Untether AI has raised $13 Million in Series-A financing led by Intel Capital. Intel Capital is Intel’s VC division. The Toronto-based startup founded in 2018 has developed a new microarchitecture for neural net interfaces. This new technology is slated to be 1000 times faster than any current AI chip while being cheaper and eliminating bottlenecks in data.


Now is the time to build a new kind of ultra-efficient chip for new frontiers in AI applications — from autonomous vehicles, data centers, mobile, vision processing and beyond — that will ultimately change how everything works.


Martin Snelgrove, CEO, Untether AI

The team has already created a working prototype that supports their claims. An actual chip is, however, much smaller than the prototype and the performance should be affected under real-time loads and working conditions. Face and Voice recognition requires an increasing number of logical operations. Untether deals with this problem by using near-memory-computing incorporating 2.5D and 3D-IC technologies. This results in increased efficiency and better performance-per-watt.

Deep Learning Training and Interference comparison

The Difference Between Deep Learning Training and Inference. Source: NVIDIA

Untether is working on an Inference chip, which is a smaller version of conventional AI chips and is similar to those used in devices like phones and tablets. The unconventional technologies employed by Untether and new advancements in AI could make Intel the harbinger of AI in mainstream computing solutions.

You can read the full press release here

Further Reading:

Leave a Reply