Decart

Decart is still early in commercialization, and its emerging model is consumption-shaped on two levels. The first is its own products: real-time world models — Oasis for physical AI, Lucy for immersive interactive video — are served live, meaning every user-minute consumes GPU time, so monetization naturally follows usage, whether through consumer experiences or partners embedding the models in their own applications. The second, more immediately commercial layer is DOS, the Decart Optimization Stack: the inference-and-training platform the company built to make real-time generation economical, positioned for high-throughput agentic and world-model workloads — infrastructure sold to others on consumption terms. No revenue figures have been disclosed, and the company is candidly capital-powered for now, having raised a $300M round at a roughly $4B valuation in May 2026 after $100M at $3.1B in August 2025. The honest framing: Decart's bet is that owning the full stack lets it price interactive generative compute profitably before pure-play rivals can, with usage-based revenue from both experiences and infrastructure still being proven.