The Bond Market Sell-Off Is Appearing in Earnings Reports

Long-dated government bond yields are at multi-decade highs globally, driven by persistent fiscal deficits, heavy debt issuance, inflation and Fed-credibility concerns; Japan's 10-year yield is at its highest since 1996 and the UK's 30-year yield is nearing 6%. AI hyperscaler capex is projected to rise from $750 billion this year to roughly $1.2 trillion next year, with Goldman Sachs estimating 35% of next year's spend will be debt-funded, adding pressure to credit markets. Klarna shares fell about 21% despite 27% revenue growth and a return to profit after it cut full-year guidance, citing FX headwinds and weaker German demand. Home Depot beat expectations but posted only 1.7% comparable-sales growth amid a frozen housing market, with higher costs expected to offset its $730 million tariff-refund benefit.

Analysis

The relevant transmission channel is not the policy rate but the term premium: higher long-end yields simultaneously raise discount rates on long-duration equities and increase the hurdle rate for debt-funded AI capacity. GOOG and META can absorb a moderate funding-cost increase, but the market should begin differentiating between AI spend that produces visible inference revenue and spend supported by lease/JV structures with deferred cash obligations. This is more negative for the broader data-center financing complex than for NVDA near term; supplier revenue can remain resilient for 1-3 quarters even as hyperscaler FCF conversion and valuation multiples deteriorate.

KLAR's issue is less the current credit book than operating leverage into a weaker European consumer and uncertainty around funding spreads through the next downturn. A lower growth trajectory at a premium revenue multiple leaves little cushion for even modest increases in loss provisions, deposit costs, or merchant incentives. The management transition elevates execution risk precisely when investors need confidence in underwriting discipline; a post-earnings decline alone is not a sufficient reason to buy.

HD and LOW are defensible quality exposures but not clean bottom-of-cycle trades until housing turnover, not merely repair demand, improves. Persistently high mortgage rates favor maintenance and small-ticket projects while suppressing the higher-margin pro/professional and large-remodel categories that drive earnings acceleration. The contrarian opportunity is that a credible decline in long-end yields would re-rate housing-adjacent equities before reported housing data recover, but current evidence supports patience rather than pre-positioning.

Key Decisions for Investors

• Maintain a 1-3 month relative-value hedge: long NVDA versus short an equal beta-weighted basket of GOOG and META only if long-duration Treasury yields continue to rise. NVDA has backlog visibility while customer capex return scrutiny should pressure hyperscaler multiples; exit if GOOG/META capex guidance is accompanied by accelerating AI monetization or if long-end yields decline materially.
• Avoid averaging into KLAR after the selloff. Reassess only after the next quarterly disclosure confirms stable loss rates, funding costs, and German transaction trends; a further revenue-guide reduction or material provisioning increase would support a short/watchlist bearish thesis rather than a long.
• Keep HD/LOW on a rate-triggered watchlist rather than initiate outright. Begin scaling long HD over LOW only after mortgage rates fall meaningfully and pending-home-sales/turnover data inflect for two consecutive monthly prints; HD's pro exposure offers greater upside in a genuine renovation recovery, but also greater downside if the freeze persists.
• Overweight GS within financials for the next 1-3 months as elevated sovereign and corporate issuance should support debt-underwriting and trading activity. Size modestly: widening high-yield spreads or a disorderly auction/liquidity event would shift the benefit from issuance fees to balance-sheet and risk-asset concerns.