Crystal Ball: Will the AI bubble burst or balloon in 2026?

Leading VCs, founders and operators predict 2026 will be a reckoning for AI as the market pivots from novelty toward measurable ROI, driving valuation markdowns for many application-layer startups while favoring vertical AI platforms with proprietary data and real margins; forecasts also anticipate a shift from giant foundation models to specialized, memory-enabled agents and wider adoption of open-source models. The newsletter also highlights continued private-market activity with Stonepeak's $10.1 billion majority stake buy of Castrol, TransDigm's $960 million purchase of Stellant Systems, VSE's $350 million acquisition of Aero 3 and Legence's $325 million purchase of The Bowes Group, underscoring selective liquidity in industrial and asset-heavy sectors amid a recalibration of AI funding expectations.

Analysis

Market structure: The 2026 pivot will reward foundation-model owners, vertical AI with proprietary workflows/data, and industrial/defense suppliers who offer steady cashflows. Expect pricing power to shift away from thin application-layer vendors toward incumbents that can prove ROI (procurement cycles lengthen to 6–18 months). GPU demand may normalize (reducing short-term capex for hyperscalers) while enterprise spend reallocates to workflow integration and on-prem/edge solutions.

Risk assessment: Tail risks include regulatory/legal shocks (EU/US AI safety rules or liability rulings) and a liquidity shock in late-stage funding that could force private markdowns >30% within 12 months (probability 15–25%). Immediate risk (days–weeks) is elevated equity/IV volatility; medium-term (3–12 months) is valuation compression for high-burn app vendors; long-term (2–5 years) winners are likely verticals with ≥20% gross margins and exclusive data moats. Hidden dependency: many apps rely on low inference pricing — a 2x increase in per-token costs would materially compress gross margins.

Trade implications: Tactical: establish a 2–3% long in TDG (TransDigm) and 1–2% long in VSEC (VSE) to capture stable aerospace/defense M&A optionality and recurring services revenue over 6–12 months; hedge by initiating a 2% short in ARKK (or equivalent speculative growth ETF) to express app-layer downside. Options: buy 6-month 25-delta puts on ARKK funded by selling 3–6 month covered calls on TDG/VSEC. Rotate 2–5% portfolio weight from small-cap AI app names into industrials/defense and selected infrastructure vendors.

Contrarian angles: The market underestimates the deflationary force of open-source models and specialized small agents — inference economics could fall 30–50% for many use cases, enabling low-cost vertical entrants. The reaction may be overdone for durable, cash-flowing application vendors with proprietary data; these could become consolidation targets, producing 20–40% upside on acquisition within 12–24 months if public/private valuations reset lower and strategic buyers step in.

Key Decisions for Investors

• Initiate a 2–3% long position in TDG (TransDigm) within 30 days; target 10–20% upside over 6–12 months driven by accretive defense M&A and recurring aftermarket margins; sell 1–2% notional of 3–6 month covered calls at a 5–8% premium to fund downside protection.
• Allocate 1–2% long to VSEC (VSE Corporation) to capture services tailwinds in aerospace MRO; hold 6–12 months and add another 1% if shares fall >10% on broad tech weakness.
• Establish a 2% short exposure to speculative AI/app-layer risk by shorting ARKK (or equivalent high-growth ETF) as a hedge; size to be delta-neutral vs TDG/VSEC longs and increase short if ARKK falls >15% to harvest volatility.
• Purchase 6-month 25-delta puts on ARKK (size ~0.5–1% of portfolio) to protect against a rapid re-rating of AI application multiples; fund these by selling 3–6 month covered calls on TDG/VSEC as above.
• Reduce aggregate exposure to small-cap, unprofitable AI application equities by 30–50% from current weights within 60 days; redeploy proceeds into industrials/defense and infrastructure names, and consider selective M&A target scouting for companies with >$10m ARR and proprietary vertical data (pay attention to acquisition windows if private rounds slow).