Aurora Labs raises $120M Series B to scale on-device LLMs for mobile apps

AI · 6 min read

Aurora Labs said today it closed a $120 million Series B led by Meridian Capital with participation from several strategic device OEMs. The startup’s selling point is an LLM architecture tuned to run across modern mobile NPUs and tiny edge accelerators while keeping model footprints under a few hundred megabytes.

The company also shipped a new SDK that integrates with Android, iOS, and common C++ embedded stacks, providing encrypted model updates and on-device fine-tuning hooks. Aurora claims reductions in inference latency of 3–5x compared with cloud-first alternatives and highlights privacy benefits for industries such as finance and healthcare.

Aurora plans to use the funding to expand partnerships with chip vendors, hire engineers for model compression and scheduling, and launch a developer program providing credits and device labs. The round signals growing investor appetite for AI that avoids constant cloud round trips and aligns with regulators pushing for data minimization.