In this week’s video, I walk through why the headline S&P, down just 1% for the month, is masking one of the most significant dispersion events in recent memory. 150 of the S&P 500 names moved at least 10% in the month, with utilities, staples, materials, energy, and industrials leading while Mag 7 and financials lagged. My turbulence model flagged this back on February 3rd, and the dominoes have only accelerated since. Credit markets are now confirming what the equity rotation has been signaling: the software unwind is not a “ghost trade.”
The leverage loan market has broken its 200-day moving average for the first time since the Fed pivot in late 2022. Option traders are piling into puts on software-exposed loan ETFs. UBS has already raised its private credit default forecast to 15%, up from 13% just three weeks ago. Meanwhile, Goldman Sachs posted its worst day relative to the S&P since the great financial crisis on Friday, and financials as a sector broke below the 200-day moving average. The contagion is spreading from software equity into credit, into private equity vehicles, and now into bank stocks.
The deeper story is about time. AI agents have compressed time horizons so dramatically that long-duration equities are repricing first, multiples compress before spreads widen. Enterprises will have a hard time adopting AI at the speed required to justify the hyperscalers’ levered capex bets, Chinese open-source models are catching up at a fraction of the cost, and memory prices have driven compute costs far higher than anyone modeled. If revenue doesn’t come in fast enough, these are levered bets, as Dario Amodei himself acknowledged. The endgame points toward Bitcoin and crypto as the infrastructure layer for an AI-native economy running at machine speed, especially if government regulation compresses hyperscaler multiples further.
Timestamps
- (00:00–05:00) Market dispersion & the webinar recap: S&P down only 1% but 150 names moved 10%+; turbulence model flagged February 3rd; credit widening accelerating; subscriber webinar covered portfolio shifts and risk-reward changes
- (05:30–09:30) Enterprise AI adoption friction: Models are good enough, but enterprise data quality, friction, and organizational speed prevent rapid adoption; Chinese open-source models 20x cheaper with comparable performance; implications for hyperscaler revenue
- (09:30–12:00) The hyperscaler levered bet: Memory costs up ~45% of compute; Dario Amodei’s bankruptcy risk comments; if revenue falls short, the capex is already committed
- (12:00–17:00) Software isn’t a ghost trade: IGV vs NDX breakdown; Salesforce at Ford’s PE; the argument for buying physical assets (silver, analog semis, chemicals, energy) over software names with uncertain 3-year viability
- (17:00–25:00) Credit contagion: Leverage loan total return index breaking down; private credit defaults rising to 15%; Apollo bad loans spreading beyond software; MFS collapse in London; private equity managers going out of business
- (25:00–35:00) Hyperscaler-to-credit linkage: Meta auditor Ernst & Young red flags on data center accounting; off-balance-sheet SPV financing connecting hyperscalers to private credit stress; financials breaking 200-day MA; Goldman’s worst relative day since GFC
- (35:00–45:00) AI agents change time: Black-Scholes model breaking down; equities behaving like call options on execution; equity market leverage at 220% of GDP; capex hangover risk as cash flow consumed by spending
- (45:00–50:00) Fed, labor, and Anthropic: Fed Governor Waller on jobs disappearing before new ones emerge; Anthropic dropping safety pledge under competitive pressure; government directing agencies to cease Anthropic use
- (50:00–1:02:00) Bitcoin and the endgame: Howard Marks on AI democratizing innovation; crypto as essential infrastructure for AI-native speed; blockchain as defense against deepfakes and agent swarms; government nationalization of AI as the most bullish catalyst for Bitcoin
