Buying stock with borrowed money reaches level of past market tops: 'Very bearish looking'

Margin debt has surged to $1.4T (May) and is up 40%+ over the past 12 months, surpassing Leuthold’s historical “market top” triggers seen in 2000, 2007 and 2021. Over the last year, the S&P 500 returned ~22%, about half the pace of margin debt growth (54% absolute growth and 26% “excess” growth), implying leverage is expanding faster than stocks. Leuthold warns this is “very bearish,” as a slowdown in borrowing can pressure prices and increase the risk of margin calls that amplify downside.

Analysis

This is less a standalone crash call than a leverage gauge turning red. The important market mechanism is that borrowed capital is concentrated in the same high-beta winners that have led the tape, so a modest break in leadership can trigger forced selling well in excess of the initial fundamental move. That makes the first 3%-5% drawdown in crowded growth/AI names more dangerous than the headline index move suggests.

In the near term, the marginal beneficiaries are brokers/prime brokers and options venues from elevated trading and financing balances, but that tailwind is fragile and usually lags the unwind. The cleaner losers are leveraged ETFs, unprofitable AI software, and small/mid-cap momentum names with thin liquidity; if the first crack hits semis or data-center infrastructure, the de-risking can cascade through adjacent baskets via ETF rebalancing and margin calls.

The contrarian read is that leverage is a symptom, not a timer. If breadth broadens and earnings keep validating the AI capex cycle, margin can stay elevated longer than bears expect; what would falsify the bearish thesis is new highs in QQQ/SMH on improving advance-decline breadth and declining implied volatility. But absent that, the setup argues for buying convex protection now rather than waiting for the vol spike, because the cheapest hedge is usually before the first crowded-fund unwind.

The main risk to this view is policy/liquidity support or a continued narrow grind higher in mega-cap tech, which can delay the unwind for months. The structural effect over 6-18 months is a lower tolerance for crowded beta and a stronger bid for defensives and cash-generative balance sheets once the market starts to discriminate.

Key Decisions for Investors

• Buy 3-6 month QQQ put spreads 5%-8% out of the money as a portfolio hedge; target a 2x-3x payoff if crowded AI/growth de-rates 7%-10%, and cut if QQQ holds above recent highs with breadth improving.
• Rotate out of leveraged-beta exposure via a pair trade: short ARKK or TQQQ against long XLP/XLU for the next 1-3 months; this isolates the margin-unwind factor while keeping market beta more controlled.
• Use VIX call spreads only on the first failed breakout or 3%-4% QQQ break lower; the better entry is after the initial gap because the trade needs a volatility expansion trigger, not a slow grind.
• Trim exposure to the most crowded AI infrastructure basket (SMH/NVDA/AMD/ANET/VRT) if it has already become a source of portfolio P&L concentration; this is where forced selling is most likely to propagate if the first crack appears.