저자: Kevin · 외국인
요약 보기
OpenAI CEO 샘 알트먼의 수조 달러 AI 인프라 투자 및 복잡한 파생상품 조달 발언이 닷컴 버블 신호일 수 있다고 경계. 현 GPT-5는 환각 문제가 있고 추론 모델은 프롬프트 엔지니어링 대행 수준이라 플래토 우려. 코어위브 대형 계약의 칩 감가상각(6년) 문제와 OpenAI 실제 capex(수백억 달러) 대비 과장을 지적. 엔비디아·AMD 주가에 AI 초지능 기대가 과도하게 반영됐을 수 있다고 우려. 팔란티어 등 실용 AI는 긍정 평가.
다듬은 글
Hey everyone, me Kevin here about to head out to go flying and I was reading this Bloomberg article about Sam Alman wanting to invest trillions of dollars in artificial intelligence while simultaneously saying that we're in essentially a dot style bubble right now. I don't know what to make of this because here's OpenAI, a company that's worth about $500 billion in the private market, which is an insane valuation.
In fairness, they get a lot of nice subscription revenue, that annual recurring revenue. But that latest GPT5 release has a lot of folks wondering, hm, are we plateauing the level of growth we're seeing? Obviously between uh level three 3 and 1/2 phenomenal 3 and 1 half to four huge then we got reasoning models which basically let's just be frank reasoning models are no more than just doing the prompt engineering for you right rather than you going back and forth three times with different prompts to try to coax the AI into doing what you want it to do they try to do that part for you.
Uh obviously the goal here is to increase the user satisfaction. I actually particularly like Elon Musk strategy where potentially what we do is we take two different reasoning models and then have them reason against each other before we provide an answer. But I think what's happening is we're reaching this plateau in uh this encyclopedic talking to you.
It's basically a talking encyclopedia, right? This talking encyclopedia. We're reaching a plateau in it. Uh, and so we're constantly trying to figure out, okay, how can we deal with eliminating those hallucinations? Oh, well, let's reason. Oh, let's have multiple reasoning bots competing against each other.
But we're from talking encyclopedia to artificial superhuman intelligence, which is what Meta is pitching. Uh, and in fairness, I have good things to say about the Meta Glasses, though the Meta AI is total trash since it only translates like three different languages. Uh, I took it to Japan. The translating sucks.
The image recognition sucks. Like everything about the meta part of the meta AI glasses sucks, but the fact that you have a video camera and glasses, which I have them with me for this flight I'm going on. Amazing. But really a topic for a different video. You know, Warby Parker, they're going to be integrating these.
Apple will be integrating these. Uh Microsoft probably will. Google probably will. Like they're all going to get into this, right? But what I'm really interested in is Sam Olen. Uh oh, by the way, sorry, quick tangent. uh plan to send out either today or Monday uh one of my first buy alerts for sort of my decade hold pie.
So, it'll be a pie of about 10 stocks and uh these will be at least 10 year old buys and in sectors that haven't had these insane runups yet uh that we've seen already in some stocks like I mean Palunteer is really sexy, don't get me wrong, okay? I'm even so enthralled by Axon, you know, Palunteer for law enforcement basically and immigration.
Let's go. Uh, I'm going to Texas. Like, all right. Anyway, uh, but we'll be sending that out. So, if you access to all of the courses, lifetime access, lifetime access to all the buy sell alerts, and the alpha report every single morning, which we had a great alpha report this morning. Now, all that out of the way, you you got to ask yourself, Sam Alman, is he really just trying to find ways to raise more money because he realizes progress is coming a lot slower.
They released five. They had to come out with one of the old models because so many people were pissed off at five. Five was great seen as great for some things like certain enterprise things but worse for other things. I'm still seeing insane hallucinations by the way. Like I try to get it to tell me basic things but like hey you know really important things in my day-to-day life like you know what is the defense level requirement to equip a certain type of armor and runescape and he can't even get that right and it's GPT5.
It's pretty embarrassing. I just asked it about the difference between positive and negative jet fuel because we're fueling up the jet right now. That's where I'm going right now. And it's like, well, any kind of charge is bad. You have to make sure you ground it. And I'm like, no, no, you dumbass. It has to do with anti- ice.
No recollection or recognition of this at all. And I'm like, can't do Runescape. can't do aviation, what good are you? But anyway, Sam Alman's talking about coming up with creative financial derivatives to help them finance trillions of dollars of AI infrastructure spend. But wait a minute, y'all have like a 12 billion deal with Cororeweave, okay?
Where Corweave buys the chips and Cororeweave will bag the hold the bag of those chips, you know, when those chips lose value. So Coreweave holds the bag on the chips over time as they're fully depreciated. Meanwhile, they tell you, you know, they're doing so well because they're depreciating on a six-year time frame.
We'll be pretty lucky if these chips are able to function at these revenue rates and utilization rates in six years. But whatever, or you know, even three years, frankly, because then you'll have to take an extra charge. Also topic for different video, depreciation schedules. They're going to hit companies like Microsoft and Meta in the future, but whatever.
Right now, nobody cares. What I'm more interested in is why is Sam Alman signing a12 billion core lease now all of a sudden turning into oh yeah we need to spend trillions of dollars on artificial intelligence and like his messaging is a little weird. He's literally saying I understand we're in like a dot style bubble and I understand some people are going to think we're crazy but let us do us.
We're going to come up with creative complex derivatives to help us finance trillions of dollars of infrastructure spend for AI, which is again weird because Stargate is, you know, we've only got commitments of like 80 to hundred billion for the Stargate Open AI initiative. We've got capex spend of under $50 billion so far from open AAI, you know, over the last couple years.
So when you combine this, it's like, cool. So yeah, we're talking about a couple hundred billion bucks in capex spend with open AI. Microsoft and Meta are kind of matching this. Yes, we are talking hundreds of billions of dollars. We know we see it in Nvidia's revenue. Okay, but where's trillions coming from?
So my take is there are a few things going on here. Number one, Sam Alman realizes in order to actually get to superhuman intelligence, we don't need to spend a hundred billion. we need to spend5 trillion dollars. Which then you sort of like teeter on the question of okay, are we gonna spend a quarter or a fifth of our US GDP on an uncertain superhuman intelligence that you might spend the money on and then realize crap, we're just stuck with a talking encyclopedia.
Again, that's my like I'm not trying to be bearish on AI. I think talking encyclopedias are great when you could trust them. I think there are huge advantages of what AI can do in medicine and accounting and policing. You know, what Palanteer is doing, what we're doing over at House Hack uh with with uh the identification of wedge deals.
You know, this is like turning my brain into an AI so we can license it to everybody in terms of what I think a good deal looks like. You know, that's those are profit generating opportunities. Obviously, you know, by the time we launch that and and if we pull it off the way we think we will, we're going to stop our fund raise because we don't think we'll need to fund raise anymore.
But that's a totally different topic. Not trying to solicit an investment. I'm just saying this is an AI video. Okay. So, speaking of AI, I mean AI's basically been driving the whole time here. I don't even know why I took over right now. I should just let it keep driving because it's a right turn on red, but it doesn't let me activate on our Oh, there we go.
Oh, it let me activate. Okay. Well, we'll see how it does. Uh I'm also going to Texas right now. We're going to be flying to Texas. is we're going to be checking out. Well, I got a project for Houseack I got to do in Dallas. Uh and then we are going to Austin to check out the robo taxis. Uh and then I might pop over to San Antonio.
Uh we'll see what we want to do after that. Uh but uh got a good weekend ahead. So, uh but I want to go back to the Sam Alman thing because it's odd to me like on one hand you're telling me you want to spend trillions of dollars. On the other hand, you're telling me we're in a dotcom bubble. And in an effort to finance your trillions of dollars, you're trying to tell me you want to come up with a complex financial derivative to help you finance your trillions of dollars of investment.
Even though that's 20x or more than 20x, 30, 40x what you've spent so far. And the best that we've gotten is G. Oh, that's a that looks like a Gulf Stream coming into the land ahead of me. That's cool. But anyway, even though you are moving what I would say is quite incrementally like let's put it this way.
I don't when I put together my alpha report for news or when I make a video like this, I give you what I think is just pure human intelligence. Now, when you're using me as a case subject, there's not even a lot of human intelligence. We're talking room temperature IQ here. But I think that my perspectives are certainly unique relative to AI.
And so, this is why when I make videos, I refuse to use AI. So, when I like on the Meet Kevin app, which you can download for free, and you see sort of like my outline of news, when you look at this outline of news, I don't use AI to put that together. That's Kevin's writing. And I make little notes as to like, is this my opinion or is this like a data point, right?
The green check mark versus the halo emoji. I do that because I want you to know what is an actual human thought. And frankly, when I've used these GPT Pro, uh, you know, the the um, oh god, what was the last model called? The, uh, four, I forgot what they call it now. 40 Pro or something like that.
Anyway, I pay for I pay the 200 bucks a month. I really feel like I'm getting ripped off. But I'll say like, "Hey, you know what's the news for today?" And it'll reason for like 30 minutes. And then I'm like, "Man, that's pretty surficial." Like in that time, I could have just read the articles myself and then had my own perspective on them.
I feel like it's just headline reading sometimes. Again, I'm not trying to be a bear. You can get into good prompting. You can get into good conversations. You can get into good legal drafting. You can get into really good work. I'm not trying to be a bear. I'm just saying the talking encyclopedia has its limitations and we're coming up against resistance.
And I think that's why Sam Alman is like, you know, we don't need Stargate and your stupid 80 to$100 billion of current commitments and maybe up to $500 billion, which part of that and most of that is probably just a Donald Trump knee bend. We need actually trillions of dollars. Trillions. and in order to finance it because we can't get enough money, we're going to come up with some complex financial derivatives, but but be careful because we're also in a.com bubble.
I'm like, when I start hearing about derivatives and complex instruments to raise money, I start thinking about bubble. And this is exactly what I would expect a bubble boy to say is, "Hey guys, everybody else is going to suck, but trust us. Give us trillions of dollars and after you give us trillions of dollars, we will make it work.
Trust us, bro. I don't know about that. Oh, they're setting up for an air show over here. Maybe I could show you some cool planes. Stick around for another minute and we'll see if I can give you a tour of some cool planes. I think they have an air show this weekend. I think I'm getting out right in front of it.
But think about that for a Why do I always do that? Yeah, they put the window thingy right where Oh my gosh. I think there are jets over here. They put the window thing right where I put my elbow when I'm driving. Uh, in the Cybert truck. I don't know, maybe it's just me. But, uh, I'm not complaining about the Cybertruck, okay?
I like the Cybertruck. So, if you're a Tesla fan, I'm not trying to be a Tesla hater. Okay. All right. Uh, but anyway, so I put all this together and then I think to myself, I don't know, Sam Alman, you're starting to make me nervous because every time Sam Alman talks like this, Elon Musk ends up inevitably coming out and saying things like, Sam Alman doesn't have the money.
He doesn't have the intelligence. We outperform them. Who's a scam? Sam Alman or Elon Musk? Reply with one word. And then, you know, it's always Sam Alman because otherwise Elon wouldn't be sharing the screenshot. Who's more trustworthy, Elon or Sam Alman? It's always Elon again. Otherwise, he wouldn't be sharing the screenshot.
So, that just leads to then more potential malinvestment from Oh, this is sick. Uh, look at this. Hold on. Oh, dude. Look at this. Hold on. Okay, there's my bird right there. But look at this. Captain Matthew Davenin. Dude, that is sick. That's an extra fuel tank. I'm pretty sure down underneath. I don't think they have bombs out here.
Anyway, let me make sure I don't cause any problems here. Yeah, because there's a plane taxing right there. Good lord. Look at this. How cool is this? This is sick. Sorry, I'm like nerding out a little bit about the aircraft. Uh, anywh who, so all of that makes me a little nervous. Now, look, I love Nvidia and AMD and the money these companies make, but I have to say I get a little bit nervous over uh like, hey, how much not how much are they affected by this investment from Sam Alman, I know positively, but how much has the stock market uh been, you know, sort of pumped up under this impression of uh Oh, there's the plane that was landing coming in.
But how much has the stock market been pumped up under this belief that Yeah, boss. We's going to super intelligence and uh we's going to win. Uh you know, I I don't know. It's uh it's a little scary to me. So, we'll see. But it is what it is, I guess. So anyway, uh, thanks for listening to this. Uh, I do think we'll get a lot of cool AI in these birds over time, by the way.
Uh, pretty excited about that. And, uh, yeah, folks, see you in Texas. Ice here.
Comments 0