Untether Launches AI Chip for Diverse Applications

By Max A. Cherney
SAN FRANCISCO (Reuters) – Startup Untether launched an artificial intelligence chip on Monday aimed at powering AI applications in cars, agricultural devices, and other edge scenarios.

The flagship AI chips from Nvidia (NASDAQ:NVDA) and Advanced Micro Devices (NASDAQ:AMD) are designed for data centers that require large amounts of energy and additional equipment. These installations are typically focused on building or training large language models, such as those powering ChatGPT.

Untether, based in Toronto, claims its chips can effectively run AI applications once a model is built, achieving greater efficiency.

According to Untether's Vice President of Product, Bob Beachler, the market for chips specifically running AI applications, known as "inference," is expected to surpass the training market by 2027, potentially growing to $102 billion.

Beachler stated, "Where that inference is going to be deployed is everywhere."

Untether's chip, named the 240 Slim, promises substantial performance with significantly lower energy consumption, making it suitable for autonomous vehicles, agricultural equipment, and some data centers.

Notably, Mercedes-Benz (OTC:MBGAF) is collaborating with Untether to develop next-generation autonomous vehicles.

The chips utilize open-source RISC-V technology, which competes with designs from Arm Holdings (NASDAQ:ARM).

In a recent peer-reviewed benchmark by MLCommons, Untether's chips showcased strong performance for their intended applications.