ALERT: AI Credit Spreads Are Suddenly Blowing Out... Just Like 2008?

Speaker A: M Everyone believes the AI bubble is mainly about stocks, and that makes sense. The obvious historical comparison is the dot com era. However, this one is all about credit, unlike 1999, and that credit is now talking loudly through credit default swaps and other debt market signals. Companies like Oracle and SpaceX have seen their CDS prices surge recently with big moves in July across all the big names in AI tech. Amazon, Meta, negative cash flow for the first time. Google, Alphabet, whatever the company calls itself, Nvidia. They're seeing higher premiums as well. Even BlackRock trying to sell an AI linked bond. Similar results to what we talked about with Amazon's bond offering. The debt problem is spreading and deepening as the debt cycle turns more and more. Back in the dot com days, investors assigned extraordinary valuations to technology companies based on profits expected years into the future. And today, once again, investors are questioning whether sky high expectations will match reality. Semiconductor stocks are falling, the Nasdaq 100 flirting with correction territory. Even companies reporting record profits are being punished because the results can't keep up with the enormous expectations surrounding AI. And increasingly, investors are no longer just asking how far can these companies grow? Instead they're asking how do I protect myself if they're wrong? The stock bubble that everyone is expecting to find in valuations is instead showing up in the availability and price of money itself. For the first stage of the AI boom, the largest technology companies appeared capable of financing almost everything themselves. Amazon, Microsoft, Meta, Alphabet, Oracle, they all generated enormous revenue, in most cases substantial operating cash flows. They also possess some of the strongest credit profiles in corporate America. But the scale of the build out and the uncompromising, almost religious way every One of these CEOs has embraced the tech future. It has changed the equation. AI related debt issuance reportedly reached 270 billion by early July 2026, which was already close to twice the amount issued during all of 2025. Amazon, the same list. Alphabet, Microsoft, Oracle, Meta. They accounted for approximately $194 billion of that total. These companies aren't borrowing because their ordinary businesses have stopped working. They're borrowing because even their tremendous cash flows are being overwhelmed by the sheer cost of competing. Each company believes it must build immediately because falling behind in the AI race could be, in their view, fatal. But when every company reaches the same conclusion, they all compete for the same chips, electricity, construction capacity and capital. The AI race therefore becomes a financing race. And a lot of that spending isn't going to feel winning. It is fueling expensive eventual losing. And in credit markets, the winner is not Necessarily the company with the most exciting technology. It's the company that can eventually generate enough dependable cash to support everything it borrowed to build. The dot com comparison is useful, but as critics contend, it's often misleading. Back then you could just add.com to your name and the stock market would take it. From there you get equity capital despite having little revenue and no viable route to profitability. Today, the largest AI spenders are completely different. Amazon, Google, Meta, the same list we keep coming back to. They're all profitable global businesses with valuable assets and huge already existing customer bases. That's why many people are rejecting the bubble comparison. And they're not entirely wrong to do so. But a bubble doesn't require the underlying technology to be fake. The Internet was revolutionary. Fiber optic networks were useful. E commerce eventually transformed the world. The problem was the price paid and when it was paid, the capacity and overcapacity being built and eventually what really brought it down. The assumptions that were used to justify everything. The same distinction applies to AI. Artificial intelligence can be historically important while investments made in its name still produce terrible returns. Every data center doesn't become profitable merely because AI adoption grows. Every chip order doesn't generate sufficient revenue. And every company doesn't win simply because the total market gets bigger. The relevant question is no longer whether AI will matter, it's going to. It's whether the incremental revenue from AI will cover the incremental cost of building and financing. It costs, by the way, that have absolutely blown out here in 2026. That is where the debt market enters the story. Stocks price the dream credit asks whether the dream can make the payments. I almost never do ads and I say no to the vast majority of those that are brought to my attention. But I was thrilled to be able to work with today's video sponsor, Monetary Metals. And the reason is because they're solving a big problem. One that I've been highlighting for a very long period of time. Gold is a safe haven, but other than sitting in a vault, it doesn't do anything. Monetary Metals fixes that. They let you earn real yield paid in physical gold without giving up ownership. It's not a fund, it's not a paper claim. It's your gold working for you. Every month more gold gets deposited into your account. It's simple, it's transparent and it is fully insured. If you believe in gold as a long term store of value, as I do, this changes everything. Go to monetary-metals.com Snyder that's monetary-metals.com Schneider for all the details, we absolutely thank them for being the sponsor of today's video. And now let's get back to it. And the clearest indication of this changing mindset comes from credit default swaps. Cds, the old market that we once talked about back in the 2008 crisis. Well, it's back at the forefront, at least here in July 2026. A credit default swap or CDS functions somewhat like insurance against a borrower failing to repay its debt. A buyer pays an annual premium to a protection seller and if a defined credit event occurs, the contract provides compensation. The higher the premium, the more expensive it is to obtain protection. And that's really the signal, not necessarily default or default probability, but the demand for protection among investors. Oracle's CDS spread reached approximately 215 basis points this week, up from about 145 basis points at the end of last year. And this is the highest it's seen in several years. SpaceX, that's a big one. Its swaps reached approximately 185 basis points, rising by more than half since trading began just last month. CDS contracts tied to Nvidia, Meta, Amazon Alphabet, all the same names we keep coming back to. They have also reached new highs, though at lower levels than Oracle, SpaceX and another one, Coreweave. These numbers aren't a forecast that default is imminent direction. However, it's significant, as is the big names that are on the list. A year ago, the main concern surrounding a high profile AI related debt offering would have been securing an allocation. The dominant questions were growth and yield. With SpaceX, dealers were circulating prices for five year default protection before the bond offering had even been formally announced. Investors wanted a hedge fund before there were bonds to hedge. That is a huge behavioral change. The conversation is shifting from how much can I buy? To how do I protect myself? A credit cycle doesn't flip only when companies default. It begins to shift when lenders stop assuming everything will work and start charging for the possibility. The growing possibility that it won't. And this growing caution is emerging because AI adoption can be interpreted in in two very significant ways. Millions of people use generative AI. Businesses are experimenting out, uh, the wazoo with copilots, coding tools, search assistants, automated customer service and a whole bunch more. The usage is real and the technology does continue to improve, but usage, it's not the same as profitable usage. Many AI services are offered for free, at least currently. They're bundled into existing subscriptions, are priced below the full cost of providing them. Inference requires computing power. Every time a user generates an image processes a document or asks the model a question. And that creates a potential problem that didn't exist in the same form for traditional software once conventional software was built. Serving an additional customer could be extremely cheap. With generative AI, each additional customer can bring continuing infrastructure costs, and the industry is betting those costs will decline rapidly. And as adoption will become deeply embedded and companies will eventually obtain substantial pricing power. And maybe they will. But cheaper models, especially from China, are challenging the assumption that the most expensive model necessarily creates the strongest competitive advantage. If comparable performance can be achieved with a lot less computing power, today's massive infrastructure plans could prove excessive. The adoption projections may be broadly correct, while the revenue estimates are still wrong. Especially if, as we're seeing now, users balk at keeping up their AI spending while not getting the productivity return they signed up for. That's exactly the kind of uncertainty credit markets are built to price. And the credit default swap market, it's not acting in isolation here. I talked last week about the Amazon bond sale, and there was another one with BlackRock. A lot of it is all the same thing from different angles. As we talked about, Amazon sold $25 billion worth of bonds across eight tranches with maturities ranging from three to 40, and initially the orders reached approximately 62 billion. But after the banks managing the offering reduced the spreads offered to investors, orders fell to roughly 41 billion. That left demand at only slightly more than one and a half times the amount of bonds available. Now, for an ordinary borrower, that might have been perfectly acceptable, but for Amazon, it was remarkably weak. High grade US bond offerings have typically attracted orders around four times their size. Just this year, Amazon also had to provide an elevated new issue concession with some of its longest bonds reportedly offering an additional 18 to 21 basis points relative to comparable existing debt. So Amazon got its bond through. That wasn't the warning. The warning was that, uh, investors required more compensation and provided much less enthusiastic demand than the market had come to expect, suggesting something has changed. Then came a 12 and a half billion dollar data center bond offering arranged through BlackRock entities. The bonds financed BlackRock's stake in a Texas data center project being developed for Meta. The offering took nearly a week and encountered weaker than expected demand, just like Amazon's did. And it ultimately priced at a yield of 7.53%, which is 1 of the highest yields for a blue chip data center financing since the current AI borrowing binge began. This is huge. Again, the bonds got sold, but credit stress first appears not as a locked door, but as a Higher price for walking through it. And after all that, with the bond market stress starting to show up and bond markets uneasiness, consider what's going on with Google or Alphabet. The negative free cash flow for the first time in the company's history, emblematic of exactly all these concerns that we keep seeing in the marketplace. That negative free cash flow was 5.9 billion, the first since Google became public in 2004. Again, this is not a solvency crisis. Google Alphabet, however you want to call it, remains enormously profitable in its advertising, its cloud, YouTube, obviously the Android business, they all remain extremely valuable. But free cash flow measures what remains after operating expenses and capital expenditures. And in the last quarter, Alphabet spending consumed more cash than the company generated because of AI and this religious drive to win the race. When capital expenditures reached 45 billion, which is approximately twice the amount from a year earlier, that's saying something. And then Alphabet raised its 2026 capital spending budget to between 195 billion and 205 billion, with management warning that expect spending to rise significantly again in 2027. It's getting to be a lot for a lot of uncertainty. It's a staggering estimate for a single company. And more importantly, spending twice as much doesn't necessarily mean Google is building twice as much capacity. Memory prices are rising. Power connections are scarce. Transformers, turbines, other equipment, they all have long waiting lists. Suitable land is disappearing and becoming more expensive. All the same companies are bidding for many of the same resources, driving up their costs far in excess of, uh, what was expected when the budgets were put together. And part of that increase, reflecting inflation is not a proportional increase in computing power and therefore revenue generation down the road. So Google, for example, faces two risks. Whether AI generates enough revenue and whether each new dollar of investment buys less infrastructure than the previous one. That's how the equity story becomes a credit and debt story. And then there are the accounting issues, because let's face it, when you have this much of a spending bill to finance in the debt market, there's a huge interest to, to be creative about how that debt shows up, or in many cases, doesn't show up. And it's another thing for the market to worry about. What is the true cost, the cost today as well as future cost of this AI tech build out? Oh yeah, they're experimenting with structures that are designed to prevent all this debt from appearing directly on, um, their balance sheets. And again last year, nobody cared. This year, people are finally starting to poke around and ask the right question questions. The Texas project, financed by Blackrock provides a clear example of this. BlackRock entities hold approximately 80% of the project's equity and finance part of that interest. By selling 12 and a half billion of bonds that we just talked about, meta retains roughly 20% of the exposure and guarantees lease payments connected to the project. Now, technically, much of the debt belongs to the project structure rather than Meta itself. Economically, the data center still depends heavily on meta. Now, this doesn't necessarily mean something's wrong or unethical here. Project financing is common and can allocate risks efficiently. But as structures become more complicated, investors are finding it harder to determine where the ultimate exposure sits. And in many cases, it is by design. Banks financing these projects can use credit derivatives to reduce or hedge their exposure, which means rising CDS activity may reveal anxiety that isn't obvious from corporate balance sheets alone. That's what we're getting here. That's the big signal from the credit default swap market. The market is no longer financing only a handful of technology companies. It is financing those, plus special purpose vehicles, infrastructure partnerships, data center developers, power projects, chip suppliers. All of it tied to the same underlying assumptions. If AI revenue grows as projected, all of these structures will pay off and they'll work really well. If adoption disappoints, if pricing power weakens, or if the technology becomes much more efficient, the industry could discover and finance too much capacity at precisely the wrong price. The debt doesn't eliminate that risk, it redistributes it. And that's where everybody is interested right now. If this doesn't work, who actually would be holding the bag? And of course, there's already indications that many investors, even the stock market, are reconsidering a lot of their previous assumptions. The tech stocks, especially semiconductors, they have not had a good month here in July. The MSCI World Semiconductor Index fell 16% in July, putting it on course for its worst month since 2022, though we should point out it remains substantially higher for the year. The the Nasdaq 100 moved toward correction territory as semiconductor shares. They fell across South Korea, Japan, Taiwan and of course, the US SK Hynix reported a six fold increase in quarterly profit just recently and margins that were above 80% and promptly saw its shares fall 19%. Why would investors punish results like these? Because the expectations are even higher. SK Hynix, good example, announced at least 31 billion in capital spending for the year, approximately 50% above its previous level. What looked like evidence of extraordinary demand could also be interpreted as another supplier dramatically expanding capacity near the top of a cycle. When it's most expensive. At the same time, reported Chinese progress in advanced lithography increases fears of future competition and in excess global chip making capacity. This is the dot com resemblance that actually matters. At the peak of a capital cycle, record profits encourage record investment. Record investment increases supply. Increased supply eventually pressures prices and returns. The best financial results often arrive just before investors begin asking whether those results can possibly be repeated and sustained. Everyone is watching AI stocks for the bubble. But stocks? They tell us what investors hope companies might become. Debt tells us what investors believe they can safely finance. Right now, Oracle's protection cost is at a multi year high. SpaceX swaps have surged. Amazon encountered unusually weak bond demand. A BlackRock backed data center financing required a 7.53% yield. And Google's unprecedented negative cash flow demonstrates just how quickly AI spending can overwhelm even the world's most powerful cash generating businesses. None of these signals individually proves there's some impending crisis. Together though what they do show increasingly clearly is that the market has started recalculating risk, especially the debt market. The AI boom is no longer a contest about who can spend the most and grow the fastest. It's fast becoming instead about who can deliver on the promises and who will be left holding the bag when inevitably some of those promises, maybe a lot of those promises, don't live up to the projections. The economy moves through four distinct regimes and each one rewards a completely different portfolio strategy. So join us for a free webinar where we're going to go over how to identify which regime we're in now, which one might be coming next, and how to position your portfolio for growth, inflation, recession, maybe even deflation. Register now for the four economic regimes and how to position your portfolio for each one.