Description

I present to you one of the world’s most wonderfully boring businesses to own, a $5-6 billion semiconductor services provider that almost no one has heard of, has never been written up on VIC, has two brokers following it, and whose earnings releases and earnings calls put me to sleep.  For a semi business, it has very low cyclicality and low capex needs yet above-industry revenue growth.  (3-year revenue CAGR of 20%, EPS CAGR of 84%.)  You get all this for the low, low price of 7-9x forward earnings.  This stock is a classic Katana core holding: It will probably never get the earnings multiple it deserves but should materially outperform the market indexes in most years, forever.  The outperformance will be lumpy.  I expect a fairly quick double out of this one soon after the fear leaves semiconductor investors, probably no later than 2H23.

Amkor was flagged to me a year ago by another VIC member in part because he knew I know semiconductors, largely through the lens of my 8-year ownership of Micron. I’ll compare some aspects of Amkor to Micron as a somewhat typical semi manufacturer to show how much better Amkor is in those ways.  I have owned Amkor for a year but don’t know it as well as I would normally know a year-old holding, because, again, it’s just so wonderfully boring. Digging into more detail doesn’t gain you much edge.

[TRADING AND NUMBERS NOTE: I drafted this pitch on a plane 11 days ago and then, well, stuff happened. It has been interesting times in crypto-land and macro-land.  Now that I’ve resurfaced in value-land, I see that AMKR’s stock price has ripped straight up 30% since then, from $20 to $26. So the $5B business on the plane is now $6.4B, and a 7x valuation has become 9x. I don't think this change matters much for a prospective long-term holder. I have already owned it for a year; price goes up, price goes down. If you are a momentum-entry value investor, you might even view this as a better entry point. The macro fundamentals are certainly better than 11 days ago, and it's looking more likely that the bottom is in. If you are a relative-to-recent-history value investor, you might worry about short-term mean reversion. Amkor has no company-specific catalysts in the near term, it's all about macro sentiment & news for the semi industry. I have near-zero opinion on whether AMKR's price will fully retrace to $20 in the next few months; the only reason I give a slight edge to "it won't" is an overall market call. My valuation thinking is fuzzy and way-up-there anyway. "It's an easy double" becomes "it's, like, an easy double."]

GREATLY IMPROVED INDUSTRY NICHE

Amkor provides Outsourced Semiconductor Assembly Test (OSAT) services: It takes the naked chips that come off a semi fab line and attaches them to other chips and components so they can be put into the products that use them (called packaging).  It also tests chips and assemblies to make sure they work. Test services are only 13% of total revenues, and I won’t mention them again.

Packaging touches almost every kind of chip and every kind of end-product.  As with memory chips, every exciting high-growth trend in semiconductors benefits Amkor; the OEM customers in the sexy segments not only grow their own units faster than the industry but also grow their content per unit faster.  For memory, that means, e.g., more DRAM per smartphone; for Amkor it means more advanced packaging services per unit.  Revenue growth for the last 5 years has been 8%, 3%, -6%, 25%, 22%.

Amkor segments its reported revenues into communications (smart phones & tablets), consumer (gaming, wearables, home electronics, etc.), automotive & industrial, and computing (PCs and data centers).  Weakness in any one industry segment is often offset by strength in others.  I dutifully input the segment numbers in my model each quarter, and I am starting to wonder why I keep doing so, because I never really pay much attention to them; the segment ebbs and flows wash out to a total number that is driven mostly by global macro trends.

Five years ago, assembly largely meant wire-bonding, literally soldering chips onto a board. This was a low-tech, low-capital-intensity, commoditized service that naturally attracted too many competitors, allowed customers to price-shop ruthlessly with few other considerations, and had inherently low margins.  The industry has migrated to “advanced packaging.”  As chips have gotten more and more sophisticated, with smaller nodes, smaller chip sizes, higher densities, faster speeds, and better power usage, less and less of them are amenable to wire-bonding.  Instead they are connected to each other, in complex and individualized packages of chips, sometimes dozens of them (called “system in a package” or SiP).  That requires packaging machines that cost 30x a wire-bonding machine; more R&D; more customization; and much more integration with customers to ensure that the packages integrate backwards to the chips and forwards to the final assemblies. (Amkor now has over 550 R&D engineers spread over 8 countries.)  That makes the customer relationships much stickier, and a lot of the smaller wire-bonders simply can’t do this work anyway.  Thus, along with being inherently higher-margin, advanced packaging’s requirements have produced an inherently more oligopolistic industry.  #1 player ASE turbocharged the oligopolization in 2018 by acquiring another large player.  Now ASE has a ~35% share, with Amkor a strong #2 at 25%.  This situation smells like DRAM did 4 or 5 years ago.

In 2017, wire-bonding (“mainstream products”) was still the majority of Amkor’s business.  Every year since then, mainstream revenues have shrunk slightly while advanced packaging has grown quickly.  By now the split is 75% advanced / 25% mainstream.

Thanks largely to this evolution, Amkor’s margins have risen relatively steadily, with some of the usual gyration for semiconductor-macro trends.  The last 5 years’ gross margins were 18%, 16.5%, 16%, 18%, and 20%, with the quarterly number peaking out at 21% in 4Q20, just before the overall semiconductor slowdown started.  EBIT margins went 7.1%, 6.0%, 5.7%, 9.0%, 12.4%, with a peak of 14.6% last Q.

Although most of the mix shift to advanced packaging has now occurred, it will keep drifting higher, and margins should keep drifting higher as well.

BETTER CAPITAL INTENSITY AND CYCLICALITY THAN SEMI MANUFACTURERS

I say advanced packaging is more capital-intensive than wire-bonding was, but it’s still not particularly capital-intensive.  To support the recent 20-25% revenue growth, Amkor has been spending 11-13% of revenues on capex.  For comparison, Micron has been spending 36-38% of revenues to support the same amount of revenue growth.

To someone (me) with a memory-chip background, Amkor’s business looks amazingly non-cyclical.  The cause is simple: no inventory whiplash.  Most OEMs keep substantial chip inventories, and their inventory management behavior is highly pro-cyclical.  In good times, they order even faster than their underlying demand is rising, because supply gets tight and they worry they won’t be able to get their hands on enough chips in their near futures.  In bad times, they begin drawing down their own inventories well in advance of their underlying demand falling.  As a result, chip makers’ new orders can collapse even as underlying usage stays pretty strong.  This effect is at its strongest right now.  Micron’s last five quarters of revenue growth were +33%, +25%, +16%, -20%, and -43%, and it is guiding for -44% this quarter.  Its gross margin has collapsed from 47% in that first quarter to 25% now.

In contrast, OSAT services have no inventory effects; demand tracks the actual underlying chip usage.  Thus Amkor’s last five quarters of growth were +24%, +26%, +20%, +7%, and +24%.  The dip to +7% in 2Q was caused by Amkor’s big plant near Shanghai being shut down by China's Covid restrictions.  The bounce to 24% was catch-up as the plant re-opened.  Without the Shanghai issue, it probably would have been a fairly clean glide path of something like +24%, +26%, +20%, +17%, +14%.  Guidance for the current quarter is +8%.  Gross margins have fallen from a 21% peak last year to ~17% now.  Yawn.

GEOGRAPHICALLY ADVANTAGED IN THIS SCARY NEW WORLD

Amkor is the only major Western-headquartered OSAT company (HQ in Arizona); all the rest are Asian.  A bunch of them are Chinese and now are basically restricted to serving Chinese customers. Amkor has only 1.1m square feet of manufacturing space in mainland China (the one plant in Shanghai) and another 9m elsewhere – Korea, Japan, Philippines, Malaysia, and a small plant in Portugal.  They are spending a lot of capex right now to build a big new plant in Vietnam; management talks about it a lot on the earnings calls.  I would not be surprised if they eventually build a plant in the U.S. as more customers keep announcing new U.S. fabs.  Thanks for the $billions, Uncle Joe.

VALUATION – A TRIPLE OR QUADRUPLE IS FAIR, A QUICK DOUBLE IS REALISTIC

Obviously the world is going to hell and a recession is either here or imminent and interest rates are higher for longer and no stock is worth anything anymore.  So what should Mr. Market pay for the profile I laid out above?  Secular growth well above GDP, still-improving industry dynamics, likely modest margin expansion, mildly cyclical, decent capital intensity, limited customer risk, limited geopolitical risk, low technology risk.  Oh yes, and zero net debt (cash = ~gross debt).  To me this screams out for an above-market multiple, call it 20x normal non-recessionary earnings.

Instead we’re at 7-9x recessionary earnings.  Remember, the semi industry is already bombed out this quarter.  Yet Amkor is telling us they will do +8% revenues and 11-12% EBIT margins (vs. 12.4% in 2021).  Looking into 2023, Micron management sounds highly confident that the upturn will start in 2Q, and Samsung and Hynix have said the same, albeit less forcefully.  They are basing those forecasts on an end to their customers' inventory drawdowns, but those customers won't end the drawdowns unless underlying chip demand is doing OK.

Consensus estimates for Amkor of -3% revenue in 2023 and $3.00 EPS look reasonable to me, maybe margin risk knocks EPS down to $2.80.  Then a snap-back in 2024; 20% revenue growth could produce EPS above $4.  20 x $3 = a $60 stock, 20 x $4 = $80, versus $20 today.  My DCF model with a 9% WACC spits out $87.

I don’t think the market is likely to pay up that much unless Amkor somehow gets a lot more investor attention.  But how about a quick double?  If a year from now the estimates are for $4 forward EPS, a $42 stock price would require only a 10.5x multiple.  Seems like a gimme.

Catalyst

The world doesn't end. The recession is merely as bad as currently forecast. The terminal Fed Funds rate is in fact less than 5%. Investors become willing to own cyclicals again at non-ridiculous multiples.

Oh, you thought I'd say something about Amkor, or even semiconductors? Nah.