Economic War Room

Speaker warns of circular AI investment loops, off-balance sheet risks, and NVIDIA's central role in capital spending.

24 min Listen to the original
AI Capital SpendingNVIDIAOpenAIOff-Balance Sheet DebtGeopolitical RiskMarket BubblesPrivate CreditSPVs

Summary

The speaker compares current AI capital spending to historical bubbles like railroads and the internet, noting that while transformative, it poses significant investment risks. He highlights that major tech firms plan over $700 billion in AI capex this year, citing a BIS study that 55% of AI investment flows circularly between firms. NVIDIA is described as a central financier and guarantor, holding vast equity investments and supply commitments. The speaker also points to OpenAI's massive spending gap and off-balance sheet debt structures via SPVs as key risks. Geopolitical vulnerabilities in chip manufacturing and rapid asset depreciation are flagged as additional concerns. Finally, five warning lights are identified, advising portfolio rebalancing and cash liquidity.

Key takeaways

  1. 1
    Between 2021 and 2025, 55% of the investment flowing into AI firms came from other AI firms. And in nearly half of those AI to AI deals, measured by dollars, the investor was also a supplier or customer of the company it was investing in.

    Unidentified speaker (speaker_0) · 2:26

  2. 2
    NVIDIA reported $99 billion in equity investments and $279 billion in supply/capacity commitments (up from $119 billion one quarter prior).

    Unidentified speaker (speaker_0) · 4:23

  3. 3
    NVIDIA committed up to $105 billion in credit support tied to an OpenAI data center campus in Ohio.

    Unidentified speaker (speaker_0) · 4:43

  4. 4
    OpenAI reported ~$13 billion revenue in 2025 but plans ~$600 billion spending through 2030 (down from a prior $1.4 trillion commitment).

    Unidentified speaker (speaker_0) · 5:51

  5. 5
    Residual value guarantees in data center SPVs act like co-signed loans; if asset value drops, the tech company pays the difference, resembling off-balance sheet debt.

    Unidentified speaker (speaker_0) · 8:48

Episode sections

  1. 0:00 – 2:13

    Intro: AI Investment History and Capital Spending

    Speaker compares AI to railroads and the internet, noting that while the technology is transformative, investors often lose money. He highlights that Amazon, Microsoft, Alphabet, and Meta plan to spend over $700 billion in capital investments for AI this year, with Wall Street expecting over $1 trillion next year.

  2. 2:13 – 4:22

    The Circular AI Money Loop

    Citing a Bank for International Settlements (BIS) study, the speaker explains that 55% of AI investment flows between AI firms, often involving circular relationships where investors are also suppliers or customers. This creates a loop where money is counted multiple times as investment, revenue, and demand.

  3. 4:22 – 5:51

    NVIDIA's Role as Investor and Guarantor

    NVIDIA is described as central to the boom, holding $99 billion in equity investments and $279 billion in supply commitments. The speaker notes NVIDIA's agreements to buy back cloud capacity if partners cannot sell it, effectively acting as a customer of last resort and financier.

  4. 5:51 – 8:03

    OpenAI's Revenue vs. Spending Gap

    OpenAI reported $13 billion in 2025 revenue but plans to spend $600 billion on computing power through 2030. To justify this, OpenAI expects revenues to grow to over $280 billion by 2030, a 24% annual growth rate, which the speaker labels as a narrative risk.

  5. 8:03 – 11:16

    Off-Balance Sheet Debt and Residual Value Guarantees

    Big tech companies use special purpose vehicles (SPVs) to build data centers, keeping debt off their balance sheets while providing residual value guarantees. Meta's Hyperion campus is cited as an example, with $28 billion in guarantees backing a deal where lenders provided $27 billion.

  6. 11:16 – 14:15

    Historical Parallels and Collateral Risk

    The speaker compares current AI financing to the late 1990s telecom bubble and Enron's use of SPVs. He warns that AI chips depreciate rapidly like smartphones, posing a risk to collateral values in off-balance sheet deals if cash flows fail.

  7. 14:15 – 16:04

    Geopolitical and Supply Chain Vulnerabilities

    The speaker highlights two vulnerabilities: the market reaction to Chinese AI models (citing DeepSeek's impact on NVIDIA) and the concentration of advanced chip manufacturing in Taiwan under threat from China, creating a strategic risk for the AI supply chain.

  8. 16:04 – 24:04

    Warning Lights and Investment Strategy

    Five warning lights are identified: big tech cutting spending, rising borrowing costs, slowing NVIDIA growth, narrowing market breadth, and 10-year Treasury yields hitting 6%. The speaker advises rebalancing portfolios, maintaining cash liquidity, and considering real assets like gold.

Assets discussed

AssetKindStanceContext
AmazonAMZNequityneutralFour companies: Amazon, Microsoft, Alphabet, and Meta, plan to spend more than $700 billion this year on capital investing, most of it for AI.
MicrosoftMSFTequityneutralFour companies: Amazon, Microsoft, Alphabet, and Meta, plan to spend more than $700 billion this year on capital investing, most of it for AI.
AlphabetGOOGLequityneutralFour companies: Amazon, Microsoft, Alphabet, and Meta, plan to spend more than $700 billion this year on capital investing, most of it for AI.
Meta PlatformsMETAequityneutralFour companies: Amazon, Microsoft, Alphabet, and Meta, plan to spend more than $700 billion this year on capital investing, most of it for AI.
NVIDIANVDAequityneutralIt reported $279 billion in supply and capacity commitments, up from $119 billion just one quarter earlier. And it's committed credit support of up to $105 billion, tied to a data center campus. in Ohio built for OpenAI. NVIDIA used to be a chip company.
OpenAIprivate companyneutralTo make the math work, OpenAI expects revenues of more than $280 billion by 2030.
DeepSeekprivate companyneutralIn January of last year, a Chinese company called DeepSeek released an AI model that claimed was built for a fraction of the cost.
GoldcommodityneutralAnd consider real diversification, including real assets like gold.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Amazon, Microsoft, Alphabet, and Meta plan to spend more than $700 billion this year on capital investing for AI.

Unidentified speaker (UNKNOWN) ·1:12

How to verify: Check Q3 2026 earnings reports or 10-K filings for Amazon (AMZN), Microsoft (MSFT), Alphabet (GOOGL), and Meta (META) for capital expenditure guidance and actuals for the fiscal year ending 2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Wall Street analysts expect total AI capital spending to top $1 trillion next year.

Unidentified speaker (speaker_0) ·1:32

How to verify: Search for consensus estimates from major investment banks (e.g., Morgan Stanley, Goldman Sachs) published between Aug 2026 and Nov 2026 regarding AI capex forecasts for 2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Between 2021 and 2025, 55% of investment flowing into AI firms came from other AI firms.

Unidentified speaker (UNKNOWN) ·2:26

How to verify: Access the Bank for International Settlements (BIS) study 'Circular Relationships Among AI Firms' published October 1, 2026, via the BIS website. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

NVIDIA reported $99 billion in equity investments.

Unidentified speaker (UNKNOWN) ·4:22

How to verify: Check NVIDIA's (NVDA) most recent quarterly 10-Q or annual 10-K filing for the 'Investments' or 'Other Assets' section. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA reported $279 billion in supply and capacity commitments, up from $119 billion one quarter earlier.

Unidentified speaker (UNKNOWN) ·4:22

How to verify: Check NVIDIA's (NVDA) quarterly filings for disclosures on 'Supply and Capacity Commitments' or 'Contractual Obligations'. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA committed credit support of up to $105 billion tied to a data center campus in Ohio built for OpenAI.

Unidentified speaker (speaker_0) ·4:43

How to verify: Check NVIDIA's (NVDA) 10-K or 10-Q filings for 'Commitments' and 'Contingencies' sections, specifically regarding OpenAI or Ohio data centers. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

OpenAI brought in about $13 billion of revenue in 2025.

Unidentified speaker (UNKNOWN) ·5:51

How to verify: Check OpenAI's financial disclosures, investor updates, or press releases for 2026 revenue figures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

OpenAI plans to spend roughly $600 billion on computing power through 2030.

Unidentified speaker (UNKNOWN) ·5:51

How to verify: Check OpenAI's public statements, investor presentations, or news reports from 2026-2026 regarding their 2026 spending commitments. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Google's cloud business reported a backlog of more than $460 billion earlier this year.

Unidentified speaker (speaker_0) ·6:31

How to verify: Check Alphabet's (GOOGL) quarterly earnings reports for Q1-Q3 2026 for Google Cloud backlog or deferred revenue disclosures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

NVIDIA lost $600 billion in market value in a single trading day after DeepSeek released an AI model.

Unidentified speaker (speaker_0) ·14:35

How to verify: Check NVIDIA's (NVDA) stock price history and market capitalization data for the date DeepSeek released its model (referenced as January of last year, i.e., Jan 2026) via financial data providers like Bloomberg or Yahoo Finance. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Do your own research

  • Verify AI capex forecasts for Amazon, Microsoft, Alphabet, and Meta in their latest 10-Q/10-K filings.
  • Review the Bank for International Settlements (BIS) study on AI investment circularity for methodology and data sources.
  • Check NVIDIA's earnings reports for details on equity investments and supply commitments.
  • Analyze OpenAI's financial disclosures or public statements regarding revenue projections and spending plans.
  • Investigate data center SPV structures and residual value guarantees for major tech firms in their footnotes.
  • Assess geopolitical risk exposure of advanced chip manufacturing in Taiwan via supply chain reports.
  • Monitor NVIDIA's growth rates and market breadth indicators for signs of slowing momentum.
  • Track 10-year Treasury yields and borrowing costs to evaluate refinancing risks for off-balance sheet debt.
Full transcriptexpand
0:00
Unidentified speaker (UNKNOWN): In
0:00
Unidentified speaker (speaker_0): the 1800s, investors were right that the railroads would change America. And many of them still went broke.
0:08
Unidentified speaker (UNKNOWN): In
0:08
Unidentified speaker (speaker_0): the late 1990s, investors were right that the internet would change the world, and telecom shareholders lost fortunes. Today, artificial intelligence is real. It works. It's going to change almost everything. But there's a question Wall Street does not like to ask out loud. Who's actually paying for it? Because when you
0:28
Unidentified speaker (speaker_0): follow the money, a surprising amount of it goes around in a circle.
0:33
Unidentified speaker (speaker_0): Welcome to the Economic War Room. I'm Kevin Freeman. Let me start by being very clear. I'm not anti AI. Our team uses it every day. It may be the most powerful tool of our lifetime. And America must win the AI race. If the Chinese Communist Party wins, it won't just own it. a technology. They will own the commanding
0:53
Unidentified speaker (speaker_0): heights of the next economy.
0:55
Unidentified speaker (UNKNOWN): But
0:56
Unidentified speaker (speaker_0): a great technology and a great investment are two very different things. History proves that over and over again. So the question for this episode is not whether AI works, it does. The question is whether the money behind it works. And the amount of money is staggering.
1:12
Unidentified speaker (UNKNOWN): And
1:12
Unidentified speaker (speaker_0): if you have a 401k, a pension or an index fund, this isn't just a story about tech investors. You're already in it. Four companies: Amazon, Microsoft, Alphabet, and Meta, plan to spend more than $700 billion this year on capital investing, most of it for AI. That's roughly $2 billion a day, every day of the year
1:32
Unidentified speaker (speaker_0): for chips, buildings, power, and cooling. And Wall Street analysts expect the total to top $1 trillion. next year. Here's something many American investors miss. Spending has grown so fast that free cash flow at several of these giants has fallen sharply.
1:48
Unidentified speaker (UNKNOWN): So
1:49
Unidentified speaker (speaker_0): more and more of the build-out is being paid for with borrowed money. And all that money has to earn a return. To earn a return, customers outside Inside the AI industry have to pay real cash for AI services. That's a lot of real cash.
2:03
Unidentified speaker (speaker_0): There's an old saying in the financial analysis: earnings are an opinion. Cash is a fact. In this episode, we're going to follow the cash.
2:13
Unidentified speaker (UNKNOWN): So
2:13
Unidentified speaker (speaker_0): where is the money coming from today? On October 1st, the Bank for International Settlements, often called the Central Bank for Central Banks, published a study titled Circular Relationships Among AI Firms.
2:26
Unidentified speaker (UNKNOWN): This
2:26
Unidentified speaker (speaker_0): isn't a short seller talking as book. This is the situation the world central bankers rely on to spot financial danger. Here's the headline. Between 2021 and 2025, 55% of the investment flowing into AI firms came from other AI firms. And in nearly half of those AI to AI deals, measured by
2:46
Unidentified speaker (speaker_0): dollars, the investor was also a supplier or customer of the company it was investing. in. Let me show you how the loop works in plain English.
2:55
Unidentified speaker (UNKNOWN): A
2:55
Unidentified speaker (speaker_0): chipmaker invests in an AI company. The AI company uses that money to rent computing power from a cloud company. The cloud company buys the chipmaker's chips.
3:05
Unidentified speaker (UNKNOWN): The
3:06
Unidentified speaker (speaker_0): chipmaker books the revenue, its stock goes up, and has more money to invest again.
3:11
Unidentified speaker (UNKNOWN): But
3:11
Unidentified speaker (speaker_0): here's a kitchen table version. Suppose I lend a neighbor a $1,000 so he can buy my used lawnmower for $1,000. I can honestly say I made a $1,000 sale. But until my neighbor pays me back from his own paycheck, my real customer is my own money. Now multiply that by hundreds
3:31
Unidentified speaker (speaker_0): of billions of dollars. Why do AI companies invest in each other this way? Partly because the inputs are so scarce. If you can't get chips, power, or computing time, you can't compete. So companies use investment dollars to lock in supply and customers at the same time. It's like a gold rush where the shovel maker lends the
3:51
Unidentified speaker (speaker_0): miners money so they can afford to buy more shovels. Every step can be perfectly legal. Every step can make business sense. The BIS says themselves that these deals can serve. real economic purposes like locking in scarce chips and power. But here's the problem the same dollar can show up as an investment, then as a
4:11
Unidentified speaker (speaker_0): revenue, then as demand, counted several times along the way. That makes it hard to answer the only question that really matters. How much of this demand comes from real paying customers outside the circle?
4:22
Unidentified speaker (UNKNOWN): Now
4:23
Unidentified speaker (speaker_0): consider NVIDIA, the company at the very center of the boom. Its last quarterly filing reported $99 billion in equity investments. Many of those are stakes in AI model makers and cloud companies which buy NVIDIA products. It reported $279 billion in supply and capacity commitments, up from $119
4:43
Unidentified speaker (speaker_0): billion just one quarter earlier. And it's committed credit support of up to $105 billion, tied to a data center campus. in Ohio built for OpenAI. NVIDIA used to be a chip company. Today it's also a venture investor, a financier, a cloud customer, and a guarantor. Its own filings show that a
5:02
Unidentified speaker (speaker_0): small number of customers account for a large share of its sales. And some of those customers are companies NVIDIA has helped Help finance. And here's one more wrinkle: NVIDIA now has agreements to buy back cloud. computing capacity from some of its AI cloud partners if those partners can't sell it to anyone else. Think about that.
5:22
Unidentified speaker (speaker_0): NVIDIA sells the chips, the partner builds the capacity. And if outside customers don't show up, NVIDIA itself becomes the customer of last resort. Now let me be careful here. I'm not saying NVIDIA's Enron. I have no evidence of fraud. NVIDIA makes extraordinary products and earns enormous. real product profits. But anyone who has managed money through a
5:42
Unidentified speaker (speaker_0): full cycle knows this rule. When a supplier is financing its own customers, you have to look much harder at the demand. Take OpenAI.
5:51
Unidentified speaker (UNKNOWN): It
5:51
Unidentified speaker (speaker_0): brought in about $13 billion of revenue in 2025. It has told investors it plans to spend roughly $600 billion on computing power through 2030. And that $600 billion commitment was a scale-back number. Earlier, the commitment was $1.4 trillion. To make the math work, OpenAI
6:11
Unidentified speaker (speaker_0): expects revenues of more than $280 billion by 2030. That's better than 24 growth in five years. Is it possible? Maybe. Is it guaranteed? No. So we have a classic tale of two narratives: the bull narrative, AI is transforming every industry. Demand outruns supply. Google's cloud business alone reported a backlog of
6:31
Unidentified speaker (speaker_0): more than $460 billion earlier this year. And the biggest buyers are cash machines. They're not pets.com. Much of that's true, and it deserves respect. AI is already writing software, answering customer calls, reading medical scans, speeding up research. Those are real productivity gains, and real businesses will pay for them. So the question
6:51
Unidentified speaker (speaker_0): isn't whether customers will pay for AI, it's whether they'll pay enough, soon enough, to justify trillions of dollars of infrastructure, much of it financed with borrowed money. The bare narrative: AI works spectacularly, but the people who finance the build-out never earn their money back. The technology succeeds, the capital structure
7:11
Unidentified speaker (speaker_0): fails. We've seen that movie before. And this isn't just a short seller's view. Ray Dalio, founder of Bridgewater, put it simply: he said, all great technology changes produce bubbles. A year ago, he estimated AI euphoria was already 80% of what we saw before 1929 and 2000.
7:30
Unidentified speaker (UNKNOWN): And
7:30
Unidentified speaker (speaker_0): the loop has a reverse gear. On the way up, NVIDIA grows, AI stocks rise, companies raise more money, they buy more chips, NVIDIA grows again, but run it backwards. backward, NVIDIA misses, stocks fall, financing dries up, orders are canceled, and NVIDIA misses again. Both narratives can be true at the same time. That's what makes this so important. All right, we're gonna have to take a break. When we
7:50
Unidentified speaker (speaker_0): come back, we'll look at where the debt is hiding and why it reminds some of us of 1999.
7:57
Unidentified speaker (UNKNOWN): So
7:57
Unidentified speaker (speaker_0): stay with us, we'll be right
7:59
Unidentified speaker (UNKNOWN): back.
8:03
Unidentified speaker (speaker_0): Welcome back to the Economic War Room. Before the break, we followed the money around in a circle.
8:08
Unidentified speaker (UNKNOWN): Now
8:08
Unidentified speaker (speaker_0): let's follow the debt, because a lot of it does doesn't show up where you would expect to find it. Here's how it often works: a big tech company doesn't build the data center itself. It helps set up a separate company, a special purpose vehicle. Private credit lenders and investors put up most of the money. The special purpose vehicle
8:28
Unidentified speaker (speaker_0): builds the data center. The tech company signs a long-term lease. The debt sits on the book of the special purpose vehicle, not on the tech company's balance sheet. And very often, the tech company promises that the asset will still be worth a minimum amount years from now. That's called a residual value guarantee. If the asset's worth
8:48
Unidentified speaker (speaker_0): less, the tech company pays the difference. Think of it like co-signing your nephew's car loan. If it doesn't show up as your debt, it costs you nothing at all until the day he stops making payments. One credit research firm compared these guarantees to selling insurance against a crash. You collect the benefit in good times.
9:08
Unidentified speaker (speaker_0): You pay in the worst times.
9:10
Unidentified speaker (UNKNOWN): A
9:10
Unidentified speaker (speaker_0): real example: Meta's Hyperion Data Center campus in Louisiana. Meta owns 20% of the venture, lenders put up about $27 billion. And Meta backed the deal with roughly $28 billion in residual value guarantees. Economically, that looks a lot like borrowing. On Meta's balance
9:30
Unidentified speaker (speaker_0): sheet, most of it isn't there. This isn't just one company. In September, the Financial Times reported that big tech has offered up roughly. $300 billion of these guarantees in just the past year. Morgan Stanley estimates that seven large tech and chip companies carry more than $3.1 trillion in
9:50
Unidentified speaker (speaker_0): off-balance sheet commitments and credit support. Then in September, Oracle, the builder for OpenAI's Stargate project, reserved the right to dis delay lease payments on its New Mexico campus.
10:02
Unidentified speaker (UNKNOWN): As my
10:02
Unidentified speaker (speaker_0): friend John Molden said, there's never just one cockroach. Why do companies do it? Because debt on the balance sheet affects credit ratings, borrowing costs, and how investors value the stock. But if you keep the debt somewhere else and the company looks stronger than it otherwise would, the BIS has warned that many of these arrangements are poorly
10:22
Unidentified speaker (speaker_0): disclosed and that the same assets may be pledged more than once. Their point in plain English debt doesn't disappear just because you can't see it. The BIS also warned that because these deals are concentrated among a very few large firms, trouble at one can spread through business ties and financial ties
10:42
Unidentified speaker (speaker_0): at the same time. And who's doing much of the lending? Private credit. Investment funds lending outside the banking system with far less transparency.
10:51
Unidentified speaker (speaker_0): Pension funds, insurance companies, which hold the retirement savings of millions of Americans, are big investors in private credit. Now you may remember another company that loves special purpose vehicles. Enron used them to hide debt and to commit fraud. Enron was named America's most innovative company by Fortune Magazine six years in a row. Wall Street
11:11
Unidentified speaker (speaker_0): loved it right up until they couldn't explain their own balance. sheet.
11:16
Unidentified speaker (UNKNOWN): I
11:16
Unidentified speaker (speaker_0): want to be precise. I'm not alleging fraud here, but when the debt sits off site and the guarantee sits on site, every investor should be paying attention. Here's what worries me most about these guarantees the collateral. An office building is still a building 20 years from now, but an AI chip? Five years from now, today's chips may be like five
11:35
Unidentified speaker (speaker_0): five-year-old year old smartphones. They still work. but nobody will pay full price. So, in a downturn, two things can fail at the same time the cash flow and the value of the collateral. That's when lenders get hurt and when guarantees turn into real bills. Michael Burry, the investor who made famous The Big Short, has raised a
11:55
Unidentified speaker (speaker_0): related accounting question. Companies spend the cash on chips today, but they spread the expense over several years. If the chips wear out economically faster than the accounting assumes, today's earnings look better than the real economics. The cash goes out now, the bill shows up later, and every new generation of chips makes the last one a little less
12:15
Unidentified speaker (speaker_0): valuable.
12:16
Unidentified speaker (UNKNOWN): If
12:16
Unidentified speaker (speaker_0): that sounds technical, remember WorldCom. Its fraud centered on booking ordinary expenses as long-term investments, which made profits look far better than they were. It ended in what was then the largest bankruptcy in American history. And again, I'm not alleging anything like that today. But WorldCom proved one thing: how a company accounts
12:36
Unidentified speaker (speaker_0): for what it spends can matter as much as what it sells. We've seen a version of this before. In the late 1990s, equipment makers like Lucent and Nortel lent money to their own customers so those customers could buy their equipment. Telecom companies borrowed heavily. and laid fiber across the country, and the bulls were right about the technology. Internet
12:56
Unidentified speaker (speaker_0): traffic really did explode, but too much was built too fast with too much borrowed money. When Credit Tightened, the supplier, the customer, the lender, and the shareholder all discovered they'd made the same bet. Loosen stock, which traded around 80 at its peak, eventually fell below a dollar. Much of that fiber set dark and unused for years. It was later bought
13:15
Unidentified speaker (speaker_0): for pennies on the dollar, and it became the backbone for the next generation of internet companies. The railroads followed the same script. Overbuilt railroad debt helped set off the panic of 1873. Companies failed, their tracks didn't disappear. They were bought cheap and used for a century. That's the pattern I want you to remember. AI wins, AI
13:35
Unidentified speaker (speaker_0): users win, society wins, and the first generation. of investors who financed it may still lose. And remember the dot-com bust. The Nasdaq fell nearly 80% from its 2000 peak. Yet Amazon survived and became one of the greatest companies in history. The lesson isn't to avoid the technology. The lesson is that the price you pay and the
13:55
Unidentified speaker (speaker_0): way the deal is financed matter just as much as the idea. A BIS working paper even modeled the AI race and found that company. competition alone could push investment roughly 50% above the efficient level under its conservative assumptions. Ray Dalio described the same trap. Companies either spend heavily to hold market share or spend less and risk
14:15
Unidentified speaker (speaker_0): losing it. Nobody wants to be left behind, so everyone overbuilds together. Now here's where the this becomes an economic war-room war room issue. In January of last year, a Chinese company called DeepSeek released an AI model that claimed was built for a fraction of the cost. In a single trading day, NVIDIA lost $600 billion in market
14:35
Unidentified speaker (speaker_0): value, at the time, the largest one day loss for any company in history. Whether their claim was true or not didn't matter. The market reacted to that possibility. Think about the lesson an adversary doesn't have to beat us in AI, it only has to make our financing bet look foolish. The Chinese Communist Party, the Red Horse, and my new book, they understand
14:55
Unidentified speaker (speaker_0): this. And there's a second vulnerability most investors never think about. Nearly all the world's most advanced AI chips, including Nvidia's, are manufactured by one company, TC T SMC, in Taiwan. Taiwan's a hundred miles off the coast of China. Beijing calls it a breakaway province and has never ruled out taking up by force.
15:15
Unidentified speaker (speaker_0): Some of that production is starting to move to Arizona, but today the heart is still in Taiwan. So trillions of dollars of AI spending depend on Taiwan, and Taiwan is under threat.
15:33
Unidentified speaker (speaker_0): Of stocks and a web of hidden guarantees is a target-rich environment.
15:38
Unidentified speaker (UNKNOWN): So
15:38
Unidentified speaker (speaker_0): the real risk isn't only an AI bust.
15:42
Unidentified speaker (UNKNOWN): It's
15:42
Unidentified speaker (speaker_0): an AI bust that someone else knows how to trigger. This is the real problem that we're facing. But with every problem, we come up with solutions. When we come back, the warning light's flashing right now, and your battle plan, how do you address it? So stay with
15:58
Unidentified speaker (UNKNOWN): us.
16:04
Unidentified speaker (speaker_0): Welcome back to the Economic War Room. We've followed the circular money, the hidden debt, and the history. So now the question everyone asks me, when? The honest answer is that nobody knows. Booms can run much longer than anyone expects, but there are warning lights, and some of them are already on. The first
16:23
Unidentified speaker (speaker_0): warning light is interest rates. On September 16, 16th, the Federal Reserve raised rates for the first time since 2023. And the 10-year Treasury yield recently climbed above 5.3%, its highest level since 2002. Why does that matter for AI? When a safe treasury pays more than 5%, every
16:43
Unidentified speaker (speaker_0): risky project has to clear a much higher bar. A data center that made sense with cheap money may not make sense. at 8% or 10%. So far, higher rates have hurt almost everything except AI. Money's been flowing out of small companies, real estate, utilities, and into a handful of AI
17:03
Unidentified speaker (speaker_0): giants. That creates the illusion that rates don't matter, but they do. The pain has simply shown up somewhere else first. Ray Dalio is a simple way of explaining how bubbles in. end. They pop when people have to turn their wealth into money, selling stocks and other assets to raise cash, pay debts, or meet tiger
17:23
Unidentified speaker (speaker_0): credit. Rising rates are exactly what forces that conversion. Borrowing gets expensive, lenders get pickier, paper wealth has to be turned into real dollars. Earlier this year, Dalio expected an easy Fed to keep the bubble inflating. Instead, the Fed raised rates. That's a condition worth watching closely. And here's a
17:43
Unidentified speaker (speaker_0): twist to watch. If the AI trade cracks and markets fall hard, the Fed could face pressure to cut rates even with inflation still above its target. That could rescue stock prices in the short run, but weaken the dollar in the long run. It's one more reason sound money matters. And speaking of sound money, we
18:03
Unidentified speaker (speaker_0): haven't just been watching from the sidelines. Our team, building on my book Pirate Money, has been leading the movement to restore gold and silver as constitutional money in the states. Texas is one of six states now implementing transactional gold based on our work. The system should be ready by next May. You can follow that work at transactionalgold.com.
18:23
Unidentified speaker (UNKNOWN): Now
18:23
Unidentified speaker (speaker_0): the second warning light is what traders call market breadth. The SP 500 has been setting records, but at the start of October, only about four in ten of its stocks were trading above their 200-day moving average. That's a common measure of a stock's long-term trend. About three out of four S P 500 stocks fell in
18:43
Unidentified speaker (speaker_0): September.
18:44
Unidentified speaker (UNKNOWN): The
18:44
Unidentified speaker (speaker_0): index is up, most stocks are not. A handful of giants are carrying the market. The top seven stocks make up roughly a third of the SP 500. So if you own an index fund, you may own a much bigger AI bet than you think. Here are the five warning lights I'm watching. One, big tech cuts its spending plans. Two, borrowing
19:04
Unidentified speaker (speaker_0): costs jump for AI cloud and data center debt. Three, Nvidia's order growth slows. Four, the market keeps narrowing as the index sets records. Five, the 10-year Treasury pushes towards 6%. If one light comes on, be cautious. If several come on together, that's a red alert. And on
19:24
Unidentified speaker (speaker_0): October 14, light number two flickered on. Coreweave, a heavily borrowed AI cloud company, fell 15% in a single day. Oracle fell nearly 9%, and yields on both companies' bonds hit record highs. If it doesn't wind, I'd expect it to happen in stages. First, AI stocks
19:44
Unidentified speaker (speaker_0): reprice, and because they dominate the market, the whole market feels it. Second, the spending slows. Orders fall for chips, power equipment, construction, and utilities. That hits jobs and growth well beyond Silicon Valley. And the first to feel it probably won't be the Giants. It'll be the smaller, heavily borrowed AI cloud companies. the ones
20:03
Unidentified speaker (speaker_0): financing chips at high interest rates and counting on customers who may be slow to arrive.
20:09
Unidentified speaker (speaker_0): Third, the credit test. Special purpose vehicles need refinancing. Collaterals mark down, guarantees come due. That's when a stock correction becomes a credit problem. Now, the good news: this is not 2008. Our largest banks are far better capitalized, and the biggest tech companies are enormously. profitable. A correction is not the same
20:29
Unidentified speaker (speaker_0): thing as a collapse, but the risk has moved to places that are harder to see: private credit, off-balance sheet guarantees, and your retirement accounts concentration in a few stocks. And there's a cost you may already feel. All of those data centers need enormous amounts of energy. We covered communities pushing back on data centers in our episode
20:49
Unidentified speaker (speaker_0): titled The Town That Said. No.
20:51
Unidentified speaker (UNKNOWN): If
20:52
Unidentified speaker (speaker_0): the build-out is overdone, ratepayers can be left holding part of the bill for power plants and transmission lines built for demand that never shows up. So what do we do? As usual, we need three battle plans: one for the nation, one for your business, and one for your family. First, the battle plan for the nation is transparency. Investors and regulators deserve
21:11
Unidentified speaker (speaker_0): clear disclosure of guarantees, special purpose. vehicles, and circular deals. Even the BIS is asking these questions. We must win the AI race, but we can't build our technology leadership on a financial house of cards. That means building the foundations here at home, chips, power, and a grid we control, and financing them with honest
21:31
Unidentified speaker (speaker_0): money instead of hidden guarantees. Run it through liberty, security, and values. Liberty means markets work when investors can. see the risk. Security means our AI future cannot be one bad quarter away from a credit crisis. And values means honest books. A balance sheet should tell the truth about what a company owes.
21:51
Unidentified speaker (speaker_0): Second, the battle plan for your business. If your company invests in AI, demand a real return. Use AI to make your business better, not to chase a headline. And if you sell to AI companies, know who's really paying you. Be careful about customers paying you with money you lent them. Third, the battle plan for your family,
22:11
Unidentified speaker (speaker_0): know what you own. Look at how concentrated your index funds and retirement accounts really are. Here's why that matters. If AI stocks grew from 10% of your portfolio to 30% over the last few years, that wasn't a decision you made. The market made it for you. Rebalancing is how you take that decision back. And don't confuse a great technology with a
22:31
Unidentified speaker (speaker_0): great investment. Don't chase what's already run. Be careful with debt late in a boom, especially debt with a rate that can rise. Keep an emergency cushion in cash. In a credit squeeze, the people who have options are the people who have liquidity. The goal isn't to predict the top, it's to make sure that whatever happens to the AI trade, it can't
22:50
Unidentified speaker (speaker_0): wreck what you're building for your family. A good man leaves an inheritance to his children's children. And consider real diversification, including real assets like gold. Dalio's been making that same case. This is not advice. I don't know your situation, but it's something you must consider. Jesus told the story about two builders, one built on sand,
23:10
Unidentified speaker (speaker_0): another on rock. The storms came to both houses, only one was still standing. AI can be built on rock. Real customers, real profit, real power, and real balance sheets. So to help you prepare, get your free economic battle plan at economicwarroom.com. And debt is the first of the six trials I write about in the Four Horsemen of the American
23:30
Unidentified speaker (speaker_0): Apocalypse. You can find it at fourhorsemenbook.com. Horsemanbook.com. That's the number four horseman book horsemenbook.com. A friend of mine recently sent me a photo of the four horsemen with a three-word question. Fifth horseman AI? That's what we'll cover in the next episode. You'll see exactly how it's operating.
23:49
Unidentified speaker (speaker_0): Artificial intelligence may change the world, but money still has rules. And when the money goes in a circle, someone eventually has to pay the bill. Remember what we see as a marketplace, our enemies view as a battle space. I'm Kevin Freeman. This is the Economic War Room.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from Economic War Room

All insights