AI could widen the funnel for OTAs, not shrink it

Bernstein outlines a bullish but conditional case that large language models could materially expand the TAM for online travel agencies by pulling a meaningful share of the roughly 40% of hotel bookings that still occur offline into digital channels and potentially forcing 10–15% of hotels that currently avoid OTAs to list. The report highlights potential cost efficiencies—AI could deliver “zero-cost” top-of-funnel traffic and lower acquisition costs—noting Booking’s 2024 marketing spend was roughly equal to its EBIT and Expedia’s marketing spend was nearly six times its operating income; the ultimate impact depends on how LLMs monetize and whether hotels preserve direct-booking controls.

Analysis

Market structure: Winners are large OTAs (BKNG, EXPE) and infrastructure chip suppliers (INTC, SMCI) because LLM-driven discovery can expand digital share of a 40% offline hotel market; a realistic capture of 20–30% of that offline share over 2–3 years implies ~8–12% incremental industry bookings and a potential 5–12% revenue lift for top OTAs after commissions. Losers include Google (GOOGL) if AI fragments the top funnel and CPC pricing weakens, smaller OTAs that lack scale/LLM partnerships, and pure-experience players (ABNB) if discovery centralizes around OTA inventory.

Risk assessment: Tail risks include LLM platforms adopting revenue-share (vs CPC) which could cut OTA take rates by 200–400 bps, hotel chains refusing API access, or swift regulation of embedded shopping—each could wipe 10–30% off the bull case within 6–24 months. Immediate moves will be partnership/news-driven (days–weeks), structural margin shifts play out in 6–18 months, and full TAM conversion is a 2–5 year outcome; hidden dependencies are data-sharing, payment flows and cancellation-friction reductions that materially change unit economics.

Trade implications: Favor concentration into large OTAs and selective semiconductor names: size positions small (1–3% each), use pair trades (long BKNG/EXPE vs short ABNB/GOOGL) to isolate travel-disintermediation exposure, and use options to cap downside—e.g., 6–12 month BKNG call spreads and 3–6 month GOOGL protective puts. Rotate away from ad-revenue cyclicals if LLM monetization trends toward zero-cost top-of-funnel; rebalance on quarterly results or LLM revenue-model announcements.

Contrarian angles: Consensus underestimates Google’s ability to monetize AI via CPC/reinforced ad formats—if Google wins the ad-LLM revenue model, OTA CAC reductions may be smaller than expected and GOOGL downside limited. Airbnb may be over-penalized; its unique supply and experiences are less substitutable, so ABNB downside beyond 20% could be an activation point. Historical parallel: meta-search + mobile did not kill OTAs; unintended outcomes include hotels losing leverage and compressing OTA margins, or AI improving conversion but also increasing cancellations by 3–6% if planning becomes speculative.

Key Decisions for Investors

• Establish a 2.5% long position in Booking Holdings (BKNG) within 1 month; target +15% price appreciation over 6–12 months, set a hard stop-loss at -12% and add 1% on any quarterly marketing spend decline >20% YoY or formal LLM distribution partnership announcement.
• Initiate a pair trade: long 2% Expedia Group (EXPE) and short 1.25% Airbnb (ABNB); time horizon 3–9 months, expected spread capture ~10%, unwind if ABNB announces equivalent global LLM inventory partnerships or EXPE/ BKNG miss bookings guidance by >3 ppts.
• Buy a 6–12 month ATM call spread on Intel (INTC) sized to 1% portfolio risk (buy ATM, sell +12–20% OTM) to capture chip upside from AI infrastructure demand; take profits if INTC rises >25% or cut if gross margin guidance misses by >150 bps on next quarter.
• Hedge ad-revenue exposure by buying 3–6 month GOOGL puts (delta ~0.25) sized so max premium = 1% portfolio; cover if GOOGL reports ad revenue growth beating consensus by >200 bps or if major LLM providers publicly commit to CPC-based monetization within 90 days.