원문 보기
Hello everyone, this is Kevin. It all begins. Big business is starting to pay more and more attention to software. A few days ago, or maybe yesterday, I mentioned how the All-In podcast started discussing the idea that open-weight models would really benefit applied levels of artificial intelligence.
Three components of AI profitability, right ? These are computing power, large language models (LLMs), and software or application layer. OK. I also got comments when I referenced them from people saying, "Oh, you lost me as soon as you mentioned All-In." And some people said, "Kevin, why did n't you tell us sooner?" And I'm like, "I told you so." We talked about this three months ago, that I think software is the main winner in AI, and LLM and then computing power will become a commodity .
I'm not thrilled with any of them. I think software is the real winner here . The problem now is that more and more people are starting to adhere to the same thesis. And I like to invest in something where the thesis is not yet overcrowded. Overcrowding is still a long way off. However, to be fair, look: IGV (a software ETF) has seen net capital inflows over the past five days, which means that there may be more going on there than just a short squeeze.
I do not invest in IGV. I just use it as a kind of indicator of the software's status. Nobody sponsors me. People always get so cynical, but honestly , they're right. You know, it seems to me that so many people are trying to scam someone. But in any case, let's talk frankly about what I just heard from the business community .
When I say "heard," I mean Bloomberg Intelligence—it's a kind of insider analytical publication that only reaches big players and hedge funds. They published a paper on how open-weight models are actually changing the rules of the game in artificial intelligence. And this is very important. Okay. By the way, I'm in this little lounge right now.
I want to say that yesterday I was sitting here with Lauren, and our team won. That's all I can say. I may have yelled " stop counting" a few times when we were doing a quiz or something. It was a lot of fun. This ship is pretty cool. Cruises should be just that , but this is not an advertisement. I really wish Disney stock hadn't been in the red for the last 11 years.
I don't own Disney stock, but it's strange to look at Disney over the past 11 years: they've fallen. If you look at them over the last 5 years , they have actually fallen a lot. A lot of this, frankly, has to do with the " streaming wars" and the money that has been spent trying to maintain leadership in streaming.
Even though innovations in the film industry do not seem to have yielded significant results. Although I must say that in the cruise and entertainment sector, profitability is excellent. This is logical, as their pricing power is enormous. Going back to the Bloomberg Intelligence article, it's really interesting because they claim that the "moat" is changing.
At first, when artificial intelligence came along , people said, "Okay, that's it." The entire " moat" will be in large language models (LLM). Anthropic and OpenAI will actually steal your data and create copies of your CRM system, your dashboards, or anything you want. You could just ask Claude to write the code for you.
Bloomberg Intelligence now says that open weights are changing things in two ways. First, they reduce infrastructure costs for software companies, allowing them to provide better service to customers. This is literally what I've been saying for years: that this is exactly what's going to happen to software.
But in any case, now serious analysts are talking about it . Although I want to praise myself for the accuracy of my predictions, I think: " No, no, no, don't talk about it yet." I still want to buy more . I'm not ready for you to realize that yet. But whatever , we can talk about it. You know, we grew up together with you.
We are like one family here. So, okay. Infrastructure costs are, of course, reduced by open scales, because, as they say, you can spend 50 cents on a certain number of tokens instead of $5. It is obvious that the end consumer, who is charged the same bill due to the pricing power of the company he uses, does not benefit from this.
The application company, the software developer, wins, as it now has lower costs for providing this service. By the way, Service Now, I do not own any shares of Service Now. Let me make this clear. I chose other software companies. You can, of course, see this within the MEK membership. You can join me there.
We have, you know what we call it? Vacation Red 2 or something like that, or Cruise Red, I think that's what we called the coupon. Check it out on me.com. Thank you to everyone who joined yesterday. You can see exactly where we are buying and I will probably send another purchase notification today or tomorrow as I am buying more.
But anyway, what's really interesting is Service Now, if you read their latest earnings reports, you'll see that the only thing they complain about is their infrastructure costs, which are constantly increasing. And guess what, the levels do something like, well, something like that? They reduce or help solve this very problem.
Pretty cool, right? Let's take the stage here, since we're so pedantic. I'll just talk while we're on stage. That's it. Oh yes, now I have lighting. Oh, that's actually pretty cool. But, oh my god, it's too early for that. Well, no, everything's fine. OK. So Service Now, one of the reasons they've been moving so well in recent days is that the very thing that's been impacting their margins is infrastructure costs, which are actually being addressed through open-source models, I have to say.
There is a big difference between these two concepts. Please find and understand the difference between open source and open source. This is very important. Because one still maintains a competitive advantage for companies like Moonshot, and the other does not. OK. So, this is the first one . But the second thing it gives is maintaining a competitive advantage.
analysts are saying. Retention of advantage—this is essentially when, if we already have this layer of applications and we have our customers in place, we get this recurring revenue. We have clients who regularly renew their contracts because they find, " Hey, even though we can quickly throw together a dashboard or something like that using cloud code..." It's not nearly as good as the integrated software, the dashboards that we had, or the applications and systems that we had set up before.
Then we actually remain more loyal customers of software companies, whether it's Palantir, Salesforce, Service Now or others. So, you do two things. You reduce these operating expenses or cost of goods sold, depending on what company it is and where they decide to attribute it. you're reducing costs for the software company, but at the same time increasing customer retention for those companies, because now individual companies can say, "Hey, we want to tweak our weights a certain way, and no OpenAI or Anthropic is going to steal that fine-tuning.
That's what makes the money in the post-training phase. Okay, it's a little bit complicated. I'll simplify it a little bit. There's pre- training, where you basically, well, again, oversimplifying. We take every book that exists and we try to get our LLMs to create neural connections between them to recognize patterns.
So when you send a query or a prompt to a chatbot, it can recognize patterns token by token. Okay, you're talking about that. You probably need information about that. Here you go . Okay, that's all pattern recognition. There's nothing intellectual about it. And God, some people get so angry when I say that.
They're like, "What do you mean? We are so close to AGI. "Uh, no. It's actually very advanced pattern matching . But anyway, so now we have this pre- training. Okay, we've got that. We're there. Now what really matters is post-training. Post-training is when we take these encyclopedias and we tweak them, saying, "Hey, we want to...
I'm making this up as an analogy to help people think about it this way." Hey, thanks for helping us with all this data. Thanks for helping us determine that this is indeed a kitchen. Let's say we're now going to set it up and say that we think repairs in this zip code will cost so-and-so. "And so we're going to tune it with our algorithms, right?
The same goes for any company , whether it's a real estate company, like what we're doing with House Hack, which, you know, we think of ourselves as basically an AI company disguised as a real estate company, which is pretty cool because the people who invested , invested on the valuation of the real estate company.
We might do another fundraiser before the end of the year, but honestly, I want to be done with the fundraiser, mostly because the business is self-sustaining. More on that in another video. But another analogy to move away from real estate, for example, and this is just more general, could be like, hey, we're a biological lab.
uh, we want the LLM to steer us in a certain direction when we're identifying protein folding or something, but we want to tune aggressiveness on this type of folding or this or that. Use this strategy or this approach. So you're essentially steering your systems to do something a certain way. And the combination of not just one of those solutions, but maybe 200 of those solutions and customizations—that's what makes your company special and gives it a price advantage.
If you can now deliver the end product, an LLM with 200 customizations in a unique and user- friendly dashboard for the customer, then they don't have to spend the effort to replicate all of that. Then you're essentially expanding your "moat" of defense. As Bloomberg Intelligence just said, your defense barrier is increasing, and the software companies' costs for that are decreasing, which means net customer retention is increasing at the same time as costs are decreasing.
Just in the last week, we've gone from 2 years of saying that software companies are this to, "Oh my God, software companies are going to be great." And that's usually what happens in the market. There's this madness where the market gets overly bearish in one sector and then it turns around in another direction, which creates big gains and opportunities.
So that's my opinion. And of course, we don't know. probably, but it's time for me to go. So thank you for being with us, and I'll see you in the next episode. Goodbye and good luck to you all.
Comments 0