[00:00] Hi. Welcome to another episode of Cold Fusion. Over the last 5 years, Big Tech has [music] received valuations that far exceed anything observed in human history. It's the kind of growth that under possibly different circumstances, [music] people would likely celebrate. Conventionally speaking, higher valuations mean better pensions, a better [music] standard of living, and a sign of a prosperous economy. But of course, when the market gets too hot [music] and euphoria explodes, it can be more of a warning sign. In late June of

[00:27] 2026, the Bank of International Settlements, an international financial institution owned by the central banks of the world, issued a stark warning. AI spending [music] and circular financing has gotten so out of control that it's becoming a risk to the economy. Big Tech [music] is in a weird position at the moment. It's gone from driving the economy to putting it at risk. And the latest fuel added to the fire, Chinese AI models. They've closed the performance gap drastically and are approaching good enough for most AI

[00:55] applications while being multiple times [music] cheaper. You can download these models yourself, use them without an internet connection, >> [music] >> be 100% in control of your data, and tweak the weights however you like. American companies have started to realize this and they're switching. [music] Just look at the massive change in the AI landscape in the past year. The yellow bar is American AI companies [music] and the gray bar is Chinese. And it's not just small companies making the switch either.

[01:19] >> [music] >> Cursor, Coinbase, Shopify, Airbnb, Uber Eats, Siemens, and even Microsoft are switching to open source and open weight models. Nothing is for certain, but to me, this seems catastrophic to the central business case of the AI boom. What's the trillions of dollars in spending for if the promised future revenue is already vanishing? To be clear, I still do think that AI will be revolutionary in the long term. >> [music] >> Coding and healthcare applications are already showing promise and the systems

[01:47] are still improving, [music] but I just think the industry has mispriced LLMs due to excessive hype. Beyond the recent collapse of data center projects, there's There's of problems with LLMs when they meet the real world. >> [music] >> This one is key and basically sums up everything. A survey of nearly 2,500 companies found that for every dollar spent on AI, only 18 cents makes it into production. The rest of the money spent disappears into fixing what the AI got wrong. That could be fixing AI-generated

[02:13] bugs, reworking what the AI has done, or just overall friction. Any other technology with a return on investment this poor would be thrown out as trash, but yet the money [music] keeps piling in. Even the top AI CEOs appear to be backpedaling on their language. Sam Altman himself said that he was delighted to be wrong when it came to [music] his predictions of an AI-driven job apocalypse. Some analysts are terrified that this market has gone beyond a bubble and this is all actually a sign that the economy itself [music]

[02:42] is essentially broken. In this episode, we see how Big Tech is behaving strangely. There's a lot of money being thrown around with not much to show for it. >> You are watching ColdFusion TV. >> [music] >> Moving into the 2010s compared to today, most tech investment was so slow-moving that it might as well have been a corpse. Aside from the significant investor interest in Apple's iPhone and the App Store release, or Amazon's logistical brilliance and faster deliveries, the idea of Big Tech looked

[03:12] like it was taking a backseat. One of the biggest stories to come out about Google, now Alphabet in 2012, was how a Google research team created a neural network out of 1,600 computer processors. They tried to mimic the functions of the human brain. After 3 days of training on YouTube videos, they successfully got their network to identify the difference between a toaster and a cat. While that was impressive at the time, it wasn't exactly the next big thing that everyday people were pumped about.

[03:39] Though by the turn of the decade, the six largest tech companies, Amazon, Apple, Alphabet, Microsoft, Meta, and Nvidia, had a combined reported revenue exceeding a trillion dollars. But their valuation by the end of 2020 was almost 8 trillion. By the end of 2025, this had blown up to 20 trillion. And now, it's over 23 trillion. In fact, this growth is so monumental that many believe that the bullish tech sector is the only part of the American economy that's keeping everything from a massive recession.

[04:09] But, [music] if the current tech boom is a bubble, why does the market, at least, seem to be behaving differently than in the past? One of the main reasons is likely to do with AI and the construction of data centers. They've meaningfully contributed to a rise in global GDP. Analysts like Harvard's John Furman have recently spoken out about how this likely contributes to an overwhelming majority of US GDP, and how it really warps the sense of how good or bad things are. >> You can just add up what went into GDP

[04:37] or into demand in the first half of the year, and 92% of the increase in it came from just two categories of GDP, information processing systems, and software, which is to say, the stuff going into um data centers. >> [music] >> Hearing this, it's almost gotten to the point where GDP appears to be a broken indicator of economic prosperity, especially when considering much of this, quote, production and consumption is being driven by the same companies that are reaching unprecedented valuations. And this ties into the whole

[05:08] circular financing idea, which I'm sure you've heard before. >> A precarious investment strategy is emerging. >> Multi-billion dollar circular deals, the merry-go-round of money continues. >> Huge, huge sums >> of money are just being passed between these enormous companies that are are pouring hundreds of billions of dollars into the promise of AI. >> As the cracks between valuations and the reality on the ground start to get wider in 2026, >> [music] >> the behavior of big tech becomes stranger with it. It's possible that a

[05:37] lot of these recent earnings and valuations are artificial. In other words, [music] a lot of these companies could be playing clever accounting tricks so that the The investor believes that everything is fine. I'm going to play a clip from two forensic accountants, >> [music] >> John Weil and Kevin Koharki. It runs for a couple of minutes, but I want you to listen closely to what they're saying. >> [music] >> They really dissect the financial engineering and tricks of the hyperscalers. It's major companies like

[06:01] Microsoft and Google. Basically, when you crunch the numbers, [music] they don't make any money. >> And for those of you who aren't already familiar with John's work, he is basically the goat of accounting-driven reporting. Uh he's been doing it since the days of Enron and WorldCom. >> a story for you about companies >> [music] >> uh that are completely overstating their earnings, aren't profitable. Uh you could see that well, they start off with X amount of net income, and it looks really impressive. And then you look at

[06:29] all their capital, you look at their cash flow, still looks impressive. Then you look at how much their capital investments are, uh CapEx it's often called. Then you say, "Oh, it looks less impressive. They're spending a lot of money on data centers." And then you have to take into account the cash costs of stock-based compensation, and then buybacks that are related to the actual uh stock-based payments themselves, you take those two elements, and now you're down to like almost no earnings, almost

[06:56] no free cash flow. >> It's insane to think that just 5 years ago, before the AI boom, [music] these companies were printing cash. >> The free cash flow for the big tech names just falls off [music] of a cliff. Meanwhile, the rest of the S&P it continues, not because they're not making lots of money, but because they're spending [music] ridiculous amounts of money on these data centers specifically, and those names [music] have been falling. >> It's like, what's the next step and we're spending a lot of money. They used

[07:22] to be free cash flow machines, they're not free cash flow machines anymore.

[07:30] >> To start our exploration, let's begin with the incestuous relationship between Google and Anthropic. Google's first quarterly profits of 2026 reported an 82% [music] increase, which is roughly a $28 billion increase from the last quarter of 2025. [music] The headlines attribute this, as well as a 77% jump in Amazon, to the rise of their cloud units. But, if this was true, isn't it rather odd that Google has been quietly laying off staff in its cloud division to, quote, "reinvest in growth areas such as AI," end quote.

[08:00] >> [music] >> And it only gets stranger when you realize that Google's income statements attribute the root cause of this profit jump as, quote, "other income." A sudden magical sum of money coming in labeled as other income is kind of like catching your teenage son having $10,000 in the family shared bank account. So, how did you get that money, son? I don't know. A classmate gave it to me. Like a disappointed father, let's look into Google's shared bank account. We find that in April of 2026, Google has

[08:27] been consistently pumping tens of billions of dollars into Anthropic, and there's no signs of slowing down. One month after Google announced their series of investments into Anthropic, in May of 2026, Anthropic conveniently announced that they've [music] committed to spend $200 billion with Google's Cloud over the next 5 years. But, what makes this even more interesting is that, at least on paper, these two companies should be rivals in the AI space. Looking even deeper, a virtually identical pattern is found with Amazon

[08:56] as well. And there's so many examples of this, deals and money flowing between different AI companies who often should be rivals. So, of course, this could all be seen as a mutual confidence in AI, approval of respective strengths, and Google can come out of this saying that the collaboration was the reason for the layoffs. However, Sasha Yanxin, who has worked in creating financial products for big banks, sees it differently. >> You know, Google goes and gives Anthropic $10 billion as an investment, and Google receives

[09:23] the $10 billion back as revenue from Anthropic for using their data centers. Google gets a massive profit without actually having to do nothing. Then, Google gives Anthropic the same $10 billion again as a new investment, but at a new valuation that is twice as high as the previous one, then they receive that money back again in revenue, and they get a massive load of profit again, and so on and so on. Repeat 20 times, and you have a $200 investment without ever needing to actually have $200 or give it to anyone in the first place. It

[09:55] doesn't pass the sniff test, does it? But, there's a question that has to be asked. With all of this valuation and all of this money flying around, where is the return on investment? Large companies like Starbucks, which spent most of 2025 promoting an inventory tool which was meant to be applied to all 41,000 plus locations, have recently dropped the idea because it was a catastrophic failure. The AI would commonly misidentify and mislabel things, so staff had to check everything, and it was just a waste of

[10:23] time. >> [music] >> Duolingo CEO Luis von Ahn promoted the idea of AI-assisted learning. The plan was to cut staff, but the AI-generated output was so bad that the company pulled a 180 and begged for their consumers back. Even Microsoft CEO Satya Nadella has recently broken file and said that LLM models aren't enough and are unstable by themselves. He states that there has to be a proper ecosystem where the humans have the ideas and the judgment, and the AI tokens are only used to build things. Another recent

[10:52] development is the loss of faith in total AI replacement of employees. Meagan Slapinsky, district president of technology talent solutions at Robert Half, tells Fast Company, quote, "In many cases, organizations have had to reassess their expectations, recognizing that while AI can be effective in certain areas, it's not [music] the be-all and end-all solution some initially believed it would be." End quote. This is arguably more notable among software engineers. Box CEO Aaron Levie warned that AI fanaticism has

[11:21] actually led to major losses, citing a need for human oversight that artificial intelligence can't match. Quote, "You can get by for a while by being non-technical about building software, but eventually someone has to understand the thing that got built, has to maintain it, has to fix security issues that come up, upgrade the systems beneath it, and so on. That's all jobs. Some will celebrate because not as many jobs will be taken as first thought, but the celebration is muted. It reasonably leaves a bitter taste in many people's

[11:49] mouths. Not only were people unreasonably fired for an unproven technology, but the very big tech companies that were proclaiming a revolution were seemingly paying each other massive amounts of money to artificially inflate the valuations behind the scenes. >> [music] >> And now we come to one of the biggest examples, Nvidia's alleged backroom dealings with SpaceX. There's enough moving parts here to cause a nosebleed, but they also appear to explain the initial public offering that otherwise doesn't make a lot of sense.

[12:18] >> [music] >> During May of 2026, the 2008 crisis predictor, Michael Burry, expanded his bearish stance on Nvidia in relation to the AI boom of the last decade. >> [music] >> He shared a graphic chain diagram connecting pension funds, insurers, specifically Ethane, the private credit firm Apollo, and the AI infrastructure between Nvidia and Elon Musk's xAI. Among the highlighted deals is Nvidia's sale of 5.4 billion dollars of its most advanced GPUs, the GB200, to a company called Valor. Its reported purpose is to buy and lease

[12:51] data center infrastructure, mostly for xAI. But here's where things start to get questionable. If Michael Burry and other skeptics' accusations are correct, these [music] chips are technically owned and leased by Nvidia using a not-so-separate company to make it legitimate on paper, meaning that Nvidia was able to claim 5.4 billion in sales revenue, xAI was able to use the chips to operate Grok, while Valor sits in the middle serving as a phony bridge to make it all work. In plain English, Nvidia

[13:21] gets to report revenue, xAI gets to power Grok, and Apollo collects the fees. All while retirees unknowingly finance the hundred billion dollar operation using a theme. They fund it with their hard-earned pensions. So, looking at SpaceX's IPO, which is floating at about the two trillion dollar mark, you look at it a little differently, especially right after they purchased xAI for 250 million, which SpaceX attributed [music] to its total valuation. I touched on this in a previous episode, but what most people

[13:50] don't realize is that [music] SpaceX is no longer a rocket company. For their IPO filing, SpaceX filed under the category of computer programming and data processing. >> [music] >> And according to that very filing, 85% of their proposed market is AI. And that's the most charitable interpretation. It's super interesting and eye-opening if you haven't watched that episode yet. But after I released it, SpaceX's very first purchase was the AI coding company Cursor. And that only solidifies my point.

[14:18] While some people might be quick to call out critics like Michael Burry as conspiracy theories, it's hardly a theory when all of these companies have done it out in the open for everyone to see. While it looks like the SEC has fallen asleep at the wheel, there's nothing explicitly illegal about what these companies have done so far. If it's all what it appears to be, it seems that Big Tech [music] has found its way to have its cake and eat ours. They seem to have fabricated demand, given exit liquidity to wealthy

[14:43] investors, and gotten Grandma to fit the bill. But ignoring all of that, it just feels like Big Tech isn't preparing for more growth, but it's just looking for insurance on what's to come. I've always been optimistic about technology and the opportunities it creates. With all of this convenience, learning how to think through problems is becoming more important than ever before. And that brings me to something really cool that Brilliant has been working on. This is the new Brilliant, a personal tutor for maths and coding that

[15:09] sits right there while you learn. It can follow your reasoning, guide you through problems step by step, and adapt in real time to how you think. This isn't like traditional studying. Instead, Brilliant turns learning into something much more hands-on, almost like solving puzzles instead of blindly retaining information. I've been trying out some of the lessons myself, and one thing that stood out to me is how the tutor nudges you in the right direction without just handing you the answer. So, instead of passively consuming

[15:33] information, you're actively building intuition as you go. And if you feel stuck, Koji is there to help you. For instance, I was working through their thinking with Python course, and I got help directly from Koji to figure out the next steps. Their courses range from core math concepts like fractions and algebra all the way through to calculus, coding, logical reasoning, and debugging. And it's all designed by educators and experts from places like MIT, Harvard, and Stanford. And it can fit into your

[15:59] everyday life. Unlike traditional tutoring, there's no scheduling or waiting around. It's just there whenever you want to jump in and learn something. Whether you're learning from scratch, brushing up on old concepts, or just trying to keep your mind sharp in an AI-driven world, it's one of the more engaging approaches I've seen. You can try Brilliant's new tutor for free by scanning the QR code on screen or heading to brilliant.org/codefusion. And if you want full access to all courses, Brilliant is also offering Code

[16:24] Fusion viewers 20% off an annual subscription. Thanks to Brilliant for supporting Code Fusion. Okay, back to the episode. So, what do you guys think? Can you sense that the market and these big tech companies are behaving strangely? There's a lot of risk, financial imprudence, and not really any care for fundamentals. But maybe that's just me. Let me know what you guys think in the comment section below. Anyway, that's about it from me. My name is Dagogo, and you've been watching Cold Fusion. >> [music]

[16:50] >> And I'll catch you again soon for the next episode. Cheers, guys. Have a good one. >> [music]

[17:09] [music]

[17:14] [music] >> Cold Fusion. it's new thinking.