Nebula Labs unveils LumaCore: a 10x faster realtime video-LLM accelerator, raises $60M Series C

Tech · 5 min read

Nebula Labs, a startup building domain-specific accelerators for multimodal AI, announced the LumaCore chip and a $60 million Series C round led by Frontier Capital. LumaCore is optimized for real-time video encoding, frame understanding, and video-language models, enabling sub-100ms inference for common tasks like face analytics, action detection, and captioning at the edge.

The company positioned LumaCore as a bridge between heavy datacenter GPUs and tiny MCUs, targeting smart cameras, retail analytics, and autonomous inspection systems. Early partners include a major surveillance OEM and an industrial drone maker running LumaCore-enabled modules for flight-time visual inspection.

Nebula Labs said funds will support a ramp in silicon tape-out volumes, expand software tooling for model quantization and compiler support, and deepen integrations with major cloud marketplaces so customers can buy devices and orchestration as a service.