Anker is stepping into custom silicon with a new chip designed to embed AI processing directly into audio devices and Internet of Things gadgets. The company claims its Thus processor is the first neural-net compute-in-memory AI audio chip, a design that collapses model storage and computation into a single, low-power component.
The move signals Anker’s ambition to reduce reliance on cloud-based AI, a trend gaining traction as privacy and latency concerns push more developers toward on-device intelligence. With Thus, Anker aims to bring voice assistants, real-time audio enhancement, and adaptive device behaviors without forcing users to send data to remote servers.
At the heart of the innovation is a fundamental shift in chip architecture. Traditional AI chips separate model storage from computation, shuttling data back and forth thousands of times per second. Anker’s CEO Steven Yang highlighted this inefficiency in a statement, noting that existing designs “carry all those parameters across many times per second.” In contrast, Thus processes computations where the data resides, cutting energy use and latency while shrinking physical footprint—critical for compact devices like earbuds and smart home sensors.
Early benchmarks suggest the chip delivers up to 50% lower power consumption during inference tasks compared to conventional solutions, according to internal testing shared by Anker. The company also positions Thus as highly scalable across product lines, from premium active noise-cancelling headphones to budget-friendly IoT hubs. Production is slated to begin this year, with devices featuring the chip expected to hit shelves in 2025.
Industry watchers see potential ripple effects beyond Anker’s ecosystem. Chipmakers like Qualcomm and Nvidia dominate the AI silicon market, but custom designs from consumer brands could accelerate the shift toward specialized, power-efficient solutions. Analysts point to Apple’s Neural Engine and Google’s Tensor chips as precedents, though Anker’s focus on audio and IoT sets it apart.
For developers, Thus offers an SDK to integrate its neural models into device firmware. The kit includes pre-trained models for voice activity detection and environmental sound classification, alongside tools to fine-tune networks for custom use cases. Early partners in Anker’s ecosystem include Soundcore, its audio brand, and Eufy, its smart home division.
While the chip’s launch is still months away, the announcement underscores a broader industry push: making AI ubiquitous but invisible. By embedding intelligence at the hardware level, Anker could help define the next wave of smarter, more responsive gadgets.
The question now is whether consumers will notice—or care—as AI transitions from cloud-dependent novelty to silent, on-device enabler.
AI summary
Anker’s Thus chip eliminates cloud dependency for AI in audio and IoT devices, cutting power use by up to 50% while enabling on-device voice and context processing.
