Bear's Portfolio through 06/2024

Investment Analysis Clubs Saul’s Investing Discussions
1

Important context for my portfolio reviews: I run a concentrated portfolio and WARNING the swings can be huge. From the 2021 high to the 2022 low, my portfolio fell more than 60%. For every $100 I had at the top I had just $40 left! Staggering. So, before trying this style, even with a small portion of your total net worth, please understand the downside – it’s much steeper than if you own an index, or a bunch of megacaps. Also, don’t follow or copy me, Saul, or anyone. We may sell a position or buy a new one at any time, so it’s impossible to follow anyway. Also, to succeed with a concentrated portfolio, you must rely on your own decisions.

Well if my above disclaimer sounded silly last month, now it’s even more so, because the portfolio has gotten even less concentrated. I added 6 new positions! And I only sold out of 1, Samsara. I mentioned IOT Samsara Earnings - #10 by PaulWBryant that Samsara’s price looked reasonable (around $28) but I didn’t have the confidence to add. The price has been up – almost 20% higher as of the last couple days, so I was happy to let it go. I was iffy on it when the price was reasonable, so I’m happy to be out of it when it’s expensive.

I also got rid of my ETFs. I’m just more comfortable with individual companies, and I think it gives me more opportunities to add and trim. But I have been needing more of them. This month I found 6 new ones.

Of the 6 positions I added, Amazon is the one I have the most confidence in. It’s not growing super fast, but I think it’s about as steady as they come, and since I expect continued expanding margins I think the price is reasonable. Paycom is a beaten down former SaaS darling, and growth has slowed mightily for them. They talked in their last CC about re-acceleration in the second half this year, so I hope that will be a positive catalyst when they give Q3 guide on their next CC. Zoom, Twilio, and Docusign are all value plays. They’re beginning to right-size their spending which is leading to rapidly improving FCF. They’re all cash-rich and doing share buybacks. This is all OT for Saul’s board, which is why I haven’t posted on any of these 5. The 6th is Supermicro which I’m sure everyone knows about by now.

The market was up a few percent in June and is now at +15% or so for the year. Such a good first half after a very strong 2023! In June I took some money out of the market permanently – a good thing to do opportunistically when you’re retired and living off your portfolio!

The companies I’m invested in, and how they did this month, and what I did

Axon Axon reported an excellent quarter on May 6. It sold off a little afterward for some reason, and I added. Axon was up just a few percent in June, and I didn’t add or trim much.

Amazon There’s not a ton to say about Amazon, other than that AWS seems to be re-accelerating, and that the company has rapidly expanding profits. Seems about as safe as it gets, and I was happy to take a medium position right away, to which I can add on dips (with intent to then trim opportunistically later). AMZN was up almost 10% in June, but you never know, maybe a dip/opportunity is right around the corner.

Monday The more I’ve thought about the quarter Monday reported in May, the more I like it. Yes, price increases are a lot of the reason they’ll have some strong quarters coming up, but I think the OK revenue growth and rapidly expanding profits could be a good one-two punch for the stock. It’s not cheap, but if profits grow as fast as I expect, I think it’s fairly valued. I added to it a little bit in June.

Zscaler ZS reported on May 30, which I talked about in last month’s review. The weird price action was a big opportunity, and I loaded up in the $160’s. This month it was up double digits and I trimmed a decent amount. I still like the company and think the stock is at a fair price.

MercadoLibre MELI was down close to 10% earlier this month, so I added a smidge. Still probably keeping this one around 5%, though it seems like a stalwart.

Nu Holdings Weirdly, NU went the other direction and was up close to 10%. I trimmed, but ever so slightly. The revenue growth and other numbers are just amazing.

ELF I’ve been taking advantage of the volatility here adding when it drops and trimming when it rises. All told, it was up double digits again this month, and so my position is noticeably smaller than it was at the end of May. I do think this one is becoming very richly valued. The trailing PE is around 67, and while we can certainly hope this means the fwd PE is more like 40-45…I’m just not so sure we can pencil in super fast growth going forward. As has happened with Celsius, rough comps are ahead. I think the upcoming quarter will be very strong, but if the market keeps bidding ELF higher, I’ll keep trimming.

Paycom Paycom was a $30b+ company a few years ago, but has fallen roughly 75% and is now an $8b company. They’ve remained very profitable, but revenue isn’t growing like it used to be. My understanding is that their product “Beti” has resulted in customers NOT having to re-run payroll as often. Great for customers, but it cannibalizes some of Paycom’s revenue. It’s possible that Beti is just an excuse and that there are other problems. They had a weird event where they promoted a co-CEO just a few months ago and he has now reportedly left the company. I’m not betting big here for various reasons, but I do wonder if all the bad news has been priced in at this point.

Twilio, Docusign, and Zoom As I said above, these companies are beginning to right-size spending, leading to rapidly improving FCF. They’re all cash-rich and doing share buybacks. This is all OT for Saul’s board, so I’ll leave it there.

Supermicro I’ve been very wrong about this little company – they’ve been a key distributor of NVDA chips (with their own liquid-cooling value-ad) and I think this will continue for several more quarters before they peak – they seem to be a part of a lot of big installs. But it’s still hard to know when that inevitable peak will come.

Wrapping up

Happy July, everybody!

Bear

“Compound interest is the eighth wonder of the world. He who understands it, earns it … he who doesn’t … pays it.” - Attributed to Albert Einstein

Previous Month Summaries

Dec 2016 (contains links to all 2016 monthly posts): Bear's Portfolio at the end of 2016 - Saul’s Investing Discussions - Motley Fool Community

Dec 2017 (contains links to all 2017 monthly posts): Bear's Portfolio through Dec 2017 - Saul’s Investing Discussions - Motley Fool Community

Dec 2018 (contains links to all 2018 monthly posts): Bear's Portfolio through Dec 2018 - Saul’s Investing Discussions - Motley Fool Community

Dec 2019 (contains links to all 2019 monthly posts): Bear's Portfolio through Dec 2019 - Saul’s Investing Discussions - Motley Fool Community

Dec 2020 (contains links to all 2020 monthly posts): Bear's Portfolio through Dec 2020 - Saul’s Investing Discussions - Motley Fool Community

Dec 2021 (contains links to all 2021 monthly posts): Bear's Portfolio through 12/2021 - Saul’s Investing Discussions - Motley Fool Community

Dec 2022 (contains links to all 2022 monthly posts): Bear's Portfolio through 12/2022

Dec 2023 (contains links to all 2023 monthly posts): Bear's Portfolio through 12/2023

Jan 2024: Bear's Portfolio through 01/2024

Feb 2024: Bear's Portfolio through 02/2024

Mar 2024: Bear's Portfolio through 03/2024

Apr 2024: Bear's Portfolio through 04/2024

May 2024: Bear's Portfolio through 05/2024

71 Likes
3

I appreciate it, Rob. That is good news. What’s the source for those numbers?

Thanks,
Bear

1 Like
4

Follow the link I gave in my prior post.

Rob
He is no fool who gives what he cannot keep to gain what he cannot lose.

3 Likes
5

Bear, glad to see that you have finally “seen the light” regarding SMCI. There are several sources of market share in the server category, they tend to disagree, but not by very much, here’s one from SA (the author provides his source, but he does not provide a link so it may be difficult to dig deeper):
https://seekingalpha.com/article/4701348-super-micro-computer-385-percent-boost-in-capex-to-maintain-above-trend-growth?source=content_type%3Areact|section%3Asummary|section_asset%3Aall_analysis|first_level_url%3Asymbol|button%3ATitle|lock_status%3ANo|line%3A1

FWIW, I think SMCI will continue to take market share and grow revenue at a high rate for several quarters if not years ahead. Why would I say that? There’s more than one reason, I’ll provide a few.

Let’s start with the market that they are serving. It’s growing at an astonishing pace, so even without taking greater share SMCI will continue to grow rapidly if it can simply maintain it’s share, but that’s not enough in and of itself to provide the growth seen recently.

However, additionally, SMCI has been capacity constrained in the recent past. They have boosted capex in order to expand their production capacity in San Jose and Taiwan. And oh yeah, they have just about completed the construction of a new factory in Malaysia. In addition, they have a factory in the Netherlands (which they are not expanding). Add it all up and they are postured to deliver 5,000 direct liquid cooled (DLC) racks a month (Supermicro Expands Global Manufacturing Footprint Increasing Worldwide Rack Scale Manufacturing Capacity to 5,000 Fully Tested AI, HPC, and Liquid Cooling Rack Solutions Per Month).

You might think DLC from SMCI, big whoop, all the vendors provide DLC. Yes, that’s true, but it’s a little like saying anyone can buy a horse. Yeah, but there’s plow horses and thoroughbreds, there’s a difference.

What’s DLC supposed to provide? Duh, cooling . . . as rack energy consumption exceeds 1 kw/rack, air cooled servers will not survive, DLC becomes mandatory. Nvidia’s GB200 shipping H2CY24, demands in excess of 1.1KW/rack. DLC is required in order to prevent thermal runaway. But once minimal cooling is delivered, the secondary function of DLC is to provide efficiency. Worldwide energy demands of computers is straining the available generative capacity. Without the juice, the machines become idle ornaments. SMCI’s proprietary DLC exceeds the cooling capability provided by all the other vendors. That equates to more compute per watt, or lower TCO, or however you want to express the benefits.

Another factor is SMCI’s product catalog. They list over 200 stock servers. Here’s a link to SMCI’s product catalog (hoover over products). Wait a minute, that’s not 200 products - no it isn’t until you realize that each listing is a product line, not an individual product. Well OK, but so what. Why is that important? The server market is not homogenous. The needs of a telecom are different than the needs of a hyperscaler, which differ from a CSP and so on. Each vertical in the server market has different use cases which demand different server capabilities and configurations (look at the drop down for “Solutions” on the linked SMCI site).

In addition, SMCI engineers will sit down with customer engineers and customize the product in order to exactly match their specific requirements. To the best of my knowledge (maybe medium depth), no other vendor will do this. Pick best fit from a limited catalog and that’s what gets delivered. As an aside, having experience in the manufacture of commercial airplanes, I am very familiar with the planning and manufacturing difficulties of offering products designed for disparate customers. Think about it. It’s just a whole lot easier to build and deliver exactly the same thing. This is true irrespective if the variations are for individual products like airplanes or product batches like servers. Customization means different manufacturing plans, different inventories, different product configuration, etc. It’s a manufacturing headache at minimum. This is precisely the reason the other vendors don’t do it. SMCI’s “building block architecture” is not just a marketing buzz phrase. It’s a meaningful differentiation which facilitates their expansive catalog and makes customization cost effective.

And another thing, if you were paying attention to SMCI’s product catalog you might have noticed that they also provide storage systems and not listed separately they provide the necessary network components (ethernet and infiniband), system management services, integration software - in other words, everything needed to stand up various types of data centers.

I could go on with additional factors that provides somewhat of a moat for SMCI in what ostensibly looks like a commodity business.

It should be noted that their ability to take share from the market leaders came at a cost beyond their extensive product and service offerings. They intentionally allowed for margin compression in order to offer irresistible prices. Large enterprises tend to have significant vendor loyalty. It’s not easy to wean them away from established relationships. But the best time to do so is when there’s a paradigm shift in requirements. AI represents a paradigm shift. The net result of SMCI winning share now is that they will likely retain those customers (and win more) with every refresh.

Years ago, when I was still in IT refresh cycles shrank from 10 years or more for mainframes to 7 years down to about 5 years for servers. With Nvidia providing new, more powerful products on a 6 - 9 month basis, refresh cycles will inevitably shrink some more, probably a lot more. SMCI will be in a position to deliver products more rapidly than most, if not all their competitors. For one thing, Charles and Jensen (CEOs of SMCI and NVDA) are childhood friends. They have a great relationship. The Two companies HQ are literally down the street from one another. SMCI engineering has direct interactions with NVDA engineering. SMCI also has very solid relations with AMD and INTC.

From my POV, SMCI will provide significant alpha for an extended period. That being said, I don’t view it as a set and forget investment. But, I’ve noticed that you are far more adept than I at opportunistic buying and selling. SMCI has a high beta, so you should be able to benefit greatly by price swings. But macro economic activity has a big influence on them. Saturation of a sort will occur. Revenue growth will slow. One needs to pay close attention to this one.

44 Likes
6

Bear, this one seems a little different for you. I see two issues:

  1. BETI is not a new thing from what I can tell, so she is the fall guy maybe unfairly characterized as such.
  2. My reverse discounted cash flow model suggests PAYC needs to grow FCF by over 12% per year for ten years to justify its current price yet analysts expect no growth for at least another year.

I agree, it’s been beaten down maybe a little too much, but why not wait until growth picks up again?

Vince

14 Likes
7

@XMFRob Thanks for linking to that SeekingAlpha article. The market share numbers given in that article are computed only by comparing revenue. Dell had 90B in revenue last year, while Supermicro had 12B, and this is why Dell is roughly listed as 8-9x the bigger market share. The article didn’t share the original labeling of what they were graphing so it took a minute to track down.

Also of note against competitors it’s technically,

Dell - 51.5%
Hewlett Packard Inc - 26.3%
Hewlett Packard Enterprise (HPE) - 14.07%

HPE is the enterprise section of Hewlett Packard which does the AI servers, but they also have many parts of the legacy server businesses to enterprises. The legacy part of HPE is not growing, but the AI server spaces sold 900M last Q2 which the CEO said was,

driven by AI systems revenue more than doubling from our first quarter

I’m assuming that means that in the prior quarter the revenue was only 450M. So the conclusion is that HPE has less AI server revenue than Supermicro.

One contrast between HPE and Dell is that HPE is way ahead on liquid cooling of Dell from what I can tell. Reading through the last earnings transcripts, HPE is constantly talking about liquid cooling, and Dell barely ever mentions it.

I believe that Telsa buying 50% from Dell and 50% from Supermicro is an incredible win for Supermicro. First, what Supermicro has said is that their company is almost always the first choice for AI servers, however sometimes they are at capacity. In a case like that the customer may go for air cooled AI servers from Dell. Most companies would rather get the ball rolling on their projects rather than wait for faster/better machines to become available.

It’s interesting the incumbent Dell took away all the press from the deal signed due to the announcement. Musk had announced that Dell got the built out for 50% of their data center needs, and immediately other Twitter users asked who the other was and he cryptically replied “SMC”. I’m fine with Supermicro protecting the privacy of their customers as well, but it does make you wonder why Musk didn’t just announce Dell got 50% and Supermicro got 50%? Would be curious your take on this.

8 Likes
8

Musk replied SMC, not SCM. He subsequently clarified that on X, indicating it was SuperMicro Computers. As for Dell getting “all the press” (which is largely true), it doesn’t matter. If you’re in the biz… you know SMCI and you know SMCI grabbed half of the contract. As you say, it was a good win.

SMCI is constantly in a balancing act. They’re growing like mad… because they customize their products on request (1/2 of personnel are engineers per the CEO), they deliver fast and… they run a lean company and STILL quote low enough to have relatively low margins. The balancing act comes in since high growth demands high capital investment… but… their low margins don’t generate as much cash as they need. So… they periodically float some more shares to develop funds because they don’t want much debt. Despite that, share count growth is not a problem, IMO.

And yes, as you observed, the three market leaders are Dell, HP… and HPE. As you reviewed the market share info, I’m sure you noticed that essentially all of SMCI’s market share gains were at the expense of the “leaders”. Note the quotes… IMO, they’re not leaders, merely incumbents who still have most of the share… for now.

Disclosure: SMCI is somewhere around 14%-15%, I think, of my portfolio. It doesn’t automatically update on my spreadsheet because it’s all call options and I haven’t gone in to develop updates to all my options… too many positions when I’m lazy. Total YTD portfolio up about (insert insane number here)… probably will end 2024 up 600%-700%. Strong performance by: NVDA, SMCI, TSLA, APP, EOSE. Not so great with PGY… and I was hosed on CELH… can’t win 'em all. :smiley:

Rob
He is no fool who gives what he cannot keep to gain what he cannot lose.

25 Likes
9

That was a typo on my part for the letters.

And yes, as you observed, the three market leaders are Dell, HP… and HPE. As you reviewed the market share info, I’m sure you noticed that essentially all of SMCI’s market share gains were at the expense of the “leaders”.

Hewlett Packard Inc and Supermicro are not competing in the same space, that’s what I was trying to get at before that the data given in that Seeking Alpha is irrelevant for comparing the market share. Hewlett Packard Inc’s business, “Primarily focuses on personal computing and printing solutions”.

Supermicro does nothing with personal computing and printing solutions. HPE (separate company from HP) is the spin off which is publicly traded, and only a small portion of their revenue is AI servers. HPE has 7.2B revenue, 900M of which is AI server, so only 13% of HPE’s revenue comes from AI servers. Dell has reported 1.7B of revenue from AI servers on the last quarter, and total revenue of 20.9B so Dell’s AI server business is 8.1% of revenue.

Supermicro has over half of it’s revenue coming from AI servers, so that’s roughly 1.9B, and that means that Supermicro has about double the market share for AI servers that HPE does.

The break down by revenue for AI servers for total revenue is,

HPE - 900M
Dell - 1.7B
Supermicro - 1.9B

So for the segment we care about most, which is AI servers, Supermicro has a slight lead in sales over Dell which are both roughly double the size of HPE.

23 Likes
10

You’re right. That article I linked was wrong and I “assumed” the numbers were right. I’m deleting the post.

Rob
He is no fool who gives what he cannot keep to gain what he cannot lose.

8 Likes
11

Wow! That’s very illuminating. Thanks for breaking it down. I didn’t take the time and trouble to actually look under the hood on those “market share” numbers from SA. I just accepted them at face value which turned out to be a highly inaccurate assessment of where the business is going. Turns out that SMCI is doing better - much better - than I thought.

9 Likes
12

@brittlerock Appreciate the feedback! I had also not broken it down just by AI server category before and it’s providing a lot of interesting insights.

If we consider that most of the AI server market is HPE + Dell + Supermicro, and a few other smaller competitors like Celestica (CLS), we may be looking at somewhere between total sales of ~5-7B for AI servers at current quarterly run rates.

However, this space is now absolutely exploding in growth rates even for the legacy players. Dell has said their revenue of 1.7B for AI servers is up 100% qoq, and HP also said AI server revenue was up exactly 100% qoq! I do not think Supermicro breaks out the growth rate for AI server alone, but I’d be curious to know what Supermicro’s qoq number is for AI server alone. The best I could find was Supermicro is doing “triple digit growth” in the AI server segment but that’s for yearly growth.

A couple other interesting stats, AI servers accounted for only 23% of total server sales in 2023. I see this cycle being very early on, especially if we consider in a given quarter there’s just ~5-7B of quarterly AI server sales from the major players in the space. Look how big the legacy businesses are of Dell and HPE where they are still selling tons of computers, printers, and other devices.

HPE, Dell, and Supermicro are all in a hugely growing market of AI servers. The modern AI server has really only been around two years or so, so the field is quite greenfield. I think Supermicro is the first choice for most companies to go to if they want AI servers, but demand is so great, there’s going to be overflow into companies like HPE, Dell, and Celestica. Of those companies mentioned, Supermicro is the only one of the group where AI severs make up over half of sales, or the only company that is really dedicated to the AI server market. While Dell has 8% of sales being AI server, they still have a huge legacy server business to manage which probably slows them on innovations like liquid cooling.

I read in one article that HPE’s innovation over Supermicro is that their racks could support air cooled and liquid cooled simultaneously while Supermicro could only do air cooled, or liquid cooled. This sounds like a niche usecase that companies may have and I’m not really buying this as a major advantage. If the data center is eventually going to switch over to liquid cooling, then why would you want to equip with air cooling as well? It seems like you’d just be building inefficiency in with the air cooled portion.

27 Likes
13

Wpr, just wanted to take a moment to give kudos. Your contributions to the board this year in terms of analysis and questions have been thoughtful and all around excellent. Even tho I don’t agree with all of your reasoning and positions (hims for example), I really appreciate what you are bringing to the discussions around here.

36 Likes
14

Thanks again - there’s more depth to this discussion than I originally thought. The whole issue around DLC is really a discussion of efficiency or compute per watt. Electrical power is the fuel of AI. From what I’ve read, availability of power is a major consideration when it comes to standing up a new data center (or as Jensen likes to call it, an AI factory).

While DLC provides the lowest power consumption per a unit of compute, it’s not an easy decision when it comes to converting an existing facility that relies on air cooling to one that employs DLC. A fair amount of ancillary plumbing is required. That has a lot of ramifications with respect to the configuration of physical floor space. Retrofitting is a much more complex problem than deciding which vendor will provide the racks.

Nevertheless, based on what I’ve read, SMCI delivers the most efficient DLC of available racks. SMCI provides the lowest TCO as compared to Dell and HPE (that’s from one of several SA analysis articles I’ve read, I don’t recall which one specifically). I assume it’s likely true for all AI equipment vendors. In other words, DLC is not homogenous. SMCI holds patents that protects their implementation.

I was in IT at Boeing for 30 years. I retired in 2010, so my knowledge is somewhat stale. But based on my past experience, large enterprises tended to have significant vendor loyalty. It was not easy for a new comer to dislodge an established supplier. But the most likely time for an enterprise to seriously consider alternatives was when there was a technical upheaval. I was one of the managers involved when Boeing transitioned from Big Blue mainframes to distributed UNIX servers. When it came to servers, Boeing didn’t go with AIX boxes (IBM). HP (there was only one company at that time) was the winner. When it came to middleware, Oracle beat out DB2 and Sequel Server. What I’m getting at is that the business that SMCI wins now sets them up as the vendor of choice for the refresh cycles to come. It makes a lot of sense for them to allow margin compression in order to gain share. Assuming enterprise MO has not changed significantly, every customer they lure away from the market leaders now is likely to remain a customer well into the future.

22 Likes
15

The news from Google this morning about a near 50% increase in their carbon emissions due to the power consumption they need for their AI data centres once again underlines the need for DLC in the AI fabs. This was why Charles Liang of SMCI has been so vocal for such a long time on his desire to cut emissions, invest in more green technologies, and help companies save money too. In his recent address at Computex in Taiwan he made significant emphasis on this both in his speech and in the video he presented.

He also said that about 15% of racks shipped this year will include DLC, and that will increase to 30% next year. Thats a massive increase on the 1% of DLC previously supplied. Liang said lead time on orders has declined to two to four weeks – that’s down from as much as a year.
The need for DLC only increases as AI data centres continue to proliferate.

Long SMCI

Jonathan

27 Likes
16

I suggest ya’all listen to Beth Kindig’s latest take on SMCI. She thinks their supply constraints are going to open the door for Dell to take over while SMCI tries to borrow and issue new shares to build-out and meet demand. Beth’s pretty insightful on this stuff and it’s something to take into consideration. Dell has a big advantage in cash.

10 Likes
17

I often struggle when listening to Kindig because as a nerd myself it’s apparent to me that she often doesn’t understand the tech she claims to understand.

For instance, in terms of Nvidia’s AI chips (start about 4:30 in that video): “floating point precision, moving from 16 point to 8 point, which increases accuracy while also increasing speed.”

She’s quite mixed up here. First, no-one talks about “points” - those numbers are bits. There is no 8-bit floating point precision. There is a 16-bit floating point (FP16) as wll as an 8-bit integer (INT8). There are also FP32s and INT 16s and INT32s. So, Kindig might be talking about using INT8 to gain performance advantages over FP16, at the price of losing precision (read Nvidia’s developer doc on it here), but that’s a well-known concept usable on any hardware. And there’s no world in which 8-bit data types have more accuracy than 16-bit data types. Maybe Nvidia has sped up FP16 to compete performance-wise with INT8, and that’s what she’s referring to?

Then Kindig (5:40) attributes Nvidia’s success in training “entirely due to CUDA…AMD is equal to Nvidia in hardware.” That’s an incorrect assessment of the situation on both counts, in my view, but I won’t go into details here.

The inferencing market will indeed expand into edge/end devices, as Kindig says. Today’s inference is 40% of Nvidia’s data center business, which might be a surprise to her given her statements. She may be right that AMD can capture a significant portion of that edge/end device inferencing market, but there are also strong companies like Qualcomm that are already pushing into that market. Competition will be fierce.

It’s ironic that Kindig correctly says (13:45) : “Inference runs best close to the data” but then continues “we don’t have powerful enough devices for inferencing.” The first issue is where the data will be - computational speed will be next. If the data is on your smartphone, you run the danger of using an out of date model. Many functions we do on our computers and mobile devices already rely on server-provided data and computation. From asking Siri for some data fact to your car planning a route based on current traffic, the data and almost all of the compute are server-based. And today, the LLMs are not only trained in big data centers, the inference is also done in those big data centers. Some of that will move to the edge/end, but only when privacy is paramount or performance improves, and even then the data requirements for storage have to be considered. Adobe’s AI image generation, for instance, requires that your image be uploaded to Adobe’s servers for the work.

As for Dell vs SMCI, I don’t have any insights. Kindig may be right that the upcoming Blackwell surge from Nvidia will benefit Dell more than SMCI, but I think it remains to be seen who will produce the most popular Blackwell-based servers.

43 Likes
18

Beth wasn’t talking about best but about supply constraints. I know SMCI is bumping up against their upper limits on the number of servers they can produce but are doing everything possible to increase that. The problem is will they be able to increase it fast enough? I am thinking they will but it is a concern.

Andy

10 Likes
19

@FallingWallenda I am not an expert in this subject but…SMCI is increasing their supply this year and beyond, they have the relationship with NVDA, they have the AI rack reputation and engineering prowess, liquid cooling advantage and announced a massive Blackwell shipment.

Sorry, the “maybe in the future something bad could happen” argument is what kept so many on this board from getting into this stock before the meteoric rise. It’s still a cheap stock compared to so many on this board, there is no reason to think it should stop here. For me, I will follow the numbers and the current tailwinds/company actions vs. some sort of future hypothetical.

21 Likes
20

She’s saying it’s not in the future. She says they’re at full capacity now.

1 Like
21

Where these two ideas might collide is is the dreaded “next year’s guide” SMCI will issue with its Q4 report. The company’s current top-end FY24 revenue guide is $15.1B. Taking a quick look, the analyst Q4 consensus of $8.1B would mean $17.7B in FY24 with an analyst consensus of $23.86B for FY25 (I used Yahoo Finance, so someone can feel free to present other numbers if you have them).

If for whatever reason management can’t meet either of those figures due to any capacity concerns or hedges, I doubt the market will be very forgiving. That seems to be a risk with SMCI that might not exist with some of its competition.

14 Likes