Aurora Labs raises $120M and debuts Helios — a modular AI chip for real-time inference

Tech · 5 min read

Aurora Labs announced a $120 million Series C round co-led by Horizon Capital and Evergreen Ventures, and simultaneously revealed Helios, a new family of modular AI accelerators built for deterministic, millisecond-scale inference. The company says Helios balances matrix multiplication throughput with an architecture specialized for sparse activation patterns common in transformer and graph models used in robotics and perception workloads.

Helios comes with a full software stack: compiler toolchains that integrate with PyTorch and TensorFlow, RTOS drivers for tight timing guarantees, and a model optimizer that can prune and reflow networks for the hardware. Aurora Labs is positioning Helios for integrators building autonomous vehicles, warehouse robots, and industrial controllers where consistent inference latency is critical.

Investors highlighted Aurora’s unique approach to co-designing hardware and runtime software, and the company plans to use the fresh capital to scale manufacturing partnerships, expand engineering in Europe and Taiwan, and pilot the chip in several robotics fleets during 2027.