World Models
“World Models” – A cartoon illustrating how Meta missed the opportunity to take a leadership role in the building of AI world models.
At its core, an AI model learns patterns in data. Early systems handled basic tasks like sorting and labeling information. Large language models pushed this further by predicting what comes next in a sequence of words. World models take another step entirely, focusing on how the world changes over time, especially when actions are taken. This shift from prediction to simulation is the real break from the past.
World models matter because planning is fundamentally different from guessing. A system that can simulate outcomes can reason over longer time horizons, test strategies, and support autonomous behavior. This is also why synthetic data suddenly matters. When Gartner predicted that 60% of AI training data would be synthetic, the point was not fake content, but the need to generate scenarios the real world does not conveniently produce. World models turn that data into rehearsal.
Demand is already visible in some adjacent markets. Digital twins and AI-driven simulation, essentially the narrow predecessors to world models, are projected to reach $250 billion by the early 2030s, growing at more than 30% annually. As simulation moves from a niche tool to core infrastructure for decision-making and agent training, estimates for the standalone world-model opportunity already exceed $100 billion.
At the frontier, leading labs are explicitly building foundation world models, training on video and action-conditioned data to simulate environments rather than generate content. Consumer use cases resemble life rehearsal, career choices, financial decisions, and health trade-offs, while enterprise use cases include scenario engines for supply chains, pricing, robotics, and risk management.
Meta should own this space, given its deep investment in virtual reality, large-scale simulation environments and acquisitions across VR hardware … as well as a social platform at planetary scale. Instead, it has spent tens of billions chasing the metaverse, delivering expensive hardware and AI video slop.
