US AI spending peak may arrive this year, Jefferies warns

Jefferies warns 2026 may mark the peak of the AI investment boom as doubts about AI capex returns rise and firms increasingly fund AI via debt/private credit, amplifying risks to private credit and private equity. The bank flags the Iran conflict and a closed Strait of Hormuz as a stagflationary risk, and rising Treasury volatility last quarter as a signal of potential equity stress. Morgan Stanley reports Asian power tightening with prices up 10%–100% since Feb 2026, Asia ex-China importing 15% of power needs, coal generation increasing, and storage+renewables now 20%–25% more competitive versus LNG, benefiting large power generators, coal producers and grid operators.

Analysis

Debt-funded AI buildouts create a fragile intersection: when capex returns miss expectations, firms that financed expansion with private credit and covenant-lite structures will face a two-way squeeze — margin pressure from weaker-than-expected revenue and mark-to-market pressure on leveraged loan/credit fund NAVs. Model runs show a 150–300bp widening in leveraged loan spreads can translate to mid-teens NAV markdowns for highly levered credit vehicles within 6–12 months, forcing either asset fire-sales or frozen distributions to LPs.

Rising Treasury volatility is not benign for long-duration growth exposures; historic episodes where 10y vol jumps precede a sustained re-rate of growth multiples point to a likely 5–12% downside for the highest multiple software and semiconductor names over a 3-month horizon absent stronger macro confirmation. Hedging short-term rate and volatility pathways (10y breakevens, swap spreads) is therefore a higher-conviction defensive move than sector rotation alone.

On energy, tighter gas availability in Asia and shipping threats around key chokepoints accelerate demand for thermal coal and favour vertically integrated utilities and exporters with spare capacity. That creates a tactical window (3–12 months) to capture elevated power spreads and coal price pass-through, but it also raises policy and stranded-asset tail risk for fuels-heavy producers once LNG/electrification economics normalize.

The combination produces cross-asset arbitrage opportunities: long physical/commodity producers and credit protections while shorting high multiple, debt-financed growth assets. Key catalysts to watch are private credit fundraising/redemption flows, leveraged loan spread moves, 10y vol spikes, and announced capex slowdowns from large AI spenders — these will crystallize P&L outcomes within 1–6 months.

Key Decisions for Investors

• Buy protection on credit-sensitive names: purchase a 3-month put spread on the BKLN loan ETF (buy BKLN 3m 90/85 puts) sized to cover 5–10% portfolio exposure; cost should be <1.5% of notional, offers asymmetric payoff if leveraged loan spreads widen 150–300bp.
• Long thermal-coal exposure: buy ARCH Resources (ARCH) and Peabody Energy (BTU) outright (3–12 month horizon). Target 20–35% position notional split; upside if Asian coal demand sustains, capped downside via stop-loss at 15% to limit rapid policy/price reversals.
• Tail hedge rates/vol: buy 3-month TLT straddle or long 10y Treasury call options (if available) sized to offset 25–50% of the portfolio's duration bucket. Expect hedge cost ~1–2% of notional but protects against rate-driven multiple compression.
• Pair trade — long physical, short high-multiple growth: go long BTU (or ARCH) and short a basket of high-multiple AI-exposed software names (example: SNOW, COUPLED ETF) in a 60/40 notional weighting for 3–6 months. Rationale: capture commodity/utility outperformance vs. multiple contraction in debt-financed growth; set stop-losses at 12% on each leg.
• Reduce new private credit commitments and reprice allocations: pause allocations to private credit funds for the next 6 months and shift 50% of dry powder into liquid credit hedges (HYG put collar or CDS indices) to avoid being caught in GP-led repricing or NAV freezes — preserves optionality if entry points widen.