SMCI Q1 2024 earnings

Super Micro reported fiscal year Q3 2024 earnings or the report for Q1 of calendar year 2024.

Guides vs Actuals
Revenue 3.7 - 4.1B guide, actual 3.85 (analysts estimated 3.89)
Adj net income 5.20 - 6.01 guide, actual 6.65 (analysts estimated 5.84)

They are raising the guide for the fiscal year from 14.3 - 14.7B to 14.7 - 15.1B, and note that this fiscal year only includes one more quarter. So they are effectively raising guidance for the next quarter by 400M.

This quarter was partially supply blocked which impacted revenue and CEO Liang said revenue would be significantly higher if not for the supply issues. In the last week of the quarter they finally received the supply they needed, but because of receiving the supply this quarter it hurt inventories and cash flows significantly. They also raised additional convertible notes and additional shares to fund their ambitious expansion plans. Their revenue guide for next quarter now is jumping up massively to 5.1 - to 5.5B in revenue, and Adj net income is projected significantly higher QoQ at 7.62 - 8.42 for the next quarter.

Some additional notes from the conference call,

CEO - Charles Liang

  • Revenue 3.85B up 200% yoy, adj EPS 6.65 up 308% yoy
  • “Continue to face supply chain challenge due to new products that require new key components”
  • Believe the situation will gradually improve in the coming quarters
  • It was mentioned that the missing components were not from Direct Liquid Cooling (DLC)
  • DLC is finally ready for high volume production
  • raised 3.28B in the quarter, will have minimal equity dilution
  • “If not limited by some key component shortage, we could have delivered more
  • YoY operation margin and net income continue to improve
  • DLC will save energy costs 40% at data center scale
  • Focus on delivering newer chips like Nvidia H200, B100, B200 GPUs, as well as Intel and AMD GPUs
  • Production teams are making aggressive progress on new Silicon Valley, Taiwan, and Malaysia factories
  • Confident Q4 revenue will be in the range of 5.1 - 5.4B” (This would be +38% qoq!)

CFO - David Weigand

  • Quarter growth was led by AI GPU platforms and was over 50% of revenue
  • Customers come primarily from enterprise and cloud service providers
  • 1.88B in enterprise channel vertical, 49% of rev (40% last Q), +190% yoy, +26%
    qoq
  • OEM appliance and data center revenue was 1.94B, 50% of revenues (59% last Q), +222% yoy, -10% qoq
  • (Still trying to understand why enterprise is up so much, and data center is down qoq, maybe supply issue?)
  • One large CSP data center customer was 21% Q3 revenues and one existing enterprise channel customer was 17% (These numbers were 26% and 11% last Q)
  • US 70% revenue, Asia 20%, Europe 7%, RoW 3%
  • YoY basis US +242%, Asia +242%, Europe +30%, RoW +87% (Curious if anybody knows why Asia is growing much faster than Europe?)
  • QoQ Asia +17%, USA +3%, Europe -3%, RoW -11%
  • Adj gross margins 15.6% - impacted by supply issues
  • Focus on strategic new designs, gaining market share and improving manufacturing efficiencies
  • Operating expenses increased 14% qoq, 72% yoy driven by higher compensation expenses and headcount
  • Adj operating margin 11.3%, in line with previous Q
  • Closing inventory was 4.1B, 67%+ qoq due to purchase of key components
  • CapEx was 93M, free cash flow of -1.6B in the quarter (this is the inventory landing huge in the last week of the Q)
  • Cash 2.1B, liabilities 1.9B, net +252M
  • Days of inventory increased by 25 days to 92 days compared to prior quarter of 67 days, due to two key component purchases for higher next Q revenues
  • We expect gross margins to be down sequentially as we focus on driving strategic market share gains

Q&A

  • “Congrats on the strong guidance”
  • TCO on liquid cooling saves the customer 40%
  • Customers pay a very little premium but save up to 40% of energy costs
  • “June quarter, we are preparing more than 1,000 liquid cooling racks for those early birds. And I believe that demand will continue to grow very strong”
  • Target gross margins 14-17% long term range
  • “Had to increase” inventory to meet next quarter demand
  • Inventory growth tied to sales
  • For September and December quarters, “we will have very strong growth”
  • “I believe this strong growth will continue for many quarters to come, if not many years.”
  • With AI demand growing so strong, will be able to “grow sequentially”
  • Malaysia factory coming online later this year
  • Analyst “You guided the June quarter to increase by 1.6B QoQ
  • Inventory build up is also to prepare for liquid cooling
  • Capital raise is preparing for next quarter shipment
  • We could have shipped more if we had more parts
  • Supply chain is an ongoing problem, but it continues to improve each quarter
  • We have an “very ambitious brand” and growing 3-5x faster than the industry average

Overall this is a more mixed report than I was expecting. They met revenue, beat EPS and then made a very strong next quarter guide which was raised 400M. EPS is also guided much higher.

However, some problems were faced this quarter with supply blocks. They are raising extra capital for expansion factories. I like that the company is playing for the long term here and seemingly not caring if all the inventory lands in the last week of the quarter. In the short term it is somewhat understandable the market does not like this report. Sounds like they are aggressively going after market share and ramping up significantly.

45 Likes

If you are going to own a stock like this, you better fully comprehend the competitive landscape. I talked to an owner today who had no idea what was meant by ODM. If you don’t know exactly what that means, you should.

3 Likes

Acronyms really only make other people feel smart because somebody doesn’t know what they mean. It really doesn’t mean anything other than Companies like Quanta, Wiwynn, Dell, and HP, that compete with SMCI. ODM means Original Design Manufacture.

Andy

22 Likes

Whatever. But how many investors understand that business model? I find few that I’ve talked to that hold this stock.

1 Like

Care to expand on the competitive threat to Super Micro you are talking about?

What does an Original Design Manufacturer have to do with it? So this refers to companies that design and manufacture products that are sold under another companies brand. From what I’ve found online Super Micro may use ODM services or act as an ODM for other brands.

Are you trying to say because their AI servers have Nvidia chips in them that they are a just a reseller and don’t have a competitive advantage?

Super Micro doesn’t use that many acronyms compared to other companies from what I’ve discovered. I’ve just spend a ton of time trying to parse all the acronyms from TransMedics call, and I doubt anybody on this board knows all those acronyms. Seems like we should be more focused on the revenue and EPS growth than who knows what acronym.

20 Likes

Kind of heard the same thing that NET had inferior technology, or that AMD and Google etc would take out Nvidia with open source and ASICs.

We are passive investors. Not industry insiders. And that is to our benefit as industry insiders see a few trees and become emotionally attached to what they do and myopically view things.

SMCI can sell everything they can produce. They will make more than $20 a share after Q4 next quarter for the year and may very well surpass $40 a share for 2025 that starts in the August quarter. All this before water cooled servers start to become material (well next quarter should be dust volume sales of water cooled).

SMCI is in a very competitive market. Yet they continue to out grow this extremely rapidly growing market. When that subsides then we may need to evaluate why. As of now no one can tell you why SMCI has done so well other than a few talking point advantages they have, and no one can reasonably articulate why SMCI won’t continue to prosper.

NET vs cannot even remember the name. Insiders on this board swore they had better technology. Problem is THEY HAD NO CUSTOMERS! So much for superior technology.

I’ll follow the narrative and the real world facts that buttress or that are inconsistent with that narrative. Anyone see anything in the guidance that does not buttress the investment narrative?

If so bring it on. But don’t tell me I need to know the nitty gritty details to smartly invest.

36 Likes

Like WPR stated, SMCI is called a white box also. It makes little difference who makes the box if they can’t sell it. Anyone following this company would know that their margins are so low it has to be a commodity business and when SMCI has a personal relationship with NVDA, which those other ODM’s do not, it makes a big difference. So its all about making money while the sun shines because these days won’t be around forever.

Andy

20 Likes

I view SMCI being the technology leader in a business where margins are typically compromised by the “reference design” and “ODM” model. In this model, established by Intel back in the Pentium processor days, which I was highly involved in, the CPU maker (or GPU in this case) holds nearly all of the IP and differentiated technology. They typically have 1-2 lead customer partners that allow them to introduce into the market most efficiently each successive next generation of processor. I consider SMCI to be a key partner of NVDA in this model.

However, NVDA is clearly proceeding down the path of the ODM/Reference Design model. This model allows people with much less design know-how to manufacturer and supply boards/boxes into the market with large factories in Asia. ODMs are tied to factories, and those factories need to be full. Thus the model always reverts to the mean of being a “low margin business” for all but the Processor supplier.

Dell and HPE, do very little technology development any more, as this model above provides them with the technology they need from a HW standpoint and they rely on service/support and software as their differentiators. They let the ODMs do much of the hardware.

SMCI should remain the best of the best, working tightly with NVDA on each new technology node. But I would expect that their growth will hit limits well before NVDA’s. As Intel would roll out new technologies (CPUs and Motherboards -and same for AMD later), they would then work to enable the rest of the market to build the motherboards and boxes. The “rest of the market” would commoditize these. The CPU chip would remain a high-margin business for Intel/AMD while those building the motherboards/boxes would crush each other on margins.

SMCI lives somewhere between the NVDAs and the ODMs. They should (for sometime) have a good business but many investors I talk to have a complete misunderstanding of the business model and no idea what an ODM is. SMCI has top level systems hardware talent and enjoys a relationship as a key partner to NVDA. But there will be market constraints of where that can take them.

Like many posts on here, the reaction from highly invested individual is to lash out at any suggestions contrary to their thesis. TIFWIW - or not. Group think has gotten the best of many of us, including me.

22 Likes

Not sure what a compromised margin is? Sure, their gross margin is 15% right now, but they are GAAP profitable with this business model, unlike most SaaS which has 80%+ gross margin but is still unprofitable.

This model allows people with much less design know-how to manufacturer and supply boards/boxes into the market with large factories in Asia

Okay but this is not happening right now. In fact, sales in Asia are growing faster than anywhere is the world. How you can you explain Super Micro’s growth in Asia, if there are supposedly large factories producing the same technology?

Dell and HPE, do very little technology development any more

That’s the point. That’s why Super Micro has a competitive advantage because Dell and HP are not innovating, and Super Micro is.

As Intel would roll out new technologies (CPUs and Motherboards -and same for AMD later), they would then work to enable the rest of the market to build the motherboards and boxes

Intel is also a partner of Super Micro, this seems like misunderstanding of the relationship between Super Micro and Intel.

hey should (for sometime) have a good business but many investors I talk to have a complete misunderstanding of the business model and no idea what an ODM is

What investors are you talking to about Super Micro who don’t understand the company? It’s on a different Fool board, or elsewhere?

Like many posts on here, the reaction from highly invested individual is to lash out at any suggestions contrary to their thesis. TIFWIW - or not. Group think has gotten the best of many of us, including me.

Where are you seeing group think here? The board has largely ignored this company because of it’s low margins, but it’s been a 8-bagger in the last year.

I’m looking at the growth in revenue and EPS and it’s a magnitude bigger than most companies the board is interested in.

21 Likes
  1. margin is 15% right now, but difficult to maintain as NVDA enables the ODM/Reference design manufacturers
  2. End market is buying what they can buy right now. item 1 is being enabled, it’s coming, but not here yet.
  3. Yes, that’s their competitive advantage today, but NVDA is going to roll out the enablement of many competitors and that will make it difficult to differentiate.
  4. Of course Intel is a partner as is AMD. I can assure you I do not misunderstand the relationship here. SMCI uses CPUs and GPUs.
  5. Investor group I am involved in, Social Media… I do not bother with any other MF boards. But I’m confident folks on THIS board have invested in SMCI and are not understanding how they fit into the eco-system being repeated for AI as it was done by Intel.
  6. I’m looking at Growth and EPS sustainability. Not the rear view mirror.

edit: I’ve posted before on SMCI and why I see limits to where they can go from here. But really, my only interest is as it relates to NVDA. I have been asked by others about the read thru from the SMCI report to NVDA and its certainly not as clear as day, but I believe NVDA will sell every chip they get at these crazy margins for some time. Some concern that the supply may not be keeping up with their growth expectations in the near term.

10 Likes

Super Micro has a building blocks architecture which allows the customer to customize the server as much as they like. For example they have TelCo customers who have very specific specifications and form factor sizes, and they can build to any spec the customer wants.

  1. I’m looking at Growth and EPS sustainability. Not the rear view mirror.

The company is saying they will grow sequentially for the foreseeable future, and that includes lapping quarters that are seasonally weak and strong. As I highlighted above, this is a direct quote from the CEO,

“With AI demand growing so strong, will be able to grow sequentially”


So in summary you are saying Nvidia is Super Micro’s main competitor who will lower their margins?

Can you name any other company that is even competing with Super Micro? Let’s rule out HP and Dell because as you pointed out these companies are not innovators.

6 Likes

Foxconn, Inventec, Quanta, Wistron are a few I’ve personally supported in this space…

And please cease with the quizzes “can you name”?

My point has simply been that while SMCI can differentiate themselves with some customers, there will be margin pressure as the ODMs come more on line and are able to supply into the market. NVDA is enabling that model and when there are more ample supplies of GPUs, these ODMs will become much more active in the market.

4 Likes

Any thoughts from anyone here regarding the extent to which $NTNX is (…or is not) a competitor?

There seems to be a consensus that forward revenue growth is projected to be (much) higher for $SMCI.

However, I see a bit of a conundrum based (…admiedly, trivially) on P/E.
P/E for $NTNX is $57 vs. P/E for SMCI is only $40.

Is the market saying $NTNX deserves a higher growth premium?
If so, how is the market reaching that conclusion, given that $SMCI has better growth numbers?

I’m not invested in either.

1 Like

NTNX is a Software company with recurring subscription revenues. There is no comparison.

3 Likes

MFChips: I posted earlier about these designs NVDA is enabling because I saw it on a podcast that I now no longer watch and got blasted that this has been common practice for like, ever. Or is this something new and I am confused again?

At the time, I researched all the other companies beyond HPE and DELL I could find and their revenue was de minimus.

Furthermore, as I earlier stated, it appears that part of SMCI’s strategy may be that lower margins that they can sustain and profit from via scale will make it very challenging for other competitors to come online. Working so far.

So what’s hard to understand about that business model? I think what you meant to say maybe was that we don’t understand the competitive landscape. But what it could be is: we do, we just have a different opinion than you about how it’s going to play out.

I share many’s belief that this will not go on forever but as long as data center build outs keep coming, SMCI will keep growing. I just read an article about Saudi Arabia wanting to do massive data center build out - chips and servers to do that. We have barely touched on sovereign build outs.

I thought it was a great quarter - last time they had issues due to supply constraints was a few quarters ago, IMHO those are good problems to have because when the constraints end, the sales shoot up.

Intjudo, it is my understanding that NTNX builds out largely for on-prem and they offer both DELL and SMCI and possibly other server solutions. So based on what I have read - and I am sure there is someone on the board who will correct me if I am wrong - they are kind of a level above, primarily focusing on data centers on prem vs. the cloud - so not a competitor.

8 Likes

Early days of AI had NVDA being very picky about who they engaged with as this is a support-intensive business (GPUs, plus networking etc). They engaged with some of the AI Kings along with SMCI. There’s a close relationship due to geography, not to mention nationality etc. SMCI is clearly a close partner, highly trusted by NVDA.

I now see evidence that we are entering a different phase whereby NVDA is rapidly bringing up both reference designs and Tier-1 ODMs. This will enable a whole host of other suppliers, many that own large factories that they will want to run at close to capacity. This is a model that was adopted by Intel many years ago and soon all Laptops (whether sold by Compaq, Dell or IBM) were in fact designed and built at ODMs (circa 1990s). I managed a design team in Taipei at that time.

Phase 2 will be constrained by how many GPUs will be available, so how this effects overall business is quite unknown to me for the near term. I imagine that it won’t make a huge difference in the near term as the Kings of AI and SMCI will get priority for the vast majority of chips available.

But I think we should look at the long history of the Data Center hardware to be our longer term guide. The hardware suppliers have never enjoyed high margins for any sustainable length of time. The Chip suppliers were the ones who had some control over margin, namely AMD and Intel. As the DC’s matured, there was very little differentiation in hardware across the supply base, and thus gross margins were difficult at best.

I imagine that SMCI will benefit for some time still as they are better innovators than most and just having allocation from NVDA is worth a lot. But investors beware, the future will not look linear to the recent past.

9 Likes

I’ve read your posts with great interest. You obviously know more about this business and market forces than many on this board, certainly more than I. I have a substantial position in Supermicro, maybe due to naivety, at least in part.

But, I would like to get your opinion on something that I think is quite relevant - but again, maybe I’m wrong.

I was employed in IT at Boeing for 30 years. I retired more than a decade ago, so obviously a number of things have changed. But my question is not really technology related. It has more to do with how large companies like Boeing do business.

Even though I was engaged with software through most of my career, I was not ignorant of hardware and networks. I was one of the key managers when Boeing moved the majority of their business applications from Big Blue mainframes to distributed Unix servers.

There was extraordinary concessions that the hardware vendors were willing to make in order to win Boeing’s decision. The reason for this was that once Boeing committed to place a very large order from a particular vendor for a critical infrastructure component, that commitment went much deeper and longer than the initial purchase order. Any vendor that won the initial provisioning contract could pretty safely assume that they would win the contract for several many refresh cycles to come. A would be competitor had a monumental task before them to break that relationship - it wasn’t impossible, it happened occasionally, but it was a rare occurrence.

Boeing had completely adopted the Japanese model of considering key vendors as business partners rather than disposable suppliers that could be easily cast aside when a competitor offered a marginally lower price, or some bell or whistle that hadn’t yet become commonplace in the market. Technologically, vendors tended to leapfrog one another constantly, it was rarely beneficial in the long run to switch based on innovative technology. Only technological advances that were of a quantum leap nature warranted serious consideration.

This strategy of forming long lasting partnerships with key vendors was not unique to Boeing. Many, if not most large enterprises had adopted this mode of conducting business with suppliers.

This was true years ago when I worked at Boeing. I had assumed that this was precisely the reason Supermicro was willing to sacrifice margins in order to win share. Am I wrong about this? Is it now such that every time a refresh is required (I don’t know what the current cycle is, when I was at Boeing it was generally five years) all vendors have the same starting line? Has the strategy of forming persevering partnerships given way to brutal competition based on price for every contract?

17 Likes

The traditional Data Center platforms (non including AI Training) has been largely commoditized for many years. Hardware margins for companies like IBM, Dell, HPE and others have been rough. There has been very little room to differentiate.

In AI right now, there is room. But with so much of the key technology coming from 1 place (NVDA), it’s not likely to look completely different in the future. Even AMD or Intel (or startups) based platforms will look largely similar to each other. All of them will develop “Reference Designs” that are given away for free. Again, the full design is given away for free to propagate. Thanks Intel. Not a model that benefits the hardware producers that buy the chips. And then there is NVDA themselves making “walls” and complete systems.

7 Likes

I agree with you, contrary to others here. What we are talking about is vendor stickiness. Companies purchasing these systems for their private install do not want the schedule risk or to spend the effort to qualify additional vendors (unless you are the government or a large hyper scalar). If there is a significant price discrepancy, yes this will move them to try a different vendor.

Now out the other side of my mouth. Today with the hyper scalars, the magnitude of the purchase and the differentiation that their customers may require (e.g. Google’s mixed hardware cloud) will always require they have a secondary fully qualified supplier. And then they can do price tradeoffs.

The market now is a bit more complex.

-zane

1 Like

While the cooling options SMCI has built out for the AI servers doesn’t actually give them a moat I think for now it keeps them a step ahead. They have built cooling towers and Racks with liquid cooling in them and said on the call one of their customers had already ordered 1000 racks of DLC ( Direct Liquid Cooling)Racks and can now produce 2000 DLC racks per month and that will be going up.

Andy

8 Likes