I too have a large SMCI position (and Nvidia). I agree with you in part. Let me explain.
First, the issue of customer demand. As @wpr101 noted, Meta is presently SMCI’s largest customer, but far from their only customer. They have a few (unstated how many) 10% customers as well, which have a tendency to change quarter to quarter. Sometimes, a current 10% customer is also a new customer. Other times it’s a new order from an existing customer. The point is, SMCI does not want for customers. They have more demand than they can get parts in order to build product to fulfill their demand. As far as business problems go, this one is not too bad. Also, they stated that the supply chain problems were getting better. In other words, they do not anticipate that the expansion in production capacity will sit idle for want of parts.
However, it is also true that they have intentionally reduced margins to win new customers. Of course, this benefits existing customers as well. But the proof is in the pudding so to speak. The increased volume has more than offset the reduction in ASP so it’s hard to argue with this strategy.
I noted in a separate thread when I worked for Boeing they changed their purchasing strategy from one of pitting suppliers against one another for every order, mostly to gain the best price, to one of forming long term partnerships with their suppliers. In other words, once a vendor was doing business with Boeing the had a very strong position with respect to repeat orders. It wasn’t superglue, the company would occasionally change vendors for a given product (or product group), but their had to be a very compelling reason. This practice was adopted from the Japanese way of doing business. Many large companies at that time employed this same mode of operation. I retired more than a decade ago. Is it still a common practice? I have no idea. Does it apply to products like servers? I have no idea.
But even if it doesn’t (it would be good to know what level of repeat business SMCI captures), this is not their only competitive advantage. Their are at least two other factors that sets SMCI apart from other vendors.
One is time-to-market. Even if a company has their order placed on back order by SMCI, they will still receive product sooner than if they go to another vendor unless they need a commodity box. It was not stated, but it is my impression that SMCI does not compete for the bottom of the market. There are a lot of very low cost Asian competitors in this market. I don’t think they want to dedicate their production capacity to capture these orders.
The other thing that SMCI offers which very much sets them apart is a willingness to customize their machines in order to meet the specifications of each customer’s unique use cases. Even though they have a very large catalog. I.e., a telecon building 5G networks has quite different requirements than a social media company - I don’t know how many segments their are, but SMCI has off the shelf products that address all of the significant ones. And beyond that, they will gladly have their engineers sit down with customer engineers to insure the customer will receive product that exactly addresses their use case requirements. More than half of SMCI’s US staff is comprised of engineers.
As for the comment that META is an example of how companies view SMCI as meeting their needs better than they can do it themselves - I see that as part true, but their are significant exceptions. For example, it is my understanding (I could be wrong) that Alphabet is designing their own chip. It is neither an CPU or a GPU, I don’t know what the acro is. Rather, it is an entirely different architecture. They will (or already are?) going to build their own machines. And there’s Tesla’s Dojo, also built on proprietary chips - at the same time, Musk has asserted he will buy all the product that Nvidia will sell to him. But so far as I know, Tesla is not buying SMCI products. Apple has a unit dedicated to ACDC (AI chips for data centers). I have no idea if this will be for internal use only or if they intend to separately market it, or even machines built around it. Apple likes hardware, so it will probably be machines if anything. So there is some validity to your observation that there is nothing so unique about servers that gives SMCI has a lock on the market. Time-to-market and customization do not make a very deep moat.
Which brings me to your other point about longevity. I do not plan on holding my SMCI position for the remainder of my life (I’m old, there might not be that much remaining). But, make hay while the sun shines. When I look back at my investments, there just aren’t many, or any companies that I’ve held for a very long time. When it come to growth companies, they inevitably stop growing at a spectacular rate. If nothing else, the law of large numbers finally overwhelms the ability to grow 30% - 50% every quarter, quarter after quarter indefinitely. It can’t be done. Even Apple, Microsoft - the other very large tech companies may have spurts when a new technology comes about that requires massive amounts of capital (like AI).
But we (including you) will probably not see another technology like this for the rest of our lives. The way I see it, if you are not exploiting the opportunities AI is providing at present, you are very much missing out. There won’t be a repeat.
So that’s why I have large positions in Nvidia and Supermicro, these are the picks and shovels companies of AI. To my way of thinking, at present there’s more opportunities with the companies feeding the frenzy than the companies that will win out. But if you want those long term winners, they aren’t so hard to identify: MSFT, GOOG, APPL, META, a few others. These companies will continue to grow and provide reasonable returns for many years to come.