AI Hardware & Chip Architecture · Wafer-scale inference architecture breaks Moore's law trajectory with >2x annual gains

episode briefing

extracted intelligence

All-In Podcast

Open Source Wins, AGI Is Here, and Scorsese’s AI Toolkit with CEOs of Cerebras & Black Forest Labs

now viewing▶ 13:00

Wafer-scale inference architecture breaks Moore's law trajectory with >2x annual gains

2026-07-10 · 5 company · 8 thematic

sentiment

3 bull0 bear2 neu

speakers

andrew feldman

Leads Cerebras, a public company (CBRS) building wafer-scale AI inference chips; previously co-founded SeaMicro (acquired by AMD); describes $25B backlog and global data center buildout at unprecedented scale.

robin rombach

Co-invented latent diffusion (basis for Stable Diffusion, Flux); leads Black Forest Labs building open-source multimodal models for images, video, audio, and robot action prediction; company has 100+ employees across Germany and San Francisco.

now playing · AI Hardware & Chip Architecture

AI Infrastructuretailwindscore 9/10andrew feldman

Global AI data center buildout exceeds 50 years of historical power capacity growth

Individual data center campuses now consume more power than midsize cities, with buildouts across US, Canada, Nordics, Europe, Middle East, and Central Asia; total projected power demand su…

AI Hardware & Chip Architecturetailwindscore 9/10andrew feldman

Wafer-scale inference architecture breaks Moore's law trajectory with >2x annual gains

Cerebras' novel architecture avoids the diminishing returns of 20-year-old GPU designs by optimizing for inference workloads directly, enabling 15x faster token generation and a performance…

Open Source AItailwindscore 8/10andrew feldman

Sovereignty drives demand for domestic open-source models as alternative to US/Chinese dominance

Countries and regulated enterprises (finance, healthcare) require on-premises, auditable models they control; with only OSS 12B and Chinese models currently available, there is a strategic…

AI Agentstailwindscore 8/10andrew feldman

Recursive 'loop maxing' and self-improving agents create exponential capability gains

Models that can critique their own outputs, iterate, and chain reasoning loops produce 'vastly better answers' not incrementally better ones; this recursive learning compounds at fruit-fly…

AI Economics & Business Modelstailwindscore 8/10andrew feldman

Abundance thesis: AI enables unlimited intelligence, energy, education, healthcare at near-zero marginal cost

The convergence of recursive AI, robotics, and compute scaling could solve cancer, provide personalized AI tutors (Aristotle-to-Alexander scale), unlimited clean energy via materials discov…

Robotics & Physical AItailwindscore 8/10robin rombach

Multimodal world models unify video generation and robot action prediction

Pre-training on video gives implicit physics understanding; the same model architecture that generates video can predict robot actions, enabling deployment on physical hardware with minimal…

AI Regulation & Policyriskscore 7/10andrew feldman

Government red-teaming of frontier models before release is reasonable risk management

As models gain cyber-offensive capabilities (finding zero-days in hours), staged releases with government red-teaming parallels pharmaceutical safety protocols; polarization obscures the le…

AI Applicationstailwindscore 7/10robin rombach

Generative AI cuts high-end film production costs 5x enabling previously ungreenlit projects

Real-world production use case: a $30M Bitcoin movie used generative AI for virtual sets instead of $150M physical production; this democratizes high-end visual storytelling and creates new…