Everyone Misread OpenAI's Leaked Financials. The Real Number is Terrifying

0:00 So, OpenAI's financials just leaked, and the internet immediately split into two camps. One says it's a ticking time bomb, that the company is hemorrhaging money, and all the way it's on the way to the biggest IPO in history. And the other says, "Actually, it's not that bad at all. Maybe this is even better than we thought it should be." Except, the truth is that both sides are fighting over the wrong number, and understanding this is critical because we're about to form an opinion on one of the single most talked about companies of our lifetimes. And being able to read through the hype to understand what is true is going to impact the stocks you buy, how you view the AI story, whether you are freaking out, or whether you just feel cool and calm despite all the hysteria. And understand if OpenAI is actually a ticking time bomb or the next generational success, we're going to do what most won't. We're going to actually look at the numbers that really matter.

1:00 And we're going to do that by running three critical tests. The first is the comparison test, the second is the Enron test, and the third is [clears throat] the Ritter test. And together, these reveal whether OpenAI is the next potential Amazon of our generation, or whether it's a ticking time bomb on its way to destruction. But first, we need to agree on the single most important number, which no one on the internet can seem to agree on, which is how much money did OpenAI actually lose?

1:32 So, everyone's talking about how much money OpenAI lost, and some are arguing it's a lot, and some are saying it's not that much. But the truth is, it's the wrong thing to care about altogether. Losses alone don't actually tell us anything. Many generationally successful companies lost apocalyptic sums of money before they turned the corner and became what they are today. But while that number alone doesn't tell us anything, agreeing about which figure to use affects all of our analysis that comes next. And, it also illuminates a critical point that I've seen raised in the comments, which is none of this really matters because these companies are too big to fail regardless. And, while it might seem like these companies are too big to fail, many companies, exactly like them, have failed and can fail even with major backers like the AI companies today. We can look back at pets.com. This was the poster child of the dot-com era, and Amazon was its largest outside investor and owned more than half, and it still collapsed because it burned through all of its cash before the business could ever turn a profit, and it couldn't raise any more money, and ultimately it disappeared.

2:48 And, that risk is very real for these AI companies. So, now, if you're not steeped in this world, this whole thing might sound pretty strange, right? Like, losing money always has to be bad, right? But, it's not necessarily. AI companies are making a bet that this is what economists call a winner-take-all market, and the idea is very simple. You get very big, very fast, even if it costs a fortune, because you need to grab the whole market before anybody else can. So, we can think about this like Google and search. Almost nobody uses, you know, Bing or Yahoo, so you have to lose money on purpose for years to win that land grab because if you win it, it's ultimately worth it. Now, whether this turns out to actually be a winner-take-all market is a subject for another video. It may very well not even turn out that way, but as a classic example of this, back to pets.com back or Amazon, Amazon lost money, or barely broke even, for the better part of a decade. And, in the year 2000, what a lot of people don't realize is it stock even fell more than 80% in a single year. And yet, today it's one of the most valuable companies on Earth. Well, they were able to survive this huge crash in value for a very specific reason. Amazon had the cash it needed, and it didn't have to raise more money at the bottom. So, they could just wait patiently for all their bets to pay off.

4:20 But, our question is, can Open AI do the same thing when, not if, but when the markets turn? And it starts with our loss number. So, if you feel as overwhelmed and over-clocked by the pace of everything as I do, the thing that I use almost every day is New Calm. So, as this channel grows, more and more folks reaching out for partnerships, and it's important to me that I only share things that I genuinely believe in, and New Calm is one of those things. Now, there's a huge amount of science behind their technology, but rather than share all that, which you can easily Google, I'm just going to tell you what I do.

4:55 When I feel tired or unfocused or stuck, or I get that like tired but wired feeling, I put on headphones, I put on eye shades, I close my eyes for 20 minutes, and I wake up clear and re-energized and ready to go. So, the two sessions that I run most often are a 20-minute rescue when I'm slammed, and a 40-minute one if I've got the time. You just It's headphones on, eye shades on, and it's full relaxation. And you wake up, you feel ready to go. I've logged literally thousands of sessions over the years. It's one of the few things that reliably pulls me out of the whole tired but wired exhausted state that comes from having two kids and having multiple businesses. I run a ton every week, and so on. The link's in the description that gets you 7 days free. Full disclosure, that is an affiliate link, and with that, back to the video. So, according to leaked financials, Open AI lost 38.5 billion dollars in 2025. And to put the size of that loss in the context, it is more than 10 times Nike's annual profit. It is more than Starbucks brings in selling coffee across the entire planet in a year. And it's the equivalent of losing $73,000 every single minute of every single day.

6:09 But that $38 billion number has to be put put in context because most of that money isn't actually money that OpenAI spent. So last October, OpenAI changed what kind of company it is. It converted from a non-profit into a for-profit. And when a company does that, accounting rules force it to report a huge one-time charge on paper. Very simply said, no money actually got spent, no money changed hands, it's an accounting technicality. So if you strip out all the accounting voodoo, you get the actual revealing number, which is an operating loss of 21 billion against about 13 billion in revenue. So the it's not that bad crowd likes to push that number even lower and they strip out a bunch of non-cash items. And what they find is the loss loss drops to around 8 to 9 billion. But that number leans very heavily on cheap computing credits from Microsoft and those credits will run out. You can think of this very simply like, you know, dad is subsidizing the rent. So kind of seems like you're making plenty of money. But dad's getting pretty impatient and you better start earning money soon, otherwise he's going to cut you off. So the reliable number that we can use here is just the most straightforward one, which is the 21 billion in loss against the 13 billion in revenue. So now our real question becomes, is that loss crazy or is it like totally reasonable when you compare it against other companies that got really big, really fast, and ultimately succeeded. And understand that we must run the first of our three critical tests, the comparison test. So, to know if a $21 billion loss is insane or totally normal, you can't look at it in isolation. You have to compare it against the right companies, and you can't do it against just pure software firms or steady industrials. You have to look at companies that at least sort of look like OpenAI, which is from my point of view three major characteristics. You need a major technological leap, you need enormous capital costs, and you need a winner-take-most race where there's only going to be a a few players that survive. And three fit, Amazon, Uber, and Tesla. None of them are perfect, but all of them ran the same playbook that OpenAI is running now, which is lose a fortune early to own the market later. And we have one question that we need to answer for each. For every dollar they made, how much did they lose? So, Amazon first. Amazon was a small online bookseller that decided to grow at any cost. And the bet there was they would build warehouses up before the demand showed up so that they could own e-commerce before anyone else could, and it worked, but it took a very long time. Amazon lost 390 million on 1.6 billion in revenue. And then the dot-com bubble burst and the stock fell over 80% in a year. But ultimately that loss didn't matter. The bet won, and we all know how the Amazon story turned out. So, at that time, during the hypergrowth phase, how much did it lose to get to domination? It was losing about 24 cents on every dollar of sales.

9:18 So, we got to keep that number on our head. Now, Uber ran the exact same playbook a decade later. And their play was flood every city, lose money on every ride so that they could own the network. In 2018, during their period of hypergrowth, right before the IPO, they showed a $3 billion loss on $11.3 billion in revenue. So, that was a 27-cent loss on every dollar spent, almost exactly the same as Amazon. So, what it seems to be is that's the price of admission for running this playbook, right? I know, it's only two numbers, but indicative. Now, let's look at OpenAI. For every dollar it makes, it loses $1.60. That is more than six times what the winners lost, and it is at a far larger scale, and we'll talk about why that matters in just a second.

10:07 So, a question that I sat with was, "Well, what about a company that was losing even more and still made it?" Well, there's Tesla. In 2010, it was absolutely lighting cash on fire to build the Model S. And by IPO, it had racked up over 290 million in accumulated losses, which was more than double its revenue, which is a far worse ratio than OpenAI on its face. But Tesla was tiny in comparison. They only had 117 million in revenue and 290 million in total accumulated losses. And when you're that small, huge losses are survivable for a simple reason. Raising another 100 million dollars is nothing compared to raising 100 billion. So, let's put these huge numbers in the context. If we took Tesla's entire revenue base, right, that 117 million, and then we stacked it up in $100 bills, it would stack up about 40 feet, which is a four-story building.

11:04 Now, if we take OpenAI's revenue base, which is 13 billion, it stacks nearly a mile into the sky, which is more than the tallest building on Earth, which raises the next critical question. Despite what seems to be frankly apocalyptic losses, is there a path where, despite all this, OpenAI eventually makes money? To answer this, we need to run test two, the Enron test. So, the Enron test answers one critical question. Can OpenAI eventually make money or is it built on a foundation so flawed that it will never be able to hold weight? Well, one of the most Wall Street investors is making the argument for the latter. So, Jim Chanos, who we've met in other videos on this channel, is the short seller who 25 years ago saw through Enron. While everyone else bought the story that basically this energy company was a technology company that had this limitless potential. It couldn't possibly be be valued. Chanos asked a very simple question. For every dollar this company spends, how much does it actually earn back? And the answer was 7 cents, which was less than what it actually cost them to borrow the dollar in the first place. And now he's pointing that same lens at AI, and he's calling it one of the biggest risks he's seen in decades. So, let's do exactly what Chanos does, which is forget the story and look at the math. But first, a very important point. Enron was outright fraud. And Chanos does cite fraud in some cases around AI, but this is in the ecosystem more broadly.

12:46 This is not OpenAI specifically. We have every reason to believe OpenAI's numbers are legitimate. So, what we are trying to do here is simply run the test. Can the engine, the revenue engine, ever earn back more than it burns? And right now, three things say no. The first is that the gap is massive. OpenAI spent $34 billion to make that 13 billion. And unlike a normal software company, where you build it once and then all your new users are basically profit, OpenAI has to pay every single time a query gets made on its system. What that means is that the single biggest expense it has is going to grow as the company grows, not get smaller. So, getting bigger doesn't necessarily fix the problem. It likely makes it worse unless, and this is a big unless, the cost of compute comes way down. Now, two, the costs are about to get worse. As Channels explains, the entire industry runs on Nvidia chips, and the data centers that own those chips spread the cost over 6 years, but the thing is the chips don't last 6 years. Nvidia keeps shipping better generation chips almost every single year. So, what turn So, what turns out they're worn out in about three, and you can already see that happening where rental rates for last year's chips fell already 28% in 12 months. And when that reckoning hits, compute will get more expensive. And OpenAI rents nearly half of all of its compute, so that's likely to keep going up. And the third is that their bill is already locked in. Per an HSBC analysis in the Financial Times, just to stay viable, OpenAI has to grow revenue from 13 billion to over 200 billion a year, and they have to raise another 200 billion in capital by 2030. All of this is against more than 600 billion in data center spending that it's already committed to.

14:42 So, just to keep the lights on, it has to 15x revenue in 4 years while costs climb, and then while raising more money than almost anyone in history. So, because that was a lot of numbers, let's just look at this thing in simple terms. You can imagine a lemonade stand. To make $13 in sales, it has to spend $34. That's OpenAI. And the argument might be, well, right now, they're paying a lot for the lemons and so forth because it's a very small stand. So, once they get bigger, all those costs will come down. Except, the cost of lemons is about to go up, not down. So, that stand has to become 15-times bigger in 4 years while lemons keep getting more and more expensive and more and more and more competitors keep entering the market with cheaper lemonade that tastes almost as good, which brings us to our final question. Despite huge losses and locked-in bills, is it possible that OpenAI is still a good investment when it IPOs, that this time truly is different? The Ritter test answers this with one ratio, price-to-sales. So, there's a finance professor named Jay Ritter who has tracked basically every IPO in America since 1960 and across all that data, he found one number that predicts better than almost anything whether a newly public company will make its investors money or lose it, and that ratio is price-to-sales. And the simple math is you take the company's value and you divide it by its annual sales and it tells you how many dollars you're paying for every dollar the company actually brings in. And here's what [snorts] he found. Of the companies that went public above 40-times sale, 12 out of 14 went on to underperform the market over the next 3 years. So, by buying those stocks, you would lose money. On average, of those above the 40X IPOs returned just 3.1%, so buyers made basically nothing. And Ritter is not shy about applying it to these companies. At a $2 trillion valuation, he said that he'd short SpaceX because even a great company can be a terrible investment at the wrong price. OpenAI is going public at a target valuation of around a trillion dollars on about 13 billion in revenue.

17:00 That's roughly 77-times sales. That is almost double the line where history says that you stop making money. This is a historic detachment of price from sales. Hey, future me here. So, since we started filming this, OpenAI has delayed their IPO. Um we'll see if it affects the ultimate target valuation, but the analysis stands as it is. I just wanted to drop that note in here uh to flag it. So, let's recap this. Test number one is the comparison test.

17:30 What we found is the losses are bigger than any survivor that we analyzed here. Test two is the enron test. The engine, as built, can't earn its way out. And test three is the ritter test. The price is 77x earnings, roughly double the profitable ratio. So, then your only question becomes this. Is this time truly different? That is the entire bet. And I genuinely like to know what you think or how you'd update this analysis and what else we should all consider. So, if you'd like to explore this question more deeply, watch the IPO video next. And if you'd like to understand how AI execs are running the big tobacco playbook for PR, watch the circus trick video next. And if you've seen those, I'll leave you with a few others for you in the description that you might enjoy.