저자: Eli · 외국인
요약 보기
마이클 버리가 엔비디아·팔란티어에 대규모 풋옵션을 보유했다고 공개하며, AI 하이퍼스케일러들이 서버 감가상각 내용연수를 늘려 이익을 과대계상하고 있다고 주장. 버리는 2026~2028년 감가상각 과소계상이 약 1,760억 달러 규모이며 오라클·메타의 이익이 각각 약 27%, 21% 과대평가됐을 수 있다고 추정. 화자는 AI 기술 자체는 유용하나 대규모 투자와 계약 규모는 경제성이 의심스럽고, 회계 조작이 나타난다는 것은 문제의 신호라고 평가.
다듬은 글
Welcome back. As you know, I am Eli, the computer guy here for the Daily Blob, where I shake my brain nipples every day to get some of that dirty dirty YouTube money so I can afford to provide you folks hands-on inerson technology education that empowers you to do whatever the hell it is you want to do in Durham, North Carolina.
Silicon Dojo, silicon dojo.com. We offer uh classes on AI and web scraping coming up next week. We have a class coming up on a extending AI capabilities uh using REST API coming up in a couple of weeks. If you're interested in any of those classes, do make sure to take a look at silicon dojo.com to look at our schedule.
Do remember free to dandies is not actually free. That's why I shake my brain nipnips every day. If you want to throw some money into the donor box link down below, that would be ever so lovely. I will I will have to tell people out there. I get I get some people in the comment section telling me, Eli, Eli, you need to go back to doing tech education.
That's what people want. And all I have to say to you is go [ __ ] yourself. Go [ __ ] yourself with a goddamn cucumber is what I got to say. I've been looking at my stats lately. My stats are currently 1.3 million views per month, which to be clear for this channel is the the highest number of views I've seen since March of 2016.
So the reality is when I shake my brain nips, [snorts] line goes up. When I talk about SQL, line goes down. So stop lying. But anyways, talking about lying, talking about lying and hiding reality, let's talk about artificial intelligence again. The great the great fraud, the great grift of 2025, the AI idiocy, right?
I've been talking about this a lot. Absolutely ridiculous. LLMs are amazing. this whole new world of quote unquote artificial intelligence is truly paradigm shifting. Is it worth $50 gazillion dollars? No. Just not just not right. I've been talking about this for a while. So, I think one of the interesting things that's currently coming up is we're starting to see all of the business dealings uh for all of these different technology companies and how they're trying to goose the numbers to make artificial intelligence seem more financially viable than it actually is.
Right? So we have this circular economy where Nvidia puts money into Open AI to buy more Nvidia equipment and then Open AAI is putting money into Oracle and Microsoft buys part of Open AI and then Open AI then signs a contract in order to provide services uh for for Microsoft, right? We've been seeing that stuff going on for a while.
I think the interesting thing now thing that we're now starting to see too is we're starting to see how uh these big large technology companies are trying to obfiscate the actual books themselves. So they already have this circular economy idiocy going on. Uh and now it appears that they're actually trying they're not fudging the they're not fudging the numbers they are tweaking the numbers.
One of the things I've talked about for a long time is if you look at the technology world, one of the things that I argue for any uh young person out there that wants to get into the technology industry is you really should be thinking about your MBA or your jurist doctor. I am telling you there is more innovation going on in the legal and financial realm around technology companies than there is with the actual technology stacks themselves.
You you think LLMs are amazing? you haven't seen the crap the lawyers and the MBAs are coming up with, right? And so I think that's kind of an interesting thing going on right now with this whole idiot AI idiocy is tech companies are now going to their disclosures and they're trying to figure out how to make their disclosures basically look better than those disclosures should look.
So that's the whole thing, right? If you need a number to be here, you can either get the number there by profit and profit and revenue and all that kind of stuff. Or you can get the number there by by changing by changing how you uh report things uh in your disclosures to get the number up there. So this is coming uh about Michael Bur.
Uh so he's the one that predicted uh the the great recession back in the day. Um and uh and yeah, he is a shorting the hell out of Palunteer and Nvidia. And when you hear this, you'll understand a little bit better why this is this isn't simply him being anti- AI hype cycle. It's he's actually looking at how they're disclosing things and realizing they are changing how they disclose things in order to make numbers look better than they probably should.
So, big short investor Michael Bur accuses AI hyperscalers of artificially boosting earnings, right? So, when we put out the the disclosures, they have to put out how much money they're making. And wouldn't it be nice? Wouldn't it be nice if you could just simply fudge those numbers to make the numbers go up?
Michael Bur, the investor made famous by the big [clears throat] short, who recently royal the market with a tech short bet, is accusing some of America's largest technology companies of using aggressive accounting to pad their profits from artificial intelligence boom. In a post on Monday on X, the Scion Asset Management founder, alleged that Hyperscalers, the major cloud and AI infrastructure providers, are understating depreciation expenses by estimating that chips will have a longer life cycle than is realistic.
And this is one of the things you have to be thinking about if you really want to be a technology leader in the modern world. You're going to have to understand things like again accounting and that type of stuff. I talk about regulations and I talk about understanding contracts and terms of service, right?
Things like depreciation are something that you also have to understand one way or the other. Basically, what depreciation is is when you buy a large proh product, right? So, you buy a rack of servers. Basically the concept with depreciation is how how do you judge the the the decreasing value of that hard asset over time?
How long does it go from being a hundred thou an asset worth $100,000 to an asset being worth zero? Right? And then basically how do you deduct that? And also how do you take that off on taxes and that type of thing. And so what he's talking about here with the depreciation is if you have a normal let's say a three-year uh depreciation scheme.
So you buy something uh for let's say $100,000. So for the accounting uh practices instead of taking $100,000 off of your profit loss statement for the year, the idea is you can depreciate by by a third every year. So it's three years. So basically one year you get to you get to claim $33,000 in expenses and then the next year you claim $33,000 in expenses and the next year you claim $33,000 in expenses.
So instead of taking a $100,000 hit one year, that's really going to skew the numbers. You take you take that smaller hit. So with a whole depreciation scheme, one of the things is what if that $33,000 still seems high, right? You're still taking losses based off of that. How do you get that number to be smaller?
Well, one of the ways you can make that number be smaller is if you extend uh the term of depreciation. So, instead of being a three-year depreciation, you put it out to fiveyear depreciation. So, now if it's a $100,000 uh hard asset, instead of depreciating by $33,000 every year, you can now depreciate by $20,000 every year.
So, that makes your books look better. You still bought a $100,000 server rack. You still got the revenue and everything else going on. It is simply how you put that uh basically into your documentation uh for for investors to feel better or worse. So again, imagine imagine being able to take off billions of dollars in costs every year just kind of have them disappear because you'd extend out the depreciation.
Um [snorts] so anyways, quote under understating depreciation by extending useful life of assets artificially boosts earnings. One of the more common frauds of the modern era, Bur wrote, quote, massively ramping capex through purchase of NVIDIA's uh chips servers on a two to threeyear product cycle should not result in the extension of useful lives of a computer equipment.
Yet, this is exactly what all the hyperscalers have done. So, basically the idea is a life cycle for these Nvidia systems is 2 to three years. They're trying to push out the depreciation schedule past that to try to boost their their uh earnings. Uh Bur estimated that from 2026 through 2028 the accounting maneuver would understate depreciation by about $176 billion inflating reporting earnings across the industry.
He singled out Oracle and Meta Platforms saying that their profits could be overstated by roughly 27 and 21% respectively. So again, you hear about this depreciation thing and I talk about $100,000 rack of servers and you're like Eli, Eli, I mean that may be true. I mean it may be true but we got bigger problems.
Oh, we have so many bigger problems, right? But we got bigger problems than that. What does this really matter? Imag imagine for the value of stock prices basically if Oracle and Meta are are are changing things in such a way to overstate earnings by a [ __ ] quarter. We're not talking about overstating order earnings by one or two percent.
Right? Maybe maybe there's something where they need their earnings to be just a dash higher than they are in order to hit some mark. So, they're [ __ ] around a little bit. Again, that is what it is, right? No, they're doing this so they can overstate their their earnings or or their their profits their profits by 25%.
That's massive. When paying for large scale or large asset upfront like semiconductor, servers, etc., A company is then allowed to under generally accepted accounting principles or GAAP uh to spread out the cost of the asset as a yearly expense that is based on the company's estimate of how rapidly the asset depreciates in value.
So basically how quickly does it go from being a $100,000 asset to being a Z asset. If companies estimate a longer life cycle for the asset, they can then lower the yearly depreciation expense that hits the bottom line. Bur last week revealed a seemingly fresh wagers against AI favorites. uh Nvidia and Palunteer Technologies.
He disclosed put options with a notional value of about $187 million against Nvidia and $912 million against Palunteer as of September 30th according to a regulatory filing. Bur Bur said in his ex post that quote more detail was coming on November 25th and that readers should stay tuned. Um so yeah, so this is going to be uh the interesting thing to happen uh in the AI world.
I honestly again in my heart of hearts I have no idea what is going on in my own industry right now with this whole AI hype cycle. It is the exact same thing right when COVID hit and then all of these companies were hiring tens of thousands of employees I didn't understand what was going on and I thought if you follow me for any length of time when everybody was hiring I I didn't get what was going on again technology um you scale what's called exponentially instead of linearly right on one if you have one server you need one administrator if you have 100 servers you need two administrators if you have a thousand servers you need three administrators 10,000 servers you need four administrators right that's basically how we scale so if meta if co happened and everybody's doing work from home.
I'm not I'm not anti- meta hiring employees. They probably should have hired about a thousand about a thousand, right? You need employees to rack servers. You need employees to to run cable. You need employees, you know, to to improve all that kind of stuff. You don't need whatever they hired like 10 or 20,000 employees, right?
That that just didn't make any sense. Google was hiring. A meta was hiring. All these companies were hired because everybody's working from home. Argo, all these technology companies obviously need to hire gazillions of employees. I said it back then. It didn't make a goddamn bit of sense. And what what when happened, right?
Elon Musk fired 75% of of Twitter employees and then everybody started dumping employees in mass, right? Meta dumped 20,000 employees and and uh Microsoft this year has dumped 16,000 employees, right? They're just dumping these employees after they said they had to have so many employees for this this new world that we're in.
That's the exact same thing. I do not understand what's going on with this whole AI idiocy. Do not get me wrong. Like a lot of people say, "Well, Eli, if you hate AI so much, why are you doing AI classes for Silicon Dojo?" Um, clickbait. Honest answer is clickbait. It's AI washing. I teach you SQL with AI.
I teach you REST APIs with AI. But anyways, right. But no, but AI is fine, right? LLMs are fine. uh structure, being able to get pull structured data from PDFs is [ __ ] amazing or there's a lot of stuff with this. Cool. But there there's a difference again between a technology being amazing and valuable in the whole nine yards and being worth $4 trillion in capex expenses.
Like you hear these contracts that people like Sam Alman are signing and the numbers literally don't make sense. $350 billion to Oracle, $250 billion to uh to to Google, right? Tens of billions of dollars to to Amazon and and other companies like that. And you just look at it and it's not that the technology is not valuable and doesn't have a place and purpose and all that kind of stuff.
It's just it's just not that. It's just not that. Especially when we look at the wider market, right, with China, China putting out all their their open- source AI models, all that kind of stuff coming out the pike and you just you just look at the wider macro environment and what whatever the intrinsic value of of AI, LLMs, and all these models might be, the actual economic val the transaction, the commerce value, I suppose um uh Markx would say, um is a hell of a lot more questionable, right?
You know, everybody needs a bottle of water. You need water to survive, right? If somebody tried to sell you a bottle of water for a million dollars, you call them a [ __ ] [ __ ] right? But but you need water to survive. Yeah. But, you know, again, it's like you look at a bottle of water. You need it to survive.
How much will you pay for a bottle of water? A buck? Two bucks? Buck 50. You're not going to be spending 40 bucks on a bottle of water, right? And I think that's when I look at the whole AI thing is again, not that it's not important, all that other stuff. you just look at the reality of of the economics and all that and it just doesn't make sense the way they were the way they're trying to sell it.
And so what's interesting here is when we start to see these companies really start [ __ ] with their accounting. That's when that's when you know things are bad. That's when you know things are really not good because this this has gone beyond the hype cycle. This has be gone beyond an over an overexited CEO.
this has gone beyond this has gotten to the point where they know there are fundamental problems economically with this model and [clears throat] they're having to go to their accounting to their accounting to try to clean this [ __ ] up, right? In a way that's optically good uh for investors. So, what do you think about this?
What do you think about Michael Bur? Right? He's shorting Nvidia and Palunteer. He's coming out talking about basically how they're trying to uh to shine their books in order to make things look better than they are. What do you think about all this? Put your thoughts down below. If you like these videos, give us a thumbs up.
If you hate these videos, give us a thumbs down. Call me amazing. Call me a dumbass. Just be a real lutnick. Give us a comment with thumbs down there. Tell me how you really feel. Uh do remember what I actually care about is Silicon Doge. Silicon Dojo.com. free DND user hands-on technology education that empowers you to do whatever the hell it is you want to do in Durham, North Carolina.
Um, we have a class coming up on AI and web scraping next week, a class on extending AI capabilities with REST APIs in a couple of weeks. We have more classes coming up every few weeks. Make sure to go to silicon dojo.com to see what the current schedule is. Do remember freezer is not free. That's why I shake the brain dip nips every day.
If [clears throat] you want to throw some money into the donor box account, uh, there is a link down below. with that. See y'all
Comments 0