AI Cracks: Where Do You Hide?

Avi argues the AI trade was built on hyperscaler commitments of $600 billion to $700 billion in data center investment, with Google alone conducting an $80 billion equity raise to sustain capital expenditure. He believes that value is now baked into the stock prices of Google, Meta, and Amazon, and that cracks are appearing in those names. The more important signal going forward is where hyperscalers are directing their spending, not the price action of the hyperscalers themselves. Hyperscaler free cash flow is heading toward near zero due to AI buildout spending, reversing nearly two decades of steady uptrend.

The market is described as being in the later innings of the AI trade, though late innings historically produce the most extreme returns. Hyperscaler stocks are behaving like the NASDAQ in 1999, exhibiting white hot upside volatility, and the NASDAQ has gone essentially sideways since around May 14th to 15th despite extremely high volatility during that period. Jonah flags upside volatility in equities specifically as a signal of extreme greed and a reason to reduce exposure rather than add, distinguishing this from crypto, where higher volatility regimes tend to be momentum-driven rather than a warning sign.

A 13-year-old encountered at an event, holding a portfolio of Google, Uber, Apple, Meta, and NASDAQ-linked single-name stocks, is used as evidence that the hyperscaler trade is widely saturated at the retail level. Memory stocks, by contrast, have not yet been penetrated by unsophisticated retail investors, suggesting room to run. The anatomy of a peak bubble in AI hardware would look like every teenager buying SanDisk and similar memory names, and that moment has not arrived. SanDisk is cited as capable of rising 40% in a few days, and Micron earnings are being watched as a near-term directional signal for the broader trade. The speakers caution that buying Micron stock as a proxy for high bandwidth memory is imprecise because hyperscalers are purchasing Micron products, not Micron stock, making data center adjacent assets like natural gas utilities and optoelectronics companies a more nuanced but preferred approach.

Current positioning includes shorting Accenture as a bet that consulting firms face reduced demand from AI, buying ARKG and biotech ETF XBI on the thesis that AI benefits biotech, and identifying fiber optic cables as the third leg of the AI trade, with AOI optoelectronics cited as an example. XBI and ARKG are held in a 70 to 30 ratio as a bet on an industry-wide re-rating rather than individual stock selection. The warning on biotech is that without fundamental knowledge such as a PhD in biochemistry, investors cannot distinguish losses caused by FDA decisions or molecule-specific failures from broader sector moves, making it difficult to hold positions through drawdowns with conviction. ARKG is noted as up 8% since the prior week's episode.

On crypto, Bitcoin and Ethereum are described as effectively sidelined as long as AI and memory stocks offer higher volatility and more interesting trades. Bitcoin broke below 60 during the recording, with a cover target range of 49 to 53. The short was established around 65, based on Bitcoin forming a range from approximately February 2nd to April 6th on the weekly chart, breaking out, and then reentering that range. Solana is predicted to see a drawdown to 46, representing roughly a 50% decline, and Farcoin is predicted to go to zero. The speaker lightened up their Bitcoin position on the way down from $124,000 highs and is not rebuying despite holding a one million dollar long-term price target. Bitcoin is also cautioned as not functioning as a hedge against broader market declines, expected to fall roughly twice as fast as the NASDAQ in a selloff.

STRC is flagged as structurally flawed because paying its dividend requires Michael Saylor to sell STRC, sell MSTR, or sell Bitcoin, with each option carrying negative price consequences for the asset sold. The speakers predict Saylor will eventually pause the STRC dividend, rendering the product useless, and note it was designed with ChatGPT as evidence of poor structural thinking. MicroStrategy's broader financial structure is characterized as involving purchasing Bitcoin with more capital than is available and paying dividends to equity holders funded by bondholders. A MicroStrategy blowup is described as feeling inevitable and expected to push Bitcoin's price down to approximately FTX-era lows, though the speakers argue the subsequent regulatory and long-term value setup for Bitcoin following such an event would be the strongest in history. The recommended approach for risk-averse investors is to buy HOOD and short crypto, on the basis that HOOD benefits if crypto rises but holds up better than crypto if the complex falls.