SED News: Apple’s AI Problem, The Real Business Model of AI, and Token Cost Reckoning

What It Covers

This episode covers three converging forces reshaping tech in 2025: Apple's structural inability to compete in AI despite $1B daily revenue, the enterprise-subsidized business model powering Anthropic and OpenAI's growth, and mounting token cost pressures that will force corporate AI spending into optimization mode.

Key Questions Answered

• •Apple's AI structural gap: Apple's R&D spend dropped from 8% to 2% of revenue during the iPhone boom, and a faked Siri demo at WWDC triggered a $250M class action lawsuit for false advertising. The company's Tim Cook-era focus on financial optimization — 75% services margins, $1T returned to shareholders — structurally conflicts with the risk tolerance required for AI innovation.
• •Consumer AI subscriptions as loss leaders: Simon Willison's token analysis reveals that a $100/month Anthropic Max plan generated $2,000 worth of actual token usage in 30 days — a 20x subsidy. The real business model is enterprise licensing, making Anthropic and OpenAI structurally closer to Salesforce-style B2B SaaS than consumer platforms, with bottoms-up adoption driving top-down corporate contracts.
• •Token cost reckoning is 6 months away: Uber's CTO exhausted the company's full-year 2026 AI budget within months, a budget set in 2025 before these tools became indispensable. Companies should begin auditing per-employee token consumption now — senior engineers running Claude Opus can accumulate thousands of dollars monthly — and evaluate prompt-routing tools that dynamically direct queries to lower-cost models.
• •Agentic context management beats model quality: The competitive moat in AI has shifted from raw model performance to how well systems manage the information environment surrounding the model. Companies report near-equivalent output from lower-tier Anthropic models versus top-tier ones, meaning the differentiator is now the agentic harness — context windows, memory, tool integration — not benchmark scores.
• •App Store revenue faces structural AI erosion: Apple's App Store generates nearly 3x more revenue than Mac hardware sales, but AI assistants are already replacing single-purpose apps for tasks like jet lag planning, weather, and information lookup. As AI interfaces become default entry points for services, Apple loses both the 30% developer cut and the platform lock-in that underpins its services growth trajectory.