Why We’re Getting Defensive While Investors Buy the Dip

Summary

Matt Caruso and Jason Shapiro describe a market that has weakened beneath the headline indexes, prompting them to become more defensive even as investors continue buying dips. Their concerns include deteriorating credit conditions, crowded positioning, weak reactions to good news, and a possible slowdown in funding for AI and data-center investment.

Market Signals and Trading Framework

• Credit and liquidity: Matt watches the JNK high-yield bond ETF as a proxy for risk appetite and liquidity. He prefers it to TLT because government bonds can rise as investors seek safety even while risk assets weaken. He says JNK had been deteriorating since late September and points to sessions when JNK fell as TLT rose as evidence of a move toward lower risk.
• Positioning: Jason focuses on positioning, sentiment, and market action. He says Nasdaq COT data showed speculators at their longest positions for an extended period, while commercial traders were at their shortest positioning of the current AI bull-market cycle, which he dates to roughly October 2022. He stresses that COT is not a standalone signal and should be combined with entries, exits, stops, and risk management.
• Market action: Jason sees stocks falling or failing to rally despite strong news as a warning. He cites strong reported results from Samsung and Taiwan Semiconductor (TSMC) that did not lead to stronger closes, as well as failed breakouts, as signs that buyers may not be absorbing selling.
• Breadth and sector rotation: Matt says defensive consumer-staples stocks began outperforming the Nasdaq before the indexes rolled over. He also saw lagging stocks and “old economy” shares reach new highs while the Nasdaq stalled—a pattern he compares with late 2021. He cautions that an index hedge such as QQQ may not protect a portfolio of growth stocks if money rotates into large, defensive names such as Apple.
• Practical takeaway: Matt says growth-stock investors should monitor the market’s underlying character and be willing to reduce exposure or move to cash when conditions deteriorate. He reports being largely in cash, with only small positions during the weak period. He says his account is tracked trade by trade in Crusoe Insights.

AI, Credit, and Funding Risks

The speakers remain open to AI’s long-term potential but distinguish that potential from whether companies can finance the buildout and earn adequate returns.

• Matt cites Gartner as warning that some next-generation cloud and data-center companies may struggle to generate enough revenue to service their debt by 2030.
• They discuss CoreWeave as a sign of financing stress. Jason says its borrowing costs had risen to roughly 15–16%. He also cites reported sales of about $907 million in shares by its CEO during the year, $3.3 billion in sales by the three co-founders since the IPO, and more than $5.5 billion in sales by Magnetar Capital. These are discussed as reported insider and investor sales, not as proof of future performance.
• Two data-center IPOs reportedly failed:
  • AirTrunk, which Blackstone bought for $16 billion in December 2024, attempted to list at a $15 billion valuation.
  • Firmus also reportedly failed to proceed with an IPO.
• A data center requiring about one gigawatt of capacity is estimated in the discussion to cost around $30 billion to build. The speakers cite a rough baseline of $2–3 billion a year in potential revenue while emphasizing that the actual economics are highly uncertain.
• They argue that higher rates can materially alter project economics: a move from 4% to 6% raises the interest cost by 50%, not merely 2%, on the relevant borrowing. Higher energy costs and inflation could further constrain financing.
• Jason points to proposed or expected AI investment of $60 billion and argues that AI companies and governments will continue to need capital. He speculates that governments may eventually have to create more money, potentially before an election. He says renewed liquidity could initially support markets, but if money creation fails to help, conditions could be much worse.
• The discussion also notes that AI’s future returns are unusually uncertain: costs could fall, Chinese models could put downward pressure on pricing, or the investment could generate substantial returns. The speakers warn that this uncertainty can produce large stock-price swings.

Outlook and Risk Management

• Both speakers are cautious about the near term, but they do not argue that AI or the long-term bull case is necessarily over. Jason says the technology and productivity gains could continue to support the S&P 500, while warning that credit cycles can interrupt the long-run upward trend.
• Jason describes a possible prolonged, uneven correction rather than a sudden, easily recognized market top. He believes dip-buying and reluctance to exit could persist until investors capitulate.
• 2027 risk: They discuss the possibility that new share issuance from potential Anthropic, OpenAI, and SpaceX listings could exceed corporate buybacks. Jason also doubts that proposed IPOs will necessarily happen on the terms currently anticipated, citing valuation and funding pressures.
• Jason warns that momentum strategies can struggle in choppy, sideways markets: repeated breakouts and reversals can lead to a series of losses. He argues that traders should focus on stocks showing current relative strength rather than continuing to trade former winners simply because they performed well previously.
• They emphasize that traders need predefined limits and the discipline to stop or reduce activity when conditions do not suit their strategy. A trader discussed at a conference reportedly stopped trading after reaching a monthly loss limit. Another example was a trader who was down 9% in a difficult year after much stronger prior results.
• Jason says he is having a good year after being mostly flat for much of it, largely because of one trade. He describes a “dream year” as making roughly 15% from one trade and then avoiding unnecessary risk.
• They caution against judging a trading decision solely by its outcome. A sound process can still produce a loss; traders should review and improve their process rather than berating themselves for not predicting an unknowable result.
• Jason cites an example of an AI-focused hedge fund that had lost 70% and whose largest holdings were reportedly down between 25% and 51% in October. He argues that expertise in a company or technology does not by itself translate into market profits.
• At the end of the discussion, they note that the S&P 500 was still less than 2% below its all-time high, despite the deterioration they described.

Instruments, Companies, and Sectors Mentioned

• ETFs and instruments: JNK, TLT, XLP, QQQ, and a three-times-leveraged semiconductor ETF identified in the subtitles as “XL” (the exact ticker is unclear); Nasdaq COT positioning; U.S. government bonds; call options.
• Stocks and companies: CoreWeave, SanDisk, Samsung, Taiwan Semiconductor/TSMC, Micron, Apple, PepsiCo, Coca-Cola, Costco, Keurig Dr Pepper, NetApp, CrowdStrike (mentioned briefly in a correction), and Meta.
• Other companies and entities: AirTrunk, Firmus, Blackstone, Magnetar Capital, Anthropic, OpenAI, SpaceX, Google, and Gartner.
• Sectors and assets: AI, cloud computing, data centers, semiconductors, technology and growth stocks, consumer staples, utilities, energy, oil, gold, silver, and potash.

Disclosures

No formal “not financial advice” disclaimer appears in the subtitles. The speakers repeatedly acknowledge that their views and indicators are imperfect and that they can be wrong.

Presenters and Sources Mentioned

Presenters: Matt Caruso and Jason Shapiro.

Other people and sources cited: Paul Tudor Jones, Tom Basso, Larry Page, Eric Schmidt, Michael Burry, Stanley Druckenmiller, Bill O’Neil, Mark Minervini, Oliver Kell, Christian Flanders, Annie Duke, Gartner, Bloomberg, and CNBC. Matt also refers to a previous episode featuring Matt Paxter.