Description

If you are reading through META and wondering how in the world they are spending so much on CAPEX, I am not sure I can explain.  If you are reading through META and wondering what are they and their competitors spending on, I might have an answer…  SMCI’s servers.  As you can see below, there appears to be a direct correlation between the SMCI’s sales and META’s capex.  My hypothesis is that SMCI is uniquely positioned to capitalize on the increasingly specialized and heterogenous server needs of customers like META and its peers.  

Super Micro Computer, Inc. (“SMCI”) designs, develops and manufactures motherboards and servers.  The company has 25% share in Silicon Valley but 7% globally.  SMCI’s growth has been surging and I suspect SMCI will quickly move to 20% share in an industry that has a nice growth trajectory, driven by the rapidly expanding analytical needs of corporations and governments globally and a competitive void that SMCI is filling.  I also suspect that 2023 will break a cycle where server demand slumps immediately after PCs slow.  Most importantly, the recent growth drivers for SMCI specifically are so strong that weaker demand for the broader server market is only likely to cause a modest deceleration at SMCI.  The stock’s valuation is absurd given the opportunity but is partly explainable for reasons I will address.  I am cautiously optimistic that we will hear more evidence to support my hypothesis this coming week on the 1q-23 earnings.  I suspect the stock can double as soon as it becomes clear that FY 6/2023 is likely to deliver $12+/share of EPS and that 2024 will bring more growth.  This could become clear as early as their 2^nd FY quarter or the 12/31/22 quarter.  If you stick with the write-up, I suspect you will begin to realize that SMCI is not an unwitting beneficiary of their timely capacity expansions.  They are beginning to win on a structural basis.  If so, there is material upside beyond $140/share.  Woodrow did a nice job identifying this trend back on 8/11/2020.

Why has SMCI seen voracious demand growth? 

SMCI is uniquely and perfectly positioned for the current technology transition occurring in servers.  These trends are the following. 

• Cloud is taking substantial share from on-premise server farms
• Servers are being asked to perform ever more diverse and complex tasks for 5g, AI, autonomous driving, machine learning, metaverse. 
• The variety of components that are embedded in a server is growing greatly as chip producers develop different chips and GPUs to serve these varying applications
• Servers are increasingly linked to create resource pools to share memory or processing power sharing, increasing the complexity of design

The above-mentioned trends require a server design/manufacturer that has the following capabilities.

• Extremely robust R&D capabilities to quickly adopt new chipsets and configurations
• Design capabilities to mix/match components to meet a customer’s increasingly varied needs
• Flexible manufacturing with scale to meet these varied requirements
• Design capabilities to provide green (low power) solutions

Historically, SMCI always operated with the above-mentioned principals.  Dell/HP dominated the historical landscape as they took their time to develop a limited set of server SKU’s that were well designed and flexible enough to meet most customer’s needs.  The limited server SKU could be built at massive scale to keep equipment cost low.  Since few of the customers had the IT capabilities to design and operate a server farm, DELL/HP added those services and competed primarily on the quality of their service and support.  For customers with the internal resources to operate a server farm, they could turn to Original Design Manufacturers (“ODM”) who could meet their server needs on a an extremely limited SKU basis but at an extremely low cost.  ODM’s were purpose built to take a design (from someone else) and build servers cheaper than anyone.  ODM’s were never well qualified to provide assistance on server design or opine on whether the customer’s equipment would integrate well or serve its purpose.  In this pre-cloud epoch with server homogeneity, this provided SMCI with the following niches to exploit.  To begin, SMCI’s competitors were ponderous and slow to adopt chipsets as they wanted a perfect design to build at scale in China.  What If you were in a bleeding edge technology industry and needed the most advanced chips?  SMCI was able to adopt new technology 6-18 months faster due to their extremely robust design capabilities and flexible and local manufacturing.  What if you operated your own sever farm and wanted to blend components and create a bespoke solution?  SMCI was the only operator that would be willing to assist in the design and delivery of these servers.  The ODM’s never offered design and DELL/HP couldn’t be bothered to develop a lower volume SKU.  On this basis, SMCI was a small operator that steadily gained share from a low base as the % of customers who wanted more specialized servers grew faster than the market.  SMCI never offered service/support once the servers were installed. 

Then two things changed.  The first was the invention of the cloud.  The cloud allowed SME and even some larger operators to outsource their server needs to cloud operators – greatly reducing the expense.  As the cloud exploded in popularity, DELL/HP position was usurped as an enormous cloud operator doesn’t need DELL/HP support/service to run their server farm.  As customers turned to the cloud, DELL/HP lost material share.  SMCI and the ODM’s were able to step into this void.  The second shift was that growth in the number of customers who were competing  on the basis of their ability to process massive amounts of data.  The servers grew more complex with GPU’s, a growing variety of new chips, resource pooling and green solutions.  In a different world, DELL/HP might have been agile competitors but their process was designed to work on limited sku and to sell service/support.  As DELL/HP lost share, they have been retrenching as opposed to adapting to the new regime.  The ODM’s were never set up to provide design advice or batch manufacturing.  As chip and component heterogeneity exploded to meet the varying analytical needs of customers, SMCI emerged as a go-to design leader uniquely positioned to capitalize on this trend.  SMCI is a preferred partner of NVDA.  A tour of SMCI’s manufacturing plant reveals their customer list which is the who’s who of metaverse, AI, autonomous driving, machine learning and 5G networks.  As I mentioned initially, SMCI has 25% share in Silicon Valley but 7% share globally.

In another example of their thought leadership, SMCI created Rack Scale.  The typical server sale is done with hundreds of units which the customer must assemble in the server room.  There is a mess of wires and material assembly time.  Rack Scale is a plug and play solution to address this issue (see picture below.)  The servers are built into the rack which come in building blocks – greatly reducing server room assembly time.  Rack Scale has substantial benefits beyond this.  Rack Scale reduces power consumption and facilitates processing and memory pooling – a key new concept.  Rack Scale also provides “future proofing” of the unit as the customer can insert new processing/memory.

In summary, SMCI is stepping into a competitive void and addressing rapidly growing needs.  Said differently, SMCI is offering a better product to a growing niche at a highly competitive price.  I may have oversimplified and generalized in the above.  Further, I don’t want to oversell.  SMCI is competing in an evolutionary matter rather than revolutionary.  They are selling design, flexible manufacturing and an evolutionary service with rack scale.  This is not a Tesla Model S.  Charles Liang (Founder & CEO) wants to dominate the industry and he appears to be keeping prices extremely low given the demand growth and less appealing offerings of his competitors.  I would also highlight that while it may be growing fastest in the unique design area it still sells a ton of commodity or relatively homogenous servers.  In addition, DELL/HP’s ASP has been moving up partly due to component inflation but also because they are adding in GPUs and other new components into their designs.  SMCI has pole position and a several lap lead but this remains a race.

Another way to frame this – if you are working at AWS or META or Roblox or TSLA, you are aware of all of these new capabilities.  GPU, DPU, CXL, resource pooling… and you want someone to design a server or farm to exploit all of these technologies and a green solution.  You can’t call the ODM’s as ODMs design a fairly simple basic server which they can produce at massive scale.  You can’t call DELL/HP any more as they largely rely on the ODM basic design.  Further, they aren’t interested for a smaller unit sale and will force a service/support contract on you… and DELL/HP will take a decade to get back to you.  Your best call is SMCI.  Now – this might be slightly over-stating SMCI’s position – and it may understate the ability of these internal teams at the customer to design the system themselves and order from the ODM’s.  However, even if you have a capable team, they might want SMCI to assist in the design too as another set of expert eyes – especially as SMCI’s solution is the same price.  See links below for two examples of how the server room is getting more complex.

https://blogs.nvidia.com/blog/2020/05/20/whats-a-dpu-data-processing-unit/

https://www.rambus.com/blogs/compute-express-link/

Frankly – I am not a server expert.  I have been following this stock for a while and the above realization took a while to grasp.  I went into length to facilitate understanding.  I welcome push back here.  However, this conclusion coalesces with Woodrow’s description

What has this meant for the financials and current outlook? 

As you saw above, SMCI’s topline has surged.  This has taken investors by surprise.  However, there is one person who forecasted this – Charles Liang.  As you can see by the investor deck from March 2021, Charles/SMCI laid out a hyper aggressive plan. 

https://s25.q4cdn.com/632471818/files/doc_presentation/2021/Exhibit-99.1-Investor-Event.pdf

He forecasted a surge in growth.  He bet aggressively by investing in a large new manufacturing and design facility in Taiwan.  Then he doubled down by eliminating his salary and bonus and converting all of his compensation to hitting a variety of highly aggressive targets.  In hindsight, one has to wonder if the board and investors were picked off by this compensation package.  However, I would take this arrangement all day over standard compensation.  The forecasts seemed preposterously aggressive at the time but they will be met years in advance and at this point appear preposterously conservative - EXCEPT FOR the stock price which has lagged revenue growth.

In fairness, Liang was aided by some server market basics which he potentially foresaw.  Most server sales are done to replace obsolete servers which typically last 5-7 years.  In 2020, server demand stalled as the economy downshifted and installing servers requires employees to be on premise which was complex during the pandemic.  As a result, customers “sweat” their server farms in 2020 and didn’t replace them as fast as normal.  At the same time, the work from home trend and surging demand for analytics created a boost in demand.  Liang’s decision to add capacity was on point as demand increased. However, the component shortage and supply chain issues held back SMCI results substantially on a trailing 12 month basis.  While we face a looming recession and PC demand has rolled, I would argue the ensuing years should be different.

• PC demand was not driven by an improving economy or technology shift.  There was a unique surge in demand caused by the pandemic.  This cycle has a life of its own.
• The PC demand surge limited the ability of server producers to meet demand.  The server fleet was not renewed substantially in 2020, 2021 or 2022 as the asset “sweat” of 2020/2021 turned into a component shortage in 2022.
• Demand for analytic processing has surged in recent years as you can see by the CAPEX spend at META, SNAP and others
• Lastly (and possibly most importantly), the new design and capabilities of these servers is rendering legacy servers obsolete far faster than before.  For instance, the energy savings alone in places like Europe can drive an incredible IRR for operators before considering the enhanced processing capabilities. 

A powerful example, META’s capex on a LTM basis is $27bn and will move to $35-39bn in 2023.  A huge portion of this is centered on data centers (see below).  The server industry is $90bn a year which means that META’s spending alone likely powers substantial growth for the entire industry.  My suspicion is that META’s demand alone could drive material revenue growth for SMCI.

• META 3Q – “Capital expenditures, including principal payments on finance leases were $9.5 billion, driven by investments in servers, data centers and network infrastructure.”
• META 3Q – “Before turning to our CapEx outlook, I'd like to provide some context on our infrastructure investment approach. We are currently going through an investment cycle, which is being driven -- which is primarily driven by two large areas of investment. First, we are significantly expanding our AI capacity. These investments are driving substantially all of our capital expenditure growth in 2023. There is some increased capital intensity that comes with moving more of our infrastructure to AI. It requires more expensive servers and networking equipment and we are building new data centers, specifically equipped to support next-generation AI hardware.”

Clearly, META is likely the largest spender in this category but their increase in spend seems more indicative of the trend in companies who compete on analytics and need SMCI.

Lastly, there are two large chipset releases coming up which should provide a substantial ballast to SMCI in 2023 as it will be first to adopt these chips into their systems, driving early adopter sales.

Valuation/Hair:

Where does SMCI trade?  Today SMCI trades at a hair of a premium to DELL/HP – the structural market share losers with limited growth who appear to be competing on an increasingly less relevant basis.  On a PEG basis, the disparity is absurd. 

SMCI was hit by two events in 2017/18.  The first was a self-inflected.  Their old CFO had an immaterial accounting irregularity on revenue recognition which caused a lengthy restatement and an exchange delisting which forced most sell-siders to drop coverage.  The historical multiple of 20x+ P/E moved to 10-12x. 

At the same time, the stock was hit by TWO articles by Bloomberg that alleged that SMCI was an agent of the People’s Liberation Army of China (“PLA”).  As a result, when Charles Liang got up to the plate at his analyst day in March of 2021 and pointed to a 1,000 foot home run – nobody noticed or cared.  The stock has followed the earnings upward but with a lag which has further compressed the multiple to 8.6x estimated EPS of $8.0/share on FY 6/2023.  $8.0/share for June 2023 represents the lower end of management guidance.  The interesting part here is that SMCI has been blowing the doors off estimates.  The other day it pre-announced 1Q-23 (9/30/22) and stated that it would produce EPS of $3.05-$3.20/share versus $2.18 of expectations.  If history is a guide, when they report this week – they will beat $3.20/share.  If so, they will have produced nearly half of the low-end of the annual guidance in the first quarter – which happens to be a seasonal low-point.  In my conversations, management has suggested that the only item that could cause risk to their guide would be a component shortage.  The slow down in PC’s and crypto mining has all but guaranteed this won’t be an issue. 

It is my supposition that the market hasn’t been paying attention to this company and what has changed.  This is especially true as HP/DELL are hated even by value investors.  Nobody is paying attention because they aren’t forced to follow DELL/HP and there isn’t really another company in the industry worth following.  As such, the market suspects that SMCI was a windfall beneficiary of the component shortage and a surge in demand and will fall victim to the obvious downturn in spending.  Hopefully, I am correct and total spend for servers continues to grow and SMCI takes an increasingly large amount of share.  If so, SMCI is likely to continue to push forward at $2.5-3.5/share of EPS each quarter.  The added scale and value add should present in higher margins than they forecasted in March 2021 and they are likely to need more capacity to fulfill demand as they grow into their leadership position.  As these trends continue and the path to 20% share becomes more obvious, more people will pay attention.  My strong suspicion is that the multiple expands to mid-teens or higher.

To address the hair, the accounting issue was immaterial but a huge distraction which limited customers ability to engage as they were off the exchange and not filing financials.  Liang is ruthless and perhaps in a bear case he was overbearing which drove his employees to make this enormous mistake.  Given the impact, I would be shocked if Liang allowed this to occur again. 

As to the spying allegations, I would say the following.  Bloomberg alleged/insinuated that SMCI cooperated with the PLA to embed spying/listening devices onto the motherboards of its servers.  The PLA used these devices to spy on the DoD and other corporations. 

Liang is of Taiwanese heritage and has far more manufacturing in the US and Taiwan than China which is highly differentiated from his peers who manufacture mostly in China.  If there was a need to remediate the manufacturing process, SMCI was in a far better position to structurally address this issue.  Liang is the largest shareholder in SMCI and would basically lose everything if the allegations were true.  History is filled with people who risked it all to spy though.  Could Liang have been compromised via coercion or did he grow up a closet PLA sympathizer?  Perhaps.  However, the scrutiny on Chinese spying has increased dramatically since 2018.  The high profile allegations had to trigger an epic investigation by the NSA and the security apparatus of the U.S as SMCI’s servers are used by the DoD.  A conspiracy that started with Liang and filtered through the organization would have a web of participants and a trail of evidence.  If real evidence existed to support a spying scandal, SMCI would never have been allowed to grow its position like it has.  It is also worth mentioning that many customers came out at the onset of the allegations in support of SMCI – an unusual step to take.  My conclusion is that there is a reasonably high probability that the PLA infiltrated SMCI’s plants in China without the aid of senior management and planted spying devices in their servers.  SMCI was warned about the issue and moved immediately to remediate the issue.  It is equally likely that other server manufacturers were unwitting victims of the same infiltration.  In short, my conclusion is that with all of the scrutiny on Chinese spying that this issue is behind SMCI either because the allegations were false or they effectively remediated the infiltration.  However, the issue does present “tape-bomb” risk or worst case …maybe it just takes 5 years to build a case.  However, if a case was building, my strong suspicion is that these large and highly strategic customers would have been strongly encouraged to use other server solutions.

Stock Price Target:  I suspect that SMCI can produce $12/share of EPS in FY 6/2023.  It will likely be able to grow in FY 2024 too.  If I assume a post scandal average P/E of 11.0x, that implies a target of $132/share or 94% upside. 

Catalysts:

• Continued beat/raise drives more attention to the stock which raises the multiple due to trend following or because more investors start to realize the shift has occurred

Risks:

• Recent results were highly cyclical and not a function of structural market share gains
• Spying tape bomb
• Devastating hit to the global economy

Catalyst

Catalysts:

• Continued beat/raise drives more attention to the stock which raises the multiple due to trend following or because more investors start to realize the shift has occurred