laser-imager/

6 http://spectrum.ieee.org/transportation/self-driving/how-we-gave-sight-to-the-mercedes-robotic-car

7 http://www.compoundsemiconductor.net/article/97880-inp-great-expectations.html

8 http://www.fiercewireless.com/tech/story/verizon-test-5g-28-ghz-texas-samsung/2016-02-25

Catalyst

Imminent and ongoing InP growth and material gross margin expansion.