Which Big AI Company Goes Bankrupt First? — Transcript
Full transcript
- 0:00It's Nebius, okay? Neo Cloud Nebius will
- 0:03be the first AI stock to go bankrupt.
- 0:06Nah, nah, it's rage bait. I'm joking. Or
- 0:08am I?
- 0:09From Anthropic to Coreweave to Oracle to
- 0:12Manebius, we're going to look at the
- 0:14first AI stock that's likely to go
- 0:17bankrupt.
- 0:18is that so important? Because
- 0:20bankruptcies are the canary in the coal
- 0:23mine for any sort of financial crisis.
- 0:26Most you young bucks out there have no
- 0:28idea what real blood on the street looks
- 0:30like, but don't you worry. I'll talk
- 0:32about last week's blood bath in a
- 0:35second, but I have a very important
- 0:37point to make here. We believe that it's
- 0:39highly likely in the case of the AI
- 0:42bubble bursting that it's accompanied by
- 0:45large bankruptcies or even driven by
- 0:48large bankruptcies. Why? Because of
- 0:51something called circular financing, and
- 0:54it's getting worse and worse as the days
- 0:56go on. Nvidia is now being called the
- 0:59bank of AI for good reason. Why it's
- 1:02going to discuss that a little bit
- 1:03later, but let's get on to talking about
- 1:06the elephant in the room. Something
- 1:08terrible happened last week, and I can't
- 1:10stop thinking about it.
- 1:12Sydney Sweeney was filmed crying in
- 1:14character during the filming of the
- 1:16House Maid's Secret last week in New
- 1:18York. I hope that poor woman's okay.
- 1:21Another equally traumatic event that
- 1:23happened last week was the AI stock
- 1:26blood bath.
- 1:28Blood bath.
- 1:29As always, we kept our Analyzed Premium
- 1:32subscribers up to date on the situation
- 1:34and our Analyzed market open, but for
- 1:37the rest of you, take a look at this
- 1:39chart here. This is the Nasdaq 100 over
- 1:42the past five years, and it's not really
- 1:45that bad when you look at it in the
- 1:47context of a longer time frame, is it?
- 1:50Now yes, I know the crisis mainly
- 1:52affected high beta AI stocks. So take a
- 1:56look at this even better slide. These
- 1:58are the four AI stocks that hit the
- 2:00hardest during last week's blood bath.
- 2:03Sandisk, AMD, Micron, and Monebius. You
- 2:08know, this always gets me. Nobody bats
- 2:10an eyelid when Sandisk soars over 2,000%
- 2:14in less than two years. But heaven
- 2:17forbid it should drop 20% giving it an
- 2:20year-to-date return of only 200%.
- 2:23We need to stop slobbering over
- 2:26short-term price changes like a bunch of
- 2:28monkeys and focus on what systemic
- 2:31changes may have happened here. Most
- 2:33recently, there are concerns that the
- 2:36Democratic mid-term wins in the United
- 2:38States are going to slow down data
- 2:41center construction because of things
- 2:43like regulation, moratoriums, and the
- 2:45like. And probably most tangible would
- 2:47be the signs that are showing access to
- 2:50easy capital for AI expansion is
- 2:53starting to tighten.
- 2:55Evidence? Well, some AI IPOs are being
- 2:58pulled back due to lack of interest from
- 3:00investors like Firmus in Australia. You
- 3:03have Microsoft pulling back on some of
- 3:05their data center investments because of
- 3:07delays. And most importantly, I think is
- 3:09that new debt is becoming more costly
- 3:12because it's more costly to insure it.
- 3:15This is something we measure with tools
- 3:17called credit default swaps. This is a
- 3:20really cool metric that any investor out
- 3:23there can use to monitor the situation.
- 3:26So here we've listed out some of the
- 3:28hyperscalers and I'll just show you
- 3:30really quickly how to read this. So
- 3:32let's take Microsoft as an example here.
- 3:34It costs $5.60
- 3:38per year to insure Microsoft against
- 3:41default over the next five years. And
- 3:44for most large AI companies out there,
- 3:47you can find these instruments. What you
- 3:49can then do is see what the cost of
- 3:52these credit default swaps has been
- 3:54doing over time. For example, how far
- 3:57are they off their all-time high, right?
- 3:59If they're close to the all-time high,
- 4:01shows you that risk is still pretty high
- 4:03or perceived as being high. And we can
- 4:06expand this out to other names and of
- 4:09course the name that's been in the news
- 4:11lately, Coreweave. So it costs $110 a
- 4:15year to protect against default for
- 4:17$1,000 in Coreweave debt. That's very
- 4:20high. And when we look at the change of
- 4:22these credit default swaps over the past
- 4:24month, you can see they've all been
- 4:26rapidly appreciating. In other words,
- 4:28this debt is getting riskier as time
- 4:30goes on. Nvidia, look at that
- 4:33appreciation in the risk for Nvidia's
- 4:35debt. What's going on there? Well, why
- 4:37it's going to talk a bit more about that
- 4:39in a second. But the other thing that we
- 4:41can do with the credit default swap
- 4:43that's quite interesting is that you can
- 4:45calculate an implied probability of
- 4:48default. In the case of Coreweave,
- 4:49that's about a 60% likelihood they
- 4:52default over the next five years. That's
- 4:55crazy, right? And when you consider the
- 4:57likelihood of default, there can be
- 5:00problems that are not so obvious to
- 5:02investors. It's not just a matter of
- 5:03looking at their debt and then their
- 5:06cash flows to see if they can make their
- 5:08interest payments. You also need to
- 5:09consider things like contractual
- 5:11commitments, covenants for their debt
- 5:14that may be quite obscure, off-balance
- 5:16sheet debt that you can't even see. And
- 5:19it's the job of credit ratings agencies
- 5:22to look through all this stuff. For a
- 5:24good number of AI stocks out there, we
- 5:27can look to credit ratings agencies for
- 5:30signs of trouble. We can see how they
- 5:32change their outlook over time. Now, I
- 5:34know what a lot of you are thinking
- 5:36right now. Too many Sydney Sweeney jokes
- 5:38lately. How about not enough Sydney
- 5:41Sweeney jokes?
- 5:43>> How about that? How about that? How
- 5:44about that?
- 5:45>> Now, I know what a lot of you are
- 5:46thinking. Credit ratings agencies
- 5:48weren't of any use during the 2008
- 5:51financial crisis. And you're exactly
- 5:53right. A sufficiently sized cataclysmic
- 5:57crisis doesn't give investors time to
- 6:00respond. That's the entire point of it.
- 6:02Because of the circular nature of the AI
- 6:05economy, a crash is likely to be sudden
- 6:07and powerful. Neo clouds are going to be
- 6:10the first to fail because they rely on
- 6:13debt for their lofty valuations. You see
- 6:16that happening right now. This simply
- 6:18captures current sentiment because it
- 6:20becomes more difficult for them to raise
- 6:22money. Those growth stories start to dry
- 6:24up.
- 6:25It comes down to this. The core business
- 6:28for these neo clouds is not generating
- 6:30positive operating cash flows or
- 6:32earnings sufficient to service their
- 6:35debt. The debt servicing is simply
- 6:37coming from customer advances or new
- 6:40financing or even selling equity.
- 6:42Whenever we bring this up that neo
- 6:44clouds are likely to be the first to
- 6:47fail, nobody disputes that. They simply
- 6:50focus on reasons why the AI trade's
- 6:52going to keep going. This is the compute
- 6:54demand story. Fair enough, but for neo
- 6:57clouds in particular, when they can't
- 6:59access capital, that growth story dries
- 7:02up. Hyperscalers on the other hand, they
- 7:04have very strong cash flows that they
- 7:07can fund build outs with. They don't
- 7:09need to take on debt even though some
- 7:11are. And hyperscalers are actually major
- 7:14customers for some neo clouds and that's
- 7:17a real problem, right? You never want to
- 7:19have your competitor as your customer.
- 7:22We talked about this in an earlier piece
- 7:24this year which looked at problems in
- 7:26the neo cloud sector. Now since that
- 7:28piece, neo clouds have lost half their
- 7:31value except for Nebius which has a
- 7:33stronger balance sheet and is more
- 7:35resilient to turmoil. The first AI
- 7:38company to go bankrupt is likely to be a
- 7:41neo cloud and it's likely to be
- 7:43CoreWeave because they're the most
- 7:45likely to struggle when it comes to
- 7:47refinancing their debt or paying their
- 7:49interest obligations. Of the neo clouds,
- 7:52Nebius is probably the best positioned
- 7:55at the moment because of their strong
- 7:56balance sheet. So ask yourself this, why
- 7:59did this recent bloodbath have you so
- 8:01worried? Is it because you're
- 8:03overexposed to the AI trade? Well, yes,
- 8:06because everyone is.
- 8:08Simply holding the S&P 500 gives you
- 8:10anywhere from 35 to 45% exposure to this
- 8:14tech sector and large tech companies,
- 8:17which aren't all classified under the
- 8:19GICS technology sector. So let's talk a
- 8:22little bit more about Nvidia, which is
- 8:24presently our largest position that
- 8:27we've been holding for over a decade
- 8:28now.
- 8:29>> Putting aside Jensen Huang's recent
- 8:31scandal, today I want to look at some
- 8:32concerns raised by Hunter Brook, which
- 8:34is an investigative journalism company.
- 8:36They came up with this really cool piece
- 8:37called the Comprehensive Guide to the
- 8:39Bank of Nvidia, which made some really
- 8:40cool points. Essentially, they're saying
- 8:42that Nvidia has evolved from just a
- 8:44chip-making company to a company that
- 8:46supports the entire AI industry. There
- 8:49have been these concerns going around
- 8:50about circular financing, right? You've
- 8:51seen the giant bubble chart where Nvidia
- 8:54supplies a customer, they have an
- 8:56investment in that customer. Hunter
- 8:57Brook has an even more complex and
- 9:00comprehensive chart here that shows
- 9:02Nvidia with relationships to the
- 9:03hyperscalers, Microsoft, Google, Amazon
- 9:06Web Services, to the LLM providers,
- 9:08Meta, Anthropic, OpenAI, to the likes of
- 9:11Tesla and SpaceX, to the likes of AI
- 9:13startups, Neo Clouds, Nebius, CoreWeave.
- 9:15All of these are interconnected, and
- 9:17they're saying that it's not circular
- 9:18financing, it's reflexive. So the
- 9:21reflexive property in math, going back
- 9:23to our high school math days, means that
- 9:26one thing equals itself, meaning that
- 9:29the value of Nvidia is dependent on the
- 9:31value of its customers, of its
- 9:32investments, of its suppliers. So
- 9:34lately, all of the value of these
- 9:37investments have been growing together.
- 9:39However, when the opposite happens, when
- 9:42demand for AI compute begins to
- 9:44underwhelm supply of AI compute, right?
- 9:46Because we're in a massive build out
- 9:48right now, the opposite is going to
- 9:49happen and you're going to see the
- 9:51reflexive financing in action. Let me
- 9:53give you an example. Nvidia has an
- 9:55investment in OpenAI. OpenAI is a
- 9:58customer of Nvidia.
- 9:59Nvidia has an investment in CoreWeave.
- 10:01CoreWeave is a customer of Nvidia.
- 10:03OpenAI has an investment in CoreWeave.
- 10:05There's this loop, and Hunter Brook
- 10:07calls it orbital financing as opposed to
- 10:09circular financing, which implies some
- 10:11motion throughout this orbit. As soon as
- 10:13one component of this orbit, of this
- 10:15reflexive financing loop starts to
- 10:18struggle, it impacts the rest of the
- 10:20group.
- 10:20>> So far, we've only talked about public
- 10:23companies going bankrupt, but honestly,
- 10:25a overly hyped private company going
- 10:27bankrupt can be a catalyst that pops the
- 10:30AI bubble as well. So I think right
- 10:32behind Neo Clouds being the most likely
- 10:35to go bankrupt first, you probably have
- 10:38startups that have overextended
- 10:40themselves. And a good reason for that
- 10:42is simply just look at how the venture
- 10:46capital community is just throwing money
- 10:48at AI hand over fist. No business model,
- 10:51no customers, no website, no problem.
- 10:54Here's a billion dollars. Now, one other
- 10:56thing I recently saw that's quite
- 10:58interesting is Polymarket, so that's a
- 11:01prediction market, says there's a 27%
- 11:04chance that the AI bubble bursts before
- 11:07June of 2027. Well, immediately you ask,
- 11:10what does that mean? Well, it's quite
- 11:13interesting. So they say that three at
- 11:16least of these events need to happen in
- 11:19a 90-day window for that bet to resolve
- 11:22to AI bubble has burst.
- 11:25They are Nvidia down 50% from an
- 11:27all-time high, Sox, the semiconductor
- 11:30ETF down 40% from an all-time high,
- 11:34Open AI or Anthropic declaring
- 11:36bankruptcy,
- 11:38Open AI being acquired,
- 11:40H100 GPU rental prices falling to a
- 11:44dollar or lower for five consecutive
- 11:46days,
- 11:47or any major AI hardware supplier seeing
- 11:51their price, their stock price collapse
- 11:5350% or more. And you can see the list of
- 11:56names here. Again, three or more of
- 11:58these events need to happen during a
- 12:0090-day window. So by all measures, the
- 12:03AI bubble is not bursting at all. This
- 12:06is just normal price volatility. One or
- 12:09more bankruptcies is major bankruptcies
- 12:12is likely to accompany any real AI
- 12:15crisis. And the first company to go
- 12:17bankrupt is likely to be CoreWeave based
- 12:19on what everything points to today. Just
- 12:22remember, no matter how strong demand is
- 12:25for AI compute, there are always unknown
- 12:28unknowns that can happen, right? Think
- 12:30like a COVID-style moratorium on AI
- 12:33because that whole cybersecurity
- 12:35situation gets out of hand. That's why
- 12:37you should always be aware of how much
- 12:39tech risk you have. And for those of you
- 12:42thinking about doubling down on Neo
- 12:44Clouds, you ought to give this next
- 12:46piece a watch which talks about some of
- 12:48the risks in that space. Thanks so much
- 12:50for taking the time to watch this today.
About this transcript
This page contains the full transcript of Which Big AI Company Goes Bankrupt First? by Nanalyze, generated from the public captions YouTube serves with the video. The transcript has 2,115 words across 326 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.