Is The AI Bubble Popping? | Jack and Max on Data Center Debt, Fragile Markets, and Insurance Companies

Episode Summary

Executive Summary: The episode argues that the “AI bubble” may be less about NVIDIA’s near-term earnings and more about the financing of massive AI infrastructure buildouts. The hosts examine rising debt issuance, widening credit spreads, and creative financing across Big Tech and neo-clouds, then contrast that with sector rotation into insurance and healthcare. They also discuss Fed-rate repricing, housing squeezes, and a major regulatory shock in Chinese fintechs.

Main Topics: AI bubble, earnings, and the credit market (Priority: 5/5): The hosts debate whether the AI trade is truly cracking or simply rotating. They argue NVIDIA’s earnings were strong, but the real risk is whether AI capex remains financially sustainable as companies fund more data centers with debt and complex financing structures. Big Tech capex and debt financing (Priority: 5/5): Discussion centers on Meta, Google, Microsoft, Oracle, and related AI spenders increasingly relying on bond issuance and joint ventures to finance data center expansion, moving beyond cash-flow funding into heavier leverage and more creative structures. Credit stress and Oracle CDS (Priority: 5/5): Oracle becomes a focal point because its CDS spreads widened and its RPO growth was seen as heavily tied to OpenAI, signaling that credit markets may be the first real constraint on AI investment, even if equity markets remain optimistic. Factor investing vs fundamentals (Priority: 4/5): The hosts repeatedly stress that short-term stock moves are driven more by factor exposure, rates, volatility, and market regime than by company fundamentals, using NVIDIA, housing, and insurance as examples. Insurance as a relative-safe haven (Priority: 4/5): Insurance and healthcare are highlighted as recent outperformers, especially due to quality/low-volatility/risk-off positioning earlier in the year. The hosts discuss Kinsale and Palomar as specialized underwriting plays with strong combined ratios but notable risks. Fed repricing and housing (Priority: 4/5): A jobs report and dovish comments from the New York Fed shifted December cut odds sharply, causing a rapid rally in housing stocks. The segment reinforces how rate expectations can dominate fundamentals in the short run. Chinese fintech regulatory shock (Priority: 4/5): The hosts say a new Chinese regulatory change limiting loans with IRRs above 24% materially damaged the bull case for Chinese fintech stocks, leading to large exits and even shorting after the facts changed.

Key Arguments: NVIDIA can continue to beat earnings, but that does not resolve the real issue: whether the AI capex boom is financially sustainable over years. The market’s concern has shifted from operating performance to how data center spending is being financed, especially via debt markets. Big Tech is no longer funding AI expansion mostly from cash flow; bond issuance has surged above historical levels. Oracle is a warning signal because its CDS widening suggests the credit market is beginning to price in greater risk, even if much of the move may reflect hedging. Creative financing structures like Meta’s joint ventures with alternative credit managers may hide leverage but still increase system-wide risk. Short-term stock moves are often dominated by factor exposure, not fundamentals; the transcript uses NVIDIA, housing, and insurance to illustrate this. Insurance stocks can be attractive because they collect float and can invest it, but performance in the sector is also heavily influenced by macro factors like rates and risk-off positioning. Kinsale and Palomar are examples of highly profitable specialty insurers, but their growth depends on underwriting discipline, market structure, and potential legal/claims risks. The rise in December Fed-cut odds shows how quickly macro repricing can change equity leadership, especially in rate-sensitive sectors like housing. Chinese fintech stocks were re-rated downward because a regulatory rule change undermined their economics, making it a case where fundamentals and policy clearly matter.

Data Points: SP 500 reversal frequency: Only 4 times in the last 30 years - Used to emphasize the rarity of the intraday reversal from up about 1.5% to down about 1.5% after NVIDIA earnings. NVIDIA-linked market reaction: AI Bulls were jubilant after the beat, but the market still reversed sharply - Illustrates that strong earnings did not stop the broader selloff in AI names and risk assets. Big Tech AI debt issuance: Over $100 billion in 2025 - The hosts say major AI-capex companies are issuing far more bonds this year than in 2020-2024. Previous annual issuance: Less than $50 billion in 2020 and 2022; even less in 2023-2024 - Contrasted with the 2025 surge to show how quickly leverage has risen. Oracle RPOs: Hundreds of billions of dollars - Referenced as the source of earlier enthusiasm about Oracle’s AI-related backlog. PIMCO trade gain: About $2 billion - Reported profit on a Meta-related bond trade that compressed after issuance. Potential borrowing capacity by 2029: $6 trillion - Back-of-the-envelope estimate cited from SemiAnalysis for major AI customers at prevailing rates. Meta/Hyperion bond/JV: $26-27 billion - A joint-venture bond deal involving Meta and Blue Owl/Bignet was cited as an example of creative financing. Oracle CDS spread: Widening / blowing out - Used as evidence that credit markets are more concerned about Oracle’s leverage and AI exposure. SP 500 drawdown from highs: About 3-4% - Despite poor sentiment, the broader index was only modestly below all-time highs. Eli Lilly performance: Up about 20% over the past month - Mentioned as a healthcare outperformer that helped offset weakness elsewhere. Eli Lilly market cap: Over $1 trillion - Shown as a major contributor to healthcare sector strength. Kinsale loss ratio: Around 60% or lower on average - Used to argue that the insurer has unusually strong underwriting discipline. Kinsale combined ratio: About 75% to mid-80s - Evidence of very profitable specialty underwriting. Kinsale expense ratio: Roughly 20-21% - Part of the combined ratio calculation discussed for Kinsale. Progressive underwriting benchmark: Seen as a very good underwriter in the 90s - Compared with Kinsale’s stronger combined ratios. Palomar catastrophe loss ratio: Near 0% in several periods; examples include -1%, 2%, 10% - Used to show how favorable the absence of California earthquake losses has been. Palomar commercial earthquake premiums: Down 20% - Indicates slowing growth in its core earthquake segment as competitors and buyers recognize profitability. December Fed cut odds: 70%+ for a 25 bps cut - Repriced sharply after Fed comments and jobs data. Earlier December no-cut odds: 60.9% no cut versus 1.7% a month earlier - Shows how quickly market expectations shifted over the month. ITB housing ETF move: Up 5.4% in one day - Rally followed the repricing toward a December Fed cut. Chinese fintech risk threshold: IRR above 24% discouraged - The regulatory change that undermined the economics of several Chinese fintech lenders.

Pivotal Quotes: "The concern isn't that NVIDIA isn't going to crush revenues the next quarter, the next quarter after that. Of course, they are. The question is about the long-term sustainabilities." — Jack: Core thesis that the AI debate is about financing and durability, not near-term earnings. "It could be the bouncers, aka the financial markets, the credit markets, kick them out at midnight." — Jack: Metaphor for credit markets ending the AI spending party if funding conditions tighten. "Whenever you lose money, you're wrong." — Max: On the Chinese fintech trade and the importance of adapting to changing facts.

Implications: Listeners should focus less on headline earnings and more on financing conditions, credit spreads, and factor exposure. AI, housing, and speculative assets may stay volatile if credit tightens or macro rates reprice further.