Stanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Economics of Generative AI

Key Concepts

• AI Value Chain (The "Triangle"): A framework representing the economic distribution in AI, consisting of Semis (chips), Infrastructure (data centers/cloud), and Applications (models/agents).
• CapEx Cycle: The massive capital expenditure required to build the physical foundation (energy, chips, power, interconnects) for AI.
• Inference vs. Training: The shift from training models (high, predictable compute) to inference (bursty, user-driven compute).
• Monetization Models: The transition from subscription-based revenue to potential ad-based models in AI applications.
• Knowledge Work: The current primary use case for AI, which the instructor argues may be insufficient to reach mass-market (4 billion user) scale.

---

1. The AI Economic "Triangle"

The instructor presents a "triangle" model to explain the current state of the AI industry, contrasting it with previous tech cycles like the Internet, Mobile, and Cloud.

• Current State: The industry is heavily bottom-heavy. Approximately 75% of the $350 billion in added revenue over the last two years has flowed directly to the Semis layer (dominated by Nvidia).
• Profitability Gap: The Semis layer enjoys high gross margins (~75%), whereas the Application layer currently struggles with margins between 0% and 30%.
• The "Cloud" Comparison: In the cloud era, it took roughly eight years (2004–2012) for the ecosystem to mature from initial CapEx investment to widespread adoption. The instructor posits that AI is currently in the early stages of this cycle.

2. The Infrastructure and Competitive Landscape

• Competitive Intensity: The Infrastructure layer is identified as the most competitive and unstable segment. Startups are fighting for market share against hyperscalers (Google, AWS, Microsoft).
• Vertical Integration: The instructor notes that the most successful companies in previous cycles (Google in Search, Apple in Mobile) were vertically integrated. He questions whether current AI players will follow this path or remain heterogeneous.
• ASIC Strategy: With hyperscalers developing their own chips (e.g., Google’s TPU, Meta’s MTIA), the instructor warns that startups in the chip space face a "small number of very large orders" rather than a traditional enterprise software sales model.

3. Consumer AI and Monetization

The instructor analyzes the growth trajectory of AI applications like ChatGPT and Gemini compared to established consumer giants:

• Usage Tiers:
  • Mandatory (3B+ users): WhatsApp, Chrome.
  • Social (1.5B–2B users): Instagram, TikTok.
  • Niche: Spotify, Twitter.
• The Challenge: ChatGPT currently sits just above the "niche" category. To reach the 4 billion user scale of Alphabet or Meta, AI must move beyond "knowledge work" and become a daily utility.
• The Ad Model: The instructor predicts that AI will eventually shift toward an ad-based revenue model. He argues that AI-driven ads will be highly effective due to superior intent-matching, logged-in user data, and high-trust conversational interfaces.

4. Course Methodology and Logistics

• Structure: The course is designed for 3 hours of weekly commitment (1 hour class, 2 hours reading).
• Guest Speakers: The curriculum features leaders from across the stack (semis, infra, models, agents). The instructor emphasizes "Adam House rules" (no recording) to encourage candid, high-level insights.
• Grading: 50% attendance, 50% final assignment.
• Objective: To provide students with mental models to evaluate AI businesses, understand the "laws of physics" governing the current cycle, and determine whether a business is a "feature" or a "platform."

5. Notable Quotes

• "Half of you are going to start an AI company. The other half are going to fund it. At the very minimum, you'll have a sense of what not to go after."
• "The incremental user of an AI application is not free... it's actually quite a bit more expensive to have AI users because turns out you've got to burn those GPUs."
• "I suspect the ads that ChatGPT will be able to serve... will have a lot better pricing because they will understand your intent."

Synthesis

The central takeaway is that the AI industry is currently in a massive, capital-intensive "build" phase where value is disproportionately accruing to hardware providers. The long-term success of the AI ecosystem depends on the "inversion" of the economic triangle—where the Application layer generates enough value to justify the massive CapEx at the bottom. The instructor remains optimistic but cautions that the transition from "knowledge work" to mass-market utility, likely supported by a sophisticated ad model, is the critical hurdle for the next decade.