uRun is an inference cloud and infrastructure layer for interactive generative AI. The company provides real-time, stateful inference at scale through APIs, with a runtime designed to keep sessions alive so that each interaction builds on the previous one. By its own account, the latency between prompt and output is eliminated, supporting continuous creation across sessions and streams.
Its product line includes a stateful inference runtime delivered as a universally available compute layer, and a real-time AI video generation platform that produces live, steerable, continuous video responding as users speak or act. The video platform is described as the proving ground for the underlying inference infrastructure, allowing scenes to be generated, steered and evolved without delay.
The company recently emerged from stealth and is backed by top-tier investors. It is led by founders with experience scaling inference infrastructure for demanding generative AI workloads, and positions itself as an infrastructure partner to model labs, builders and research teams, with an emphasis on production-grade reliability.
uRun's technical work spans real-time stateful inference, streaming and low-latency compute, inference infrastructure scaling and API-based runtimes. Its stated verticals include generative AI and model labs, AI video generation, developer tooling, AI research and interactive media.
