Ambi Raises $26M to Build Always-On AI Assistant That Listens Continuously

Former Plaud executive Peter Mo has closed $26 million in funding for Ambi, an AI application designed to listen continuously across iPhone, Mac, and Apple Watch to deliver real-time, context-aware assistance. The round was led by Luminous Ventures — an investor in Meituan and Unitree Robotics — alongside original design manufacturer Huaqin. Unlike conventional meeting note-takers that process audio after a call ends, Ambi’s agent aims to surface relevant knowledge, suggest actions, and initiate tasks while a conversation is still underway, drawing on calendar data, prior meetings, personal knowledge bases, and web sources.

The architecture implies a shift toward persistent, low-latency inference workloads. Continuous audio capture, on-device transcription, and real-time retrieval-augmented generation demand efficient model serving pipelines — whether on edge silicon or in the cloud. Ambi’s roadmap includes a desktop app that observes screen activity to build long-term memory, and a dedicated hardware device equipped with microphones, camera, speaker, and display. Each layer adds compute intensity, from always-on keyword spotting to multimodal fusion.

Why it matters for GPU / AI infrastructure

Always-on assistants like Ambi represent a growing class of workloads that blur the line between edge and cloud inference. They require sustained token throughput, tight latency budgets, and privacy-preserving data flows — all of which drive demand for heterogeneous GPU fleets, confidential computing enclaves, and optimized model quantization. For infrastructure providers, the signal is clear: the next wave of consumer AI will not be batch-oriented; it will be streaming, stateful, and personal.

Privacy remains a focal point. Ambi states ambient capture is off by default and user-controllable, yet the category faces scrutiny over third-party consent and on-device indicators. Apple’s own Voice Intelligence rollout has amplified the debate. As these products mature, infrastructure that supports local-first processing with optional cloud burst — backed by verifiable audit trails — will become a competitive differentiator.