TIP825: Meta, Adobe, Booking Holding w/ Stig Brodersen, Tobias Carlisle & Hari Ramachandra

Hari Ramachandra presented Meta as a strong business despite a roughly 20 percent decline from its peak at the time of recording. Operating margin stands at 41 percent, net margin at 30 percent, and free cash flow in 2025 at $46 billion, with revenue growing at an 18.45 percent CAGR over five years. Meta's forecasted 2026 ad revenue of $243 billion would exceed Google's projected ad revenue by $3 billion. TIP's internal model valued Meta at $775 against a trading price of approximately $600, implying a 46 percent upside in Hari's base case.

The central debate around Meta was its projected $135 billion in capex for data centers and AI infrastructure. Tobias Carlisle acknowledged Zuckerberg as one of the best managers currently operating but argued that AI chips age faster than traditional infrastructure like railways or fiber optic cable, making the capex cycle riskier. His primary concern was not a catastrophic loss but a multi-year period of multiple compression while the capex is absorbed, with the key uncertainty being whether the investment generates revenue above what the business already produces. Hari countered that AI model quality is becoming commoditized, shifting the advantage toward distribution and scale, which benefits Meta. He argued Meta's walled garden data remains an advantage even as compute costs fall, enabling better ad targeting without requiring a subscription model. Stig offered a counterpoint that if computing becomes essentially free, Meta's data advantage could erode if everyone could collect and process equivalent data.

Tobias Carlisle pitched Booking Holdings, which owns Booking.com, Priceline, Kayak, and OpenTable, as a reasonable risk-adjusted bet. He estimated intrinsic value at approximately $220 per share against a trading price of roughly $167, a discount of about 30 percent. The business is capital-light because it does not buy hotel rooms upfront, carries high return on invested capital, strong free cash flow, high switching costs, and has been buying back a significant amount of stock as the price traded down. The primary AI risk is disintermediation, meaning LLMs could allow travelers to book directly without visiting Booking's sites. Tobias noted that Google posed the same threat throughout most of Booking.com's history and the company continued to grow. His counter-thesis is that LLMs will need to access Booking's database of specialized relationships to fulfill bookings, and the bull case is that Booking becomes a seamless API channel for LLMs and grows faster than historically. Hari agreed that Booking could function as a headless API plugin to ChatGPT or Claude, remaining in the transaction flow even if users never visit the site directly.

Stig was less convinced, noting he increasingly used LLMs for travel planning during a recent Paris trip and found them very helpful, and that a user could theoretically instruct an LLM to book a hotel without needing Booking.com at all. Hari cautioned that even without full disintermediation in the short term, Booking risks losing mind share and on-site real estate used for cross-selling and advertising over the longer term. Tobias acknowledged that a roughly 30 percent discount may or may not be sufficient margin of safety given these risks, and suggested Booking Holdings may serve as a bellwether for how LLMs are impacting other businesses.

Stig Brodersen presented Adobe, which has a market cap of roughly $100 billion and was trading near a seven-year low at the time of recording. Adobe generates $23 billion in revenue, 96 percent from subscriptions, with more than 41 million paying users and 850 million monthly active free users. The digital media segment, including Creative Cloud and Document Cloud, accounts for 76 percent of the business. Stig argued switching costs are Adobe's most important moat, reinforced by the fact that entire teams are trained on the software and the annual cost is negligible for larger companies. He noted that professional designers have little incentive to adopt AI tools that level the playing field with entry-level workers, and employees generally lack incentive to embrace efficiency gains that may threaten their jobs. Stig is buying closer to $200 against an intrinsic value estimate of $350 to $400, while the stock trades around $224. He caveatted that if AI becomes ten times better than Adobe products, switching costs would eventually be overcome.

The central debate around Adobe is top-of-funnel risk. If new users never enter the Adobe ecosystem in the first place, both terminal value and the multiple assignable to the business are impaired. Stig previously felt comfortable assigning a high multiple given 96 percent subscription revenue but said LLM risk makes that harder to justify. Tobias believes any decline would take much longer than the market currently prices in and frames Adobe as potentially a cash cow in slow decline, in which case aggressive buybacks are the correct capital allocation response. Stig noted the outcome depends partly on whether the world comes to value abundant cheap content or high-quality differentiated content, and allowed that in five to ten years there could be significant pushback against AI-generated work as people seek the human element.