Artificial intelligence semiconductor company Furiosa AI (CEO Baek Jun-ho) announced on the 17th that it has officially released its second-generation AI inference accelerator RNGD (Renegade) on Microsoft's Azure Marketplace.

Azure users will be able to easily utilize high-performance AI infrastructure by leveraging Renegade.

RNGD is a next-generation AI accelerator for data centers designed for efficient inference of LLM and multimodal models.

It was first unveiled at the Hot Chips 2024 conference in August last year and attracted great attention from the global tech industry.

It is optimized for various environments such as cloud-centric, on-premises, and hybrid, so it can be flexibly used for various purposes, and provides functions such as △deployment of production environment for inference within minutes, △flexibly adjusting the size of inference infrastructure according to demand, and △seamless integration with existing Azure data and software stacks.

Furthermore, FuriosaAI plans to soon provide a Llama 3.1 pre-compiled model-based API, allowing enterprise customers to immediately test and leverage Renegade’s inference performance within their existing workflows. We also plan to sequentially expand support for various model architectures in the future.

“The Azure Marketplace is a place where customers around the world can easily find and deploy trusted partner solutions,” said Jake Zborowski, general manager, Microsoft Azure Platform. “We’re excited to welcome FuriosaAI’s Renegade to this ecosystem.”

With this launch, Furiosa AI plans to accelerate the distribution of its products. It is currently conducting product evaluations of Renegade for domestic and international corporate customers, and has entered large-scale mass production through collaboration with TSMC.

“Renegade’s Azure integration is a major milestone in realizing our vision of providing an efficient AI inference infrastructure that is accessible to everyone,” said a FuriosaAI representative. “From startups to large enterprises, they will be able to quickly and easily secure the performance they need without complex AI infrastructure.”