Trend Trend 13: world models let AI learn physics and predict what happens next
Summary
The thirteenth trend in CB Insights' outlook is that AI is learning how the physical world works. So-called world models simulate environments so that AI can predict what will happen next. These systems learn physics from video, images and simulation, building an understanding of the world that lets them anticipate future states. CB Insights calls them AI's next frontier, and leading researchers are founding startups on the bet that they will enable spatial intelligence beyond what language models can do. Big technology companies are also investing, and mentions on earnings calls are rising. Autonomous vehicles and robotics are the first commercial battlegrounds, though many labs are still in the research phase.
Classification
Evidence 3
- 2026 Tech Trends CB Insights page=88;section=Physical AI & robotics / AI learns how the physical world works 2026-01 accessed 2026-07-26
- 2026 Tech Trends CB Insights page=89;section=Physical AI & robotics / AI learns how the physical world works 2026-01 accessed 2026-07-26
- 2026 Tech Trends CB Insights page=88;section=Physical AI & robotics / AI learns how the physical world works 2026-01 accessed 2026-07-26
Observed signals 4
- SignalA gaming platform's video archive becomes a training asset worth rejecting a 500 million dollar offer
- SignalLarge technology incumbents are hiring and publishing in world models
- SignalMentions of world models on earnings calls rise as autonomous vehicles and robotics lead commercialisation
- SignalSenior AI researchers leave incumbents to found world model startups
Part of issues 1
- IssueWorld models enter commercialisation while most laboratories remain in research1 trends · 1 signals
Relation types: constitutes · supports
Public id: fm-8c6c646231fc
