저자: MeetKevin · 외국인
요약 보기
AI 수익화는 컴퓨트·소프트웨어 계층이 유리하고 LLM 토큰 계층은 상품화로 수익성이 낮다고 주장. 팔란티어는 고객 집중·매출 84%↑·이익 99%↑로 가격결정력이 강하며, 향후 주가 4배 가능성을 PEG 기반으로 제시. 소프트웨어주(팔란티어·인튜이트·세일즈포스)의 할인 매수를 언급하고 IGV 바구니를 제안.
다듬은 글
This video will accurately explain where you can make money in the future in the field of artificial intelligence. Today is July 3, 2026. You should know that there are three ways to make money in artificial intelligence, two of which are more likely to end in losses. In this video, you'll learn what it means , and we're going to use Palantir to explain everything.
So, whether you like Palandir or not is not important for this video. Watch this video till the end because you will learn a lot of things . It's time to talk about stocks that can yield four times the return , and we'll look at this based on what was said in this Palantir interview . So, yes, most of this video will be about Palandir .
Since many people misunderstood look at it in detail . Let's watch this video, who doesn't like a Sara video? Come on, let's start with Alex Karp. After that, we will talk about Palandir's financial situation. First, let's look at how other companies compete, based on what has been said to understand what is happening in artificial intelligence, and compare it to Palantir's financial situation.
At the very least, let's explore where we can see a four-fold profit, because we all want to make more profit . So, let's get straight to the point . I'm Meet Kevin, let's get started. Mr. The first thing you need to know about Karp is that he is a great salesman . This is not a criticism against Palantir.
Actually, I have more faith in Palantir. However, when he warns about how bad some things are , there is a kind of bias in it ; The more aggressively he speaks, the more people want to use Palandir , because he is selling the solution while pointing out the problem. So, keep this in your mind . However, he has a good starting point for why there is disillusionment with large language models or leading labs like OpenAI, Anthropic, and Google's Gemini; It's well known that Meta has admitted that its AI is lagging behind the competition.
He is dissatisfied with the LLM leading laboratories . He says they steal your weights, data, and security. There is some truth to what he says . Did you know that 'The Information' magazine published a story about a design company called Figma ? Well, both Figma and another design company, Canva, have partnered with Anthropic.
They thought, "Oh, this is great. We're going to be Anthropic's design engines , work with them, and get more business ." Therefore, it was assumed that Figma and Canva would win by alliance with Anthropic. Anthropic's head of production was a member of Figma's board of directors. So, this seemed like a great partnership .
But suddenly, the production head, Fikma, resigned from the board of directors. Then, shortly before the launch, Figma ended its partnership with Anthropic . Immediately, the Anthropic company announced, "Here is ' Claude Design'." It basically performs the same service that Figma and Canva sell. Is that so!
This confirms what Karp says . LLM companies are keen to acquire the same companies they say they want to partner with. I'm not saying that anthropic is completely bad, but that's what Karp is saying . " Look, these leading labs will steal your data," says Karp. "If they steal your data, they will learn the secrets to your business's success and start a business that will compete with yours ." This may be, in a way, the reason why that stock I'm about to mention is falling.
What if, for example, a company like Intuit said, "We want to integrate your LLM into all of our QuickBooks products?" Some stock investors will start saying, "Get out of Intuit stock ." Because those LLM companies will copy the QuickBooks software exactly . I don't think it's that simple , but that's what Palantir Corp is insisting on .
That is, LLMs are dangerous. They will steal your data, your secrets, and your security. Instead of the value they provide, they charge you in tokens . Okay. Part of this is sales strategy, part is truth. Just as it is easy to charge for electricity, it is easy to charge for tokens. It is difficult to ask for a share of the profits because it is complicated to calculate whether those profits came from those tokens .
It is very difficult work. But that's exactly what Palantir does. Palantir companies base their business deals on how much profit they generate for their client companies . At least that's their sales description. You might be surprised to learn that Palantir doesn't have many customers. You think Palantir is the company that plans to generate more than $7 billion in software revenue next year .
That's what they're aiming for . You would think that this would have millions of customers. But they do n't. Because they are selecting the highest-valued of their Fortune 500 customers and trying to sell them more and more . So, the sales team says, " We don't pick up the phone for all business ." "We only pick up the phone for very large businesses, like the U.S.
Air Force, the Department of Defense, or one of the largest hospital chains in the country." This is where it's useful to understand what Palantir does . I'm not going to go into this in depth because it's a bit confusing , but you should see where LLMs fit in, because it also applies to other stock-related things we're talking about.
Look at this. This is an approximate picture, and I will explain it very simply . Palantir talks about a concept called ' ontology' . This already sounds very complicated. Just think of it as Windows XP . It's an operating system, right? Windows Vista. Oh, no. No Windows Vista. Okay, okay. Windows XP.
We will stay on Windows XP. So, think of the ontology as essentially a platform . Palantir takes your data and combines it with the actions and functions you want to perform, algorithms, etc., that tie it all together. So think of it this way . That's an easy way for me to think about it , right? Let's say we need to produce Model 3 Tesla cars , and we need a widget in the seat of each Tesla .
That's a butt heater, right? Every car should have heated seats , right? So, I have the materials ; Car and seat heater. These are connected to distributors . I need a certain number of items based on the orders I receive . Orders are a number we have. Now we know that cars and seat heaters are what we need to produce.
We can connect them together . And we can use algorithms or weights , like how many of these seat heaters we order are failing ? How many are we going to get from each distributor , right? All of these are important. All of this comes together through Palantir. When you don't use all of these on the same platform , it actually becomes very complicated.
Imagine having all of this in useless spreadsheets . For example , okay, let's say I have a spreadsheet . This is the number of orders we have, and I need to somehow connect this to Butt Heaters and various distributors . I have to go into that spreadsheet and connect everything the old fashioned way , which is really bad.
Palantir basically integrates all of this for you , which is why it is so powerful . Whether it's a hospital counting beds and patients , a shipping company , or a war department deciding whether to kill someone . They integrate all these details . This is the reason why we missed 9/11, besides the fact that 10 people in Chad say it was a domestic conspiracy .
Even though we have information about the 9/11 events , we may have missed it . We knew that suspicious pilots were being trained . We knew it was all connected , but no one had put the pieces of the puzzle together. That's where Palantir helps. If Palantir helps, where do LLMs come in the Palantir model?
Hmm. Well, this is the interesting thing, because you can have all this information , which seemed a bit complicated, right? But, what do you have here ? LLM comes in as something called hot-swappable. Hot- swappable is actually very important because it means Palantir doesn't have to rely on Claude, Anthropic, or ChatGPT.
They are subservient to computing power , aren't they? They have to use Microsoft or Azure or AWS to host their operating system . And they have to pay tokens for these LLMs , but they can switch between those different LLMs. LLMs are essentially a tool to help those using Palantir. For example , I asked, "Dude, do we have enough butt heaters?" If you ask me.
Instead of going through the entire operating system layer, I just go straight in and say, " Hey, are there enough butt heaters this month ?" I ask. Then LLM will say, "Yes, everything seems to be fine." Or, " You have a problem." You can use this predefined function to transfer some of your orders to another distributor .
All those algorithms are already determined , but LLM is a separate external network. Now, why am I mentioning this ? Because, says Karp, LLMs are a scam . We should see them as a commodity . This corresponds to a two-tiered system of earning money . This is where it starts to become a little more important.
I know it takes a while to go into the details, but I think this is where the money is really made. This is important because it would take hours to create a video thinking about how to convey so much information, right? Okay. There are three ways to make money with AI . Do you understand? One is the compute layer.
Do you understand? This is what is called the hardware layer in the race , right? GPUs, server racks, the components in those racks, fiber optics, whatever . Compute layer. Then there's the token layer, which is your LLM world, like OpenAI, Anthropic, etc. , right? After that there is the software layer.
The software layer has been depreciating in value recently . The software market, that is, the stock market, is selling shares of software companies . These could be CRM or even companies like Palantir . Palantir has suffered a major decline from its peak price . To see if this is fair , we are now going to examine the financial statements.
But this indicates that something is wrong, right? Like, "Oh, it looks like all the software is going to be replaced ." But look, Karp and the CEO of Salesforce think the software layer is the most profitable layer , and the stock market is wrong about that . That's where things get interesting . That's why I took so much trouble to explain the operating system and the LLM system .
Therefore, LLM is just a component or tool that you can use with the operating system. The operating system contains all your files and data, you don't need to give everything to LLM. That is the main feature of this. Obviously, SpaceX and Meta, as well as Microsoft, AWS, and Google, make money in the compute tier , but competing at that level is very costly.
Therefore, the software industry is definitely a more profitable industry than the hardware and computing industry. In addition, this area is mostly at a commodity level. That's why I think Anthropic and OpenAI want to do an IPO soon, so they can get the money and buy a jet . Let them do whatever they want with it.
However , Palantir's Corp tells us that instead of charging for profits, these companies charge for tokens . I repeat , that is Palandir's sales tactic. In his opinion, the two places where you really make money are the compute and application layers. That's what I just explained. Not an LLM tier. He says people are angry because they have to pay for tokens that create no value, and instead create a wealth tax by sucking the value out of a company, i.e.
stealing data. Okay. So, we've already talked about data stealing, alpha, and the filtering part. We talked a little bit about how Palantir is trying to solve that. We also talked about where the money is made . So, I think this 'no value' concept is a bit extreme. He strongly criticizes this layer .
But, as a commodity, it will create less value over time. You will make more money in the software compute tier . I think that in the future, computing will become a commodity. Therefore, software will only be important in the future, where large- scale lock- in effects will create the greatest value for companies.
That's where I think you'll make big money in the stock market. Again, we're going to talk about Palantir's financials in a moment . But, he says, they create a huge wealth tax. What does he mean by this? In his opinion, what he's basically saying is, "If you go out and annoy a lot of people and burn a lot of money building data centers , if you create a social aversion to AI and don't create value at the same time, people will resist data centers , and you're actually crippling an environment that should be delivering value and profit .
Okay, interesting . Now, he points to companies like Corp Anthropic , Dario is a good guy, but look at what they're doing. They're coming up with Cloud Security, Cloud Legal, Cloud Code, Cloud Science, Cloud Design, right? They're just copying what other companies are doing . Becky Quick notices this and says, " You sound angry ." This seems a bit of implicit anger.
To which he says, " No, I'm expressing the voice of American businesses ." Short-term investors do n't know what we're talking about. LLMs are probabilistic, where being 51% correct is enough, but that's not enough when you're trying to attack someone with a defense department. " Well, there's a lot to understand here.
Okay, let's go over this again in detail . LLMs are probabilistic. He's basically saying, " Look, it's enough if they're mostly right ." This is where hallucinations come in. That creates AI chaos . If you listen to Palantir's earnings call, you can hear them constantly complaining about how bad the AI mess is.
This is a software layer that actually takes an LLM and puts constraints on it to prevent it from being chaotic and instead make it deterministic . Algorithms are deterministic . Generative AI is not like that. But if you use them with a software layer , you can get a predictable result. So, it's really like we're creating our own proprietary AI.
This is not an advertisement. We are building our own proprietary real estate AI to instantly find real estate deals across the country . Someone said, " Can an agent do this?" They ask . Of course. But how long does it take an agent to review 200,000 listings across the United States? It should be a combination of LLM , but mostly algorithmic .
So, you have some AI and some vision AI, but there is a lot of algorithm . Because algorithms are what we know to be stable. They are determinable , we enter our values , 1 + 1 is always 2 . There's no way to say it's mostly two , right? That's what algorithms are. I mention this because we work in the same field .
Of course, on a minuscule scale compared to companies like Palantir, right? We are just a small startup . But I mention this because algorithms are important, and balancing algorithms and artificial intelligence , vision, generative, and language models, is where the value comes from. Palantir does that.
They combine LLMs with algorithms and use them in the right places . Okay, great. And of course Big Data . Remember, in the past, everyone called AI big data . Analyzing big data, right? Okay, now where do we go from here? Okay. We talked about how ontology works and how, from Palantir's perspective, it is an operating system for bringing data to life .
How much money do they actually make ? Is this company worth investing in? So, the first thing I want to do is look at their cash flow statement . Their cash flow statement, let's just say it's already pretty impressive . The company last repurchased shares in 2023 . They are not currently buying back shares , but I think they will start doing so soon in this company, which will be very beneficial for investors .
Please note that this is not personal financial advice . Okay. Their net cash from operating activities has tripled , which they are investing in bonds . Specifically, investing in Treasury bonds that mature within the next 6 to 12 months is Buffett-style. It's great that they keep their marketable securities very short-term and low-risk .
Another thing to note in their cash flow statement is that their net income in the last 3 months has increased several times over the last year . Yes, net income has increased almost 4 times , which is very impressive . And their stock-based compensation only increased by 33% . Therefore, their stock compensation is not keeping pace with the increase in income.
This is a good sign that a share buyback will be coming soon . Now let 's look at the revenue. This company combines everything we know so far , right? The weights, operating system, algorithms , data, and use of others' computing power will all be standardized . The only part that is not standardized is the software layer, and that is Palantir.
What do we see here? Revenue has increased 84% year-on-year. Total profit increased by 99% . This means their pricing power has increased . They have made more profits with less cost or their costs have decreased. If they can do this, their pricing power increases . Generally, if revenue and gross profit increase at the same rate, it means their pricing power is stable.
Or if revenue doubles and expenses triple , that's a different story. If so, it means they are unable to pass that cost on to customers and their pricing power has diminished. Their pricing power decreased. Nobody likes it . No one. Their sales and marketing expenses, including stock compensation, have only increased by a third .
Because, as surprising as this may sound , they only have approximately a thousand customers . They might be a little more . I think I have it noted down somewhere. I will find it soon . But, with just a thousand customers , you can have a large sales team and consistently sell to those customers at a higher price , which is very valuable .
Here it is on the earnings call. The number of customers increased by 31% to 1,007 . That's all. They don't really need many small customers . They only like big fish . This is one reason why Intuit 's shares are falling . But people don't understand this . Intuit has a lot of customers . Let's say millions of customers .
Now, to simplify the calculations, let's assume there are one million customers . Their biggest customers spend the most money . Their poorer, lower- value customers spend less money . So, they leave the site . They lose the poor, retain the rich , and sell to the rich in large quantities . Including women.
Palantir is a unique company because it essentially only sells to the best companies. If you apply to use the Palantir platform , you are more likely to be rejected . Because they only look at who will make the most profit . They only choose the best . That's why they make so much money , continue to grow , and their sales pipeline is excellent; If they need more money , they can go a little further down that path and continue to add customers .
Here, they only increased their customer base by 31% , meaning they only added 300 customers in a year . Even though they only added 300 customers , look at their revenue . By adding 300 customers , they doubled their revenue . But do you know how many thousands of other companies are out there that could make money for this company ?
Apparently it works. Palantir is definitely making them money, check this out. Their top 20 customers increase their spend by 55%, spending more than $108 million per customer . Dude, how big of a company do you have to be to spend $108 million on your company's database ? It's like spending $108 million on an old- fashioned Excel spreadsheet , and it's that good .
To be honest , it saves you that money. This is the thing people miss about artificial intelligence. People think, "Oh, artificial intelligence will make everyone rich ." This may be due to Elon Musk trying to popularize it among the public . No. I think artificial intelligence is making companies even richer .
I mean, talking about this would n't be so welcome right now. This will often lead to more socialism, which is actually quite contradictory. Because if this leads to more socialism, it will ultimately require more institutions, or more accurately, more governments, to provide social services to the people.
For example, like food stamps or benefits , right? To deal with fraud. Or these governments will have to deal with fraud . I'm trying to find this specific thing here . Yes, here it is. Look at this, which I wrote on the earnings call. Every aspect of government faces fraud . More government programs and more socialism, because corporate AI leads to more socialism , right?
More corporate AI or Palantir is needed to combat the fraud caused by more socialism . I will read this part to you. On the civilian side, the USDA awarded Palantir a $ 300 million contract last month to support American farmers, protect farmland, improve the supply chain, and protect agricultural programs from fraud, abuse, and the influence of foreign adversaries.
Now think about this trend , because it's very interesting in my opinion , right? The more successful AI becomes, the more the stock market rises, and the richer the rich get , the more AI companies make money . So, as money increases for AI companies , voters will vote in favor of social safety nets and socialism , right?
Programs, more financial aid, it's like a cycle , right? Governments will have to spend more on fraud prevention . They can do that with Palantir, which again makes companies even richer . Oh my god, this is a strange, strange cycle. I think it's an AI cycle . Anyway , let's continue with these financial statements , because they are incredible .
Let's see this. Therefore, net income rises. Operating expenses have only risen by 24%, which is nothing. Net income has increased 4 times , which is incredible . This is a reference to their ending share buybacks in 2023 . This is about the company's division and how it is growing. Their revenue has grown by 76% on the government side and 95% on the business side .
Both sides of the business are doing very, very impressively well . In the first 3 months of 2026, 53% of revenue came from government and 47% from business. So, you are not relying heavily on government or business alone . You rely on both . It could be the United Kingdom , it could be the United States, it doesn't matter.
Maybe not Russia . Therefore, that equal mix is very effective . If you look at the balance sheet, there's really nothing to see there. After deducting their deferred revenue, they have about $1.1 billion in total debt, meaning that for every dollar of debt they have, they have about $8.69 in cash and receivables.
And most of that money remains in cash , because receivables are actually a small part of it. This is truly amazing . This is a company that prints money. They are only harvesting the best now , so they will be printing money for a long time . They are rightly complaining about tokens being a commodity .
Right? They say tokens are like coal . When the price of tokens decreases , according to Jevons' paradox , we can do more work with those tokens , but we need to be careful about AI and LLM garbage , so we will control them through our platform . Right? Special, but we do n't want any more garbage. That's fair.
Okay, we talked about customer volume. What else was there to talk about ? Customer release. We talked about customer growth. We talked about fraud. Okay, good. Now, let's talk about the ratings of other companies . So, should I buy shares of this company, is there a possibility that it will give 4 times the profit?
Here are the facts. Currently, the company is expected to earn $1.47 per share . This is its predicted growth rate. So, this is until the end of 2026 . This is the forward P/E for those who say, "But trailing ratios are different." I know. I only use forwarded numbers . Wall Street's 4- year growth rate predictions may be wrong, but they average 48% per year .
This means that this company is currently operating at a PEG ratio of around 1.8. That's the price- earnings growth ratio, which I think is very useful because you take the PE ratio and divide it by the growth . I don't like to use PE ratios alone , because how do you know if a company is overvalued just by looking at the PE ratio?
You also need to take growth into account. This company is growing at 48% . So, if a company is growing at 48% at the end of the year , the fair value of this company is $188. Then what I decided to do was, I saw this, this is my assumption, right? There are no guarantees here. I looked at their 2030 EPS and I think they will only grow 30%.
If their growth drops below that, like Tesla went from 50% to zero, then the valuation will be wasted. But by the end of 2030, the company's value could be $563, based on its EPS and a 30% growth rate . That's about 4 times growth in 5 years. This is not bad when you expect continued growth. So, I wanted to compare this to some other companies .
Let 's look at Intuit. Let's say they grow 12% , which I think is the lowest since the 2030 EPS projection . If so, the company's share price would come to around $1,259, which is also a 4.5- fold increase . Both of these companies are generating about 4x growth. But, I'm asking for a 12% growth rate for Intuit, and a 30% growth rate for Palantir .
To be honest , both of these may be less . So, I wanted to try Salesforce. Salesforce has an EPS target of $21.40 for 2030. Let's assume that 15% growth continues . That makes Salesforce 5x growth , which is very interesting , because Salesforce recently took out $27 billion in debt for the next few decades .
Well, you know, every now and then you get three or four billion dollars . And they bought back $ 25 billion worth of their shares at prices between $175 and $210 . So, they are at a loss on the share buyback, although not by a large amount , which shows how confident they are in that application layer .
Remember, LLMs in between may become commonplace. Computing will probably become commonplace. The software layer, that's the profitable part. Again, my company, Reinvest, which we previously called HouseHack, I believe has that place in the real estate industry. We are developing our own AI . It is already in operation .
It works . It is on the househack.com site. It is in early beta stage. I think it's like Tesla's early FSD, we know it doesn't do some things right. It can go through red lights, just like Autopilot did in the old days . Doesn't stop at red lights. This is an early stage. We are building it . We are growing it .
It's exciting because it's out there , people can use it to give feedback , say this is wrong, this is right , we can adjust the weights and algorithms and train on more data . This is really funny . But then I did SpaceX. Now SpaceX is getting really interesting . Remember, all of these companies are very large, understand?
So I don't try to compare myself to them . What I mean is , Palantir is a $300 billion company. This is actually a lot less than 207 . Now, you can buy this company at a 38% discount from its all-time high . Look at a company like Micron , we know that memory is a commodity. What, leave it alone, sir.
Palantir is only 1/3 of Micron's value. Palantir has a strong lock-in effect . Now, they don't show their exit rate . Salesforce shows. It has been stable at 8% since around 2014 . So, despite all this AI drama about AI killing software, their exit rate has actually remained stable . I don't think so .
Anyone can create an app. Well, anyone can create an app. But no one can distribute it in a way that can be used by people's businesses, like companies like Palantir or Salesforce do. They are different, with Salesforce historically doing more for customers and Palantir focusing more on the enterprise side of the business .
I think using Palantir for CRM, like customer follow-up in CRM, would be like taking a cannon to a boxing match . But anyway, look at Salesforce now. Salesforce's valuation is more than a third of Palantir's, at $136 billion . So, these are huge companies. Check out Intuit. Intuit is a $75 billion company .
It is getting worse . Let's look at SpaceX. SpaceX is now valued at $2.1 trillion. These are very large companies. What is our valuation of my small company, Reinvest? I don't know . Only about a few hundred million dollars. Do you understand? So, a small part and most of it is due to our actual property value .
Because we have a lot of paid-for and completed properties . But anyway, we are a very, very small company compared to these giants. But I like to study these legends because I think the things you learn by watching these videos will help you understand yourself as an investor . That's interesting .
I never thought about those three layers . Three layers of AI . Two of them have become common items. Okay, now let 's look at SpaceX. We look at SpaceX and predict what Wall Street expects their earnings per share ( EPS) to be by the end of 2030 . I think that would be a year before the end of the decade.
I give them a software range or valuation of 2.67 times the PEG ratio. That gives them a fair price of 339, which is just 2x. So based on this analysis, it shows that CRM could grow by about 5 times, Intuit by about 4.5 times, Palantir by about 4 times, and SpaceX by about 2 times. This shows how valuable SpaceX is compared to others.
Furthermore, it also shows the availability of discounts on all the software stocks here. You can buy shares of Palantir, Intuit, and CRM as a package. By doing that you can diversify growth risk, right? Because you need 12% here, 30% here, and 15% here . That way you can diversify part of that risk .
You can also add Axon to this . Or you can buy IGV directly. But there are no axons in the IGV. I am sure IGV has all of these . I have no connection with IGV . I don't really care whether you buy IGV or not. But IGV has all three of them . You can look it up yourself. Search for IGV Holdings on Google.
Another interesting thing is that I posted it on the ' Meet Kevin' app , which you can download for free. I have posted this screenshot here . "Generative AI (GenAI) is driving a surge in mobile app releases ." But apps with significant usage and app reviews have actually declined . Meanwhile, on the 'Meet Kevin' app, the way people are currently using our app has increased 7 times , which is very exciting.
Time spent on the app has increased about 7 times year-over-year . This is really very good. I think we are trying to provide real value to people. Whether it's Alpha Wire or Data Tab or Stock AI or Reinvest AI or Homes AI . There are a lot of things in that app . I also use it as my task manager . I don't think many people will use Task Hack .
I like this very much . Anyway , yes, you can download it for free on Apple or Android . I hope you found this video helpful and useful. It took a while to create this , so I hope you learned something from it. If you learned something , consider sharing this and subscribing . I'll see you in the next video .
Why don't you publicize the things you said do a little advertising effort and see how it goes. Congratulations, friend. You have achieved a lot . People love you . People look to you as a role model . Kevin Baffrath, here he is. a pleasure to receive your feedback .
Comments 0