Episode

Bitter Lessons in Venture vs Growth: Anthropic vs OpenAI, Noam Shazeer, World Labs, Thinking Machines, Cursor, ASIC Economics — Martin Casado & Sarah Wang of a16z

Podcast: Latent Space: The AI Engineer Podcast
Published: Feb 19, 2026
Duration seconds: 3318
Processing state: processed
Canonical source: https://www.latent.space/p/a16z
Audio: https://api.substack.com/feed/podcast/188504140/c66a34d6406cd1b378cffe52d8a8c00b.mp3
JSON: /v1/public/podcasts/latent-space-ai-engineer/episodes/bitter-lessons-in-venture-vs-growth-anthropic-vs-openai-noam-shazeer-world-labs-thinking-machines-cursor-asic-economics-martin-casado-sarah-wang-of-a16z
Markdown: /podcast/latent-space-ai-engineer/bitter-lessons-in-venture-vs-growth-anthropic-vs-openai-noam-shazeer-world-labs-thinking-machines-cursor-asic-economics-martin-casado-sarah-wang-of-a16z.md

Actions

POST https://stenobird.com/v1/public/podcasts/latent-space-ai-engineer/episodes/bitter-lessons-in-venture-vs-growth-anthropic-vs-openai-noam-shazeer-world-labs-thinking-machines-cursor-asic-economics-martin-casado-sarah-wang-of-a16z/transcription-requests
Idempotently request low-priority transcript generation for this episode.
GET https://stenobird.com/podcast/latent-space-ai-engineer/bitter-lessons-in-venture-vs-growth-anthropic-vs-openai-noam-shazeer-world-labs-thinking-machines-cursor-asic-economics-martin-casado-sarah-wang-of-a16z.md
Read the agent-friendly Markdown representation of this episode resource.

Summary

a16z partners Martin Casado and Sarah Wang analyze the blurring lines between venture and growth capital in the AI era. They explore how massive compute-driven rounds are reshaping startup economics and the strategic tension between foundation model labs and the application ecosystem.

Topics

Artificial Intelligence
Venture Capital
Foundation Models
Compute Economics
Machine Learning Infrastructure
Robotics
Software Engineering
Generative AI

Highlights

Main idea: The distinction between venture and growth investing is disappearing as AI startups require billion-dollar 'compute contracts' almost immediately after inception
Failure mode: Foundation model companies risk 'borrowing against the future' by running gross margin negative to fund massive training runs
Practical takeaway: 'Agent labs' may capture more value than model labs by pricing against human labor costs rather than the commoditizing price per token
Strategic tension: The industry faces a fork between an oligopoly of general models and a fragmented landscape of highly specialized vertical applications
Market opportunity: While much attention is on frontier models, 'boring' enterprise software and vertical robotics remain significantly underinvested

Chapters

1:00 The New AI Investment Thesis: Discussion on the aggressive, broad-scale investing approach required to back frontier model companies like Anthropic and OpenAI.
5:10 The Convergence of Venture and Growth: How massive capital requirements and complex compute negotiations are blurring the lines between early-stage and late-stage investing.
9:20 The Token-to-Product Flywheel: Analyzing how low friction between inference and product creation allows for rapid, high-frequency funding rounds.
17:45 Underinvested Opportunities in AI: Identifying the gap in enterprise software investment and the potential for value in non-hype sectors.
22:20 The Future of Robotics and 3D: Evaluating the investment landscape for vertical robotics and the impact of generative 3D on scene creation costs.
30:35 The Economics of Training vs. Inference: Examining the sustainability of negative gross margins in model training and the long-term impact on model labs.
42:55 Verticalizing Up: The Cursor Case Study: How application-layer companies like Cursor can build upward by training their own models on proprietary product data.