Ep5. Stock Comp, AI Cold War, Valuations for LLM | BG2 with Bill Gurley, Brad Gerstner, & Bob Mylod

Episode Summary

Executive Summary: The episode centers on how stock-based compensation distorts incentives and reported profits in Silicon Valley, arguing that SBC should be treated as a real cash expense and measured honestly. The hosts also discuss Apple’s rumored AI partnerships with Google, TikTok’s geopolitical future amid U.S.-China tensions, and Microsoft’s hiring of Inflection’s team as a sign of fierce but uncertain AI economics.

Main Topics: Stock-based compensation as a real expense (Priority: 5/5): The longest discussion argues that SBC should be counted like cash, not excluded via adjusted EBITDA, because it materially affects dilution, incentives, and shareholder returns. Options vs. RSUs and incentive alignment (Priority: 5/5): The speakers contrast startup-era options, which only pay if value is created, with RSUs, which can reward executives even when stock underperforms and weaken alignment with shareholders. Public-company comp culture and peer benchmarking (Priority: 4/5): They criticize compensation consultants and peer-group benchmarking for creating an upward ratchet in pay, encouraging boards to follow inflated norms rather than shareholder-first discipline. Apple, Google, and the AI assistant race (Priority: 4/5): The hosts debate rumors that Apple may use Google’s Gemini, concluding Apple is more likely building its own small language model for Siri while also potentially integrating external models into search. TikTok/ByteDance and the U.S.-China AI cold war (Priority: 4/5): They argue TikTok’s future is increasingly a sovereign-policy issue, with little path for the app to continue unchanged and likely retaliation from China if the U.S. moves to ban or force divestiture. Microsoft’s acquisition of Inflection talent (Priority: 4/5): Microsoft’s hiring of Inflection’s team is framed as a yellow flag for overheated AI valuations and a sign that large strategics may outcompete venture investors in frontier AI. LLM economics, switching costs, and data gravity (Priority: 4/5): The episode closes by questioning whether LLMs have durable moats; the view is that switching costs are low and value may accrue more to data platforms and enterprises than to standalone models.

Key Arguments: Stock-based compensation is a genuine economic cost and should be treated like cash in company reporting and compensation decisions. Excluding SBC from adjusted metrics creates false profitability, encourages excess spending, and misallocates capital. RSUs weaken shareholder alignment because employees can still gain value even when the stock falls, unlike options. Comp committees and consultants often optimize for peer comparables and upward market norms rather than shareholder value. Public companies that aggressively compensate via SBC can outbid startups and inflate talent costs across Silicon Valley. Apple is unlikely to outsource its core personal-assistant AI; it is more likely to build a native small language model and selectively partner in search. TikTok has moved from a business issue to a sovereign-national-security issue, making a clean U.S. future unlikely without major structural change. The Inflection deal suggests AI valuations are fragile and that large platforms like Microsoft may favor strategic hiring over traditional acquisitions. LLM businesses currently have weak switching costs; durable value may depend on data, memory, or integration rather than model quality alone.

Data Points: Booking share count: declined from about 50-52 million to 35 million shares - Used to show how buybacks and disciplined SBC can reduce dilution over time. Booking stock price appreciation: from $10/share to $3,400/share - Cited as evidence that employee rewards should come from shareholder value creation. Microsoft stock-based compensation: $2.8 billion per quarter / over $11 billion per year - Example of how large companies can use huge SBC budgets to compete for AI talent. Microsoft SBC as share of free cash flow: 10% of free cash flow - Despite the dollar size, the company can absorb SBC because of its enormous cash generation. Microsoft dilution: 35 basis points - Illustrates how a giant market cap can make a very large SBC budget look small in percentage terms. TikTok U.S. revenue: $8-$10 billion annually (rumored) - Used to contextualize the size of the U.S. business within the broader ByteDance empire. TikTok/ByteDance revenue outside U.S.: over 90% of revenues outside the United States - Supports the argument that the company is globally dominated and geopolitically exposed. TikTok/ByteDance profits outside U.S.: over 100% of profits outside the United States - Emphasizes that the U.S. unit may be economically secondary or unprofitable. Inflection financing scale: one of three LLM companies with over $1 billion raised - Shows how capital-intensive frontier AI has become. Inflection last round valuation: $4 billion - Referenced in contrast to Microsoft potentially hiring the company’s employees rather than buying the business. Greylock AI round example: $250-260 million at a $1 billion pre-money valuation - Used as an example of aggressive AI market pricing. Netflix SBC dilution: about 1.5% per year - Cited as an example of a more disciplined, lower-dilution compensation model. Comp committee peer pressure: 100% of boards think their teams should be above the mean - Explains the structural upward drift in executive compensation. OpenAI consumer pricing: $20 per month - Mentioned as a consumer chatbot price point in the AI assistant competition.

Pivotal Quotes: "A dollar of stock-based compensation is actually not only should it not be ignored, it's the most single valuable expense." — Bob Mylod: Summarizing why SBC must be counted honestly rather than excluded from core performance metrics. "Show me the incentives and I'll show you the behavior." — Brad / attributed to Charlie Munger: Used to explain why adjusted metrics and comp structures shape management decisions. "I'd rather stick a viper down my shirt than hire a comp consultant." — Warren Buffett (quoted by the hosts): Illustrates their criticism of compensation consulting and peer-driven pay escalation.

Implications: The episode argues for a reset in tech compensation: count SBC transparently, tie pay to true value creation, and beware AI hype, geopolitics, and weak model moats. Boards, investors, and founders should focus on real economics, not adjusted optics.