Is AI a Mistake? (EP. 452)

What It Covers

Michael Batnick and Ben Carlson debate whether AI represents a net negative for society, examining software stock selloffs in companies like Salesforce and ServiceNow, the rapid rise of Anthropic to a $14B revenue run rate, prediction market growth threatening DraftKings, and whether current market volatility signals a broader reckoning or recency bias overreaction.

Key Questions Answered

• •AI Task Completion Horizon: A data scientist listener at a major payments company reports that AI's task completion time horizon doubles every seven months — three years ago it saved 60 seconds on a line of code; today it handles four-hour workloads. Enterprise headcount for one engineer pencils at $1M annually, making 12-figure CapEx investments rational as features now cost a fraction of previous prices.
• •Mutually Exclusive Market Narratives: Bank of America analyst Vivek Arya identifies a logical contradiction in current market pricing: investors cannot simultaneously believe AI CapEx won't deliver ROI and that AI will destroy all SaaS companies. If AI is powerful enough to eliminate software incumbents, infrastructure builders see massive returns. If infrastructure spending won't pay off, AI lacks the power to disrupt SaaS. Both cannot be true.
• •Enterprise Software Switching Costs: ServiceNow implementations take 12–18 months; Workday migrations span multiple years. Enterprise buyers optimize for career risk, not unit cost — "we went with Salesforce" is defensible in any boardroom, while a vibe-coded weekend app functions as a resignation letter. AI coding tools accelerate competitive clock speed symmetrically, meaning incumbents with existing distribution and data hold structural advantages over startups.
• •Anthropic Revenue Trajectory: Anthropic reached zero revenue in January 2023, $100M by early 2024, $1B a year later, and now runs at a $14B annual rate. One in five businesses on the Ramp corporate card platform now pays for Anthropic, up from one in 25 a year ago. Anthropic's customer base overlaps 79% with OpenAI's, suggesting direct competitive displacement rather than market expansion.
• •Prediction Market Accuracy vs. Professional Forecasters: An NBER working paper finds Kalshi participants match highly trained economic forecasters on accuracy for key indicators including Fed rate decisions. A London Business School and Yale study shows Polymarket bettors forecast corporate earnings more accurately than paid Wall Street analysts. The structural advantage: bettors can abstain when uncertain, while professional forecasters must produce estimates regardless of conviction level.
• •S&P 500 Earnings vs. Stock Volatility Disconnect: Despite 115 S&P 500 stocks declining 7% or more in single sessions over eight trading days — a pattern historically associated with bear markets — the index holds near all-time highs. Simultaneously, Q1 2025 revenue growth for the S&P 500 tracks at roughly 9%, the strongest quarter since 2022, creating a significant divergence between underlying business performance and individual stock price behavior.