DeepMind releases Perception API and scores $250M in internal investment for embodied AI

AI · 6 min read

The Perception API aggregates object recognition, scene segmentation, audio event detection, and spatial mapping into a single inference endpoint optimized for low latency. DeepMind emphasizes the API’s compatibility with AR headsets and robotics SDKs.

The internal $250 million commitment will expand model training datasets, fund safety research for embodied agents, and subsidize cloud credits for early partners in robotics and AR. DeepMind highlights use cases in assistive robotics, industrial inspection, and immersive AR experiences.

Developers welcomed an integrated perception offering but sought pricing clarity and guarantees around latency and privacy. DeepMind responded with an enterprise SLA, edge inference options, and data-handling schemas for sensitive deployments.