Kneron Inc, a San Diego-based AI chip start-up, has unveiled a new energy-efficient alternative to Nvidia’s GPUs. The company’s KL730 AI chip, which utilizes Taiwan Semiconductor Manufacturing Co’s (TSMC) 12-nanometer process technology, aims to provide a cost-saving solution for generative pre-trained transformer (GPT) applications at scale.
Founder and CEO of Kneron, Albert Liu, challenged the notion that only GPUs can handle AI tasks, stating that their NPU (neural processing unit) can perform the same functions as a GPU, but at a much lower cost. Quanta Cloud Technology Inc and Chunghwa Telecom Co have already reported operational cost savings of up to 75% after implementing Kneron’s KL720 NPU in their servers, and Liu believes that the KL730 will deliver even better performance.
In addition to cost savings, the KL730 is touted as being 150 to 200 percent more energy-efficient than products from other industry players. This enhanced energy efficiency makes it an attractive option for automakers in Germany and Japan, who have expressed interest in utilizing the AI chip for full-frame detection, a crucial feature for advanced driver-assistance systems.
Liu envisions autonomous vehicles becoming the primary application for NPUs, followed closely by servers. Notable companies such as Quanta Cloud, Chunghwa Telecom, and South Korea’s Naver Corp already count themselves as Kneron’s server customers. Hanwha Group, a South Korean conglomerate, has also adopted Kneron’s AI chips for use in surveillance cameras.
The KL730 AI chip is scheduled to be available for sampling by customers by the end of this year. It can be used in a variety of applications, including enterprise-edge servers, smart home solutions, and advanced driver-assistance systems.
Kneron, established in 2015, has raised over US$140 million in funding from investors such as Qualcomm Ventures, Hon Hai Precision Industry Co, Horizons Ventures Ltd, and Sequoia Capital. Aside from TSMC, GlobalFoundries Inc and United Microelectronics Corp are also Kneron foundry partners.
With the debut of their energy-efficient KL730 AI chip, Kneron aims to disrupt the market dominated by Nvidia’s GPUs by offering a cost-effective and high-performing alternative. The chip’s potential applications in autonomous vehicles and surveillance cameras, coupled with its impressive cost savings and energy efficiency, make it a compelling choice for businesses and individuals looking to leverage AI technology without breaking the bank.
