Booking, Expedia face margin pressure from AI agents, Bernstein says

Bernstein warns that AI platforms could materially disrupt OTAs, compressing margins and eroding the long-standing "supply moat"; Booking's EBITDA margin is cited as over 300bps below 2015 levels and Expedia's B2B now represents >33% of revenue. The firm maintains Market-Perform ratings on BKNG and EXPE but highlights terminal-value risk as AI narrows informational advantages, likely pressuring take-rates and marketing efficiency and forcing higher ad/spend. This is a structural, medium-term sector risk that may weigh on multiples and margin forecasts rather than an immediate credit or downgrade catalyst.

Analysis

The structural threat from high-context AI search is less about immediate traffic loss and more about a multi-year erosion of the pricing funnel and advertising efficiency that underpins OTA economics. Expect a gradual flattening of CPC and CPC-like auction dynamics as AI agents favor a narrow ranked set of suppliers — that compresses marketing ROI and forces OTAs to pay more to maintain room-night flow or concede margin to hoteliers directly.

Second-order winners will be middleware and distribution utilities that make hotels and niche specialists “AI-ready” (channel managers, real-time rate APIs, and curated inventory partners): these reduce friction for direct-book conversion and become new bargaining chips for hotels. Conversely, legacy metasearch feed sellers and high-marketing-cost customer acquisition models will see the biggest margin contraction; business models that rely on long-tail CPC inventory are most exposed.

Timing: market moves should unfold over 6–24 months as conversational AI prototypes gain commercial integrations, but terminal value questions hit analyst models over 2–5 years when take-rate normalization is priced in. Reversal catalysts include slow monetization of AI agents, hotel coalition countermeasures (exclusive direct-booking offers, API-blocking), or regulatory limits on bundled results — any of which could restore a premium to scale OTAs faster than current consensus assumes.

Key Decisions for Investors

• Pair trade (6–12 months): Short EXPE vs Long GOOGL (ratio 1:1). Rationale: EXPE faces larger immediate take‑rate risk and B2B reliance, while Google can re-monetize AI via ads/search integrations; target P&L: 20–30% downside on EXPE vs 15–20% upside on GOOGL. Use 6–12 month put spread on EXPE (buy 6–9 month 30–40 delta puts, sell 25–30 delta puts lower strike) financed by selling a small portion of long-dated GOOGL calls if needed to improve cost.
• Hedge BKNG exposure (3–9 months): If long BKNG, buy 3–9 month protective puts (30–35 delta) rather than exiting; BKNG’s merchant model cushions share shocks but margin erosion is likely. Risk/reward: pay ~2–4% option premium to cap a potential 25–35% drawdown while keeping upside optionality if AI monetization is slower than feared.
• Event-driven short (earnings shelf, next 2 quarters): Initiate a concentrated short/inverse exposure to EXPE ahead of the next earnings cycle via options (buy 3–6 month puts) sized so max loss = 8–12% of equity allocation. Rationale: analysts will likely lower guidance as marketing efficiency degrades — asymmetric payoff if EPS beats become misses.
• Volatility/term-structure trade (12–24 months): Sell short-dated volatility on OTAs (sell 3–6 month straddles/strangles around BKNG or EXPE) and buy longer-dated puts (12–24 month) to capture mean reversion if short-term AI headlines spike fear but long-term fundamentals remain intact. Target carry: collect premium while limiting tail risk via long-dated protection.
• Monitor triggers to flip long OTAs (3–24 months): Revisit long BKNG/EXPE when either AI platforms fail to secure inventory integrations (measured by fewer than 2 major hotel chains live in an AI booking flow) or when OTA take-rates compress >200bps in reported quarters. Until then, keep OTA longs sized underweight relative to tech/AI beneficiaries.