EN / RU
← Back to essays
Β· Essay Β· 2 min

First Lecture Club Call on Stanford MS&E 435

The first Lecture Club call on Stanford MS&E 435 gathered about 80 people and sparked a lively discussion.

πŸŽ™ We held the first Lecture Club call on Stanford MS&E 435: Economics of the AI Supercycle.

Thanks to everyone who attended β€” there were about 80 people, and the discussion was lively and at times contentious. Specifically: we caught a couple of "uninvited guests" on Zoom β€” I will enable the waiting room and password for the next meeting to prevent this from happening again.

Here are some thoughts highlighted by the participants:

β€’ Mikhail β€” the current monetization of AI looks like "we're reinventing the wheel again": advertising and subscriptions. The marketing budget in the world is limited β€” new growth is only possible through autonomous agents replacing employees.

β€’ Alexander β€” what if the triangle doesn’t flip? Apps earn only about ~30% gross margin, ChatGPT is deeply unprofitable, advertising will undermine trust. Plus, there's an open question: how to show ads to an agent, not a human?

β€’ Sergey β€” real monetization is already on the horizon: Agentic Commerce. ChatGPT is biting into Amazon and TikTok Shop through the Agentic Commerce Protocol β€” 5–15% from each transaction. It’s surprising that Amazon hasn’t yet released Rufus as a standalone app.

β€’ Svetlana (an economist at Amazon) β€” each previous upheaval of the pyramid required a new monetization model (license β†’ ads β†’ SaaS). A new one for AI is not yet visible β€” hence all lectures feel like fortune-telling.

β€’ Mika β€” "not all the work is knowledge work". A large part of the story is about physical AI: robots and interaction with the world. And advertising in the AI era will likely be nothing like the current contextual banners.

β€’ Pasha β€” there are not 3, but 4 levels in the pyramid: model sellers through API (OpenAI, Anthropic) β€” a separate layer, not applications. And the future is salaries for agents, not software subscriptions.

β€’ Ira β€” a counter question: will users buy a subscription to avoid ads? How does this work in YouTube Premium?

β€’ Kostya β€” the cost of inference is falling, open models are readily available. "Middle" providers with decentralized infrastructure will emerge. Money is in agents: SaaS + agent as the new face of services.

πŸ“Ί Recording of the call:
https://youtu.be/4oawuoWhYmg

πŸ“† Next meeting β€” Monday, May 11, 10am ET / 5pm MSK
πŸ“ Registration on luma: https://luma.com/ilz7seef

To watch β€” Lecture 2: Silicon, The GPU Economy
πŸŽ₯ https://www.youtube.com/watch?v=4faCRNl9Bi4

Additional materials:
β€’ Acquired: Altimeter (with Brad Gerstner) β€” https://www.youtube.com/watch?v=npBC3R-dOr0
β€’ Jonathan Ross (Groq): "OpenAI & Anthropic will build their own chips. Will NVIDIA reach $10T?" β€” https://www.youtube.com/watch?v=VfIK5LFGnlk

Discuss in the chat: @rvnikita_blog_chat

First Lecture Club Call on Stanford MS&E 435 β€” illustration