Anker's Custom AI Chip Aims to Reshape the Accessory Market
Anker is stepping into the silicon arena with a new processor designed to embed AI directly into its product line. The company's 'Thus' chip represents a distinct architectural choice, employing a...
Anker is stepping into the silicon arena with a new processor designed to embed AI directly into its product line. The company's 'Thus' chip represents a distinct architectural choice, employing a compute-in-memory design for neural network tasks. This approach, Anker claims, allows the chip to perform complex calculations using less power and in a smaller physical footprint than conventional architectures, making it suitable for compact devices like earbuds and IoT accessories.
Steven Yang, Anker's CEO, explained the technical rationale. He contrasted Thus with standard AI chips, where data must constantly shuttle between separate memory and processing units. "Thus puts the computation where the model already lives," Yang stated. "The model never has to move again." This efficiency gain, the company argues, enables a leap in on-device capability.
The first application will be in Soundcore's forthcoming premium earbuds. Anker selected this category due to its extreme engineering constraints: minimal space and strict power budgets. Previous earbud AI was limited to simple neural networks. Anker asserts the Thus chip can manage networks with millions of parameters, a jump intended to dramatically improve functions like voice isolation during calls. The planned earbuds will combine this processor with an array of eight microphones and bone conduction sensors to target the user's voice specifically.
While promising, the chip's real-world performance against established competitors like Apple and Sony remains unproven. Leaks suggest the first Thus-equipped models, likely the Liberty 5 Pro series, will be unveiled at Anker's event on May 21, with pricing expected to start around $170.
Source: The Verge
Ready to Modernize Your Business?
Get your AI automation roadmap in minutes, not months.
Analyze Your Workflows →