We all know that AI has permeated every aspect of our lives, from smartphones to self-driving cars. And the AI chip is the brain of these smart devices. It is specifically designed for AI tasks and is able to process large amounts of data efficiently, allowing machines to think and learn like humans. Recently, the startup Untether launched a new artificial intelligence chip called the speedAI240 Slim AI inference acceleration card. The chip is specifically designed to power AI applications in automobiles, agricultural equipment, and other specific scenarios. Unlike Nvidia and AMD's flagship AI chips, which are often used in data centers in connection with thousands of other chips to build and train large language models, which requires a lot of energy and additional equipment. Toronto-based Untether says its chips are designed to run AI applications efficiently once models are built.
Untether's new chip is based on open-source RISC-V technology, a technology that has attracted much attention because it is more flexible, open, and has the potential to break the monopoly of traditional chip giants. Untether's chips claim to be high-performance but consume very little energy. This means that it not only accelerates AI adoption, but also extends the life of the device.
What's more, Untether is a chip that can be used in multiple markets, including automotive, agriculture, and machine vision. Chris Walker, CEO of Untether AI, emphasized that the true potential of AI extends beyond the data center to the cars we drive and the fields where food is produced. These environments require very different approaches at the hardware and software levels.
As AI moves from training to inference, the importance and unique needs of edge acceleration become clear. Edge AI applications cannot tolerate the high latency, high capital costs, and non-determinism of cloud-based AI services—they need to meet very different size, power, and operational cost requirements.
Picture: Startup Untether unveils AI chip (Source: Digital Watch Observatory)
The speedAI240 Slim accelerator card offers the best performance and reduced power consumption solution with its low-profile, 75-watt thermal design power (TDP) PCIe design. In the MLPerf inference benchmark, the speedAI240 Slim accelerator card was considered to achieve the world's lowest latency and highest throughput. The chip's design features make it a market leader with powerful hardware support for AI inference.
The imAIgine Software Development Kit (SDK) from Untether AI provides a one-click process that simplifies the process of transforming a trained neural network model into an optimized, inference-ready model to run on the speedAI acceleration solution. This software support greatly improves development convenience and accelerates the deployment of AI applications.
Bob Beachler, VP of Product at Untether AI, expects the market for inference-focused AI chips to reach $102 billion by 2027, far surpassing the training space. In addition, Mercedes-Benz plans to use Untether's chips in its future autonomous vehicles, which shows that Untether's chips have gained some acceptance in the automotive field.
Untether AI's speedAI240 Slim AI inference accelerator card is not only a technological breakthrough, but also a redefinition of the AI hardware market. With the continuous advancement of AI technology and the continuous expansion of application fields, we can foresee that this chip will play an important role in autonomous driving, agricultural technology, and other fields, leading a new era of edge computing.
With the addition of Untether AI, the global AI chip market will become more competitive, and at the same time, it will also bring infinite possibilities for the future development of AI technology.
