South Korean artificial intelligence (AI) chip company FuriosaAI has joined hands with global semiconductor giant Broadcom to develop next-generation AI inference chips. The partnership is seen as an effort by both companies to tap into a new market through technological cooperation, as demand for inference infrastructure rapidly grows in the AI accelerator market dominated by Nvidia.
FuriosaAI announced on the 28th that it has signed a strategic partnership with Broadcom to jointly develop next-generation AI accelerators. The collaboration goes beyond simple joint chip development, focusing on building a next-generation AI infrastructure platform that integrates AI computing, networking, and software.
FuriosaAI's third-generation AI accelerator under development will adopt a Tensor Contraction Processor (TCP) architecture along with HBM4 and HBM4E memory. Broadcom will incorporate its advanced packaging technology into the design. Sampling of the third-generation accelerator is scheduled to begin in the first half of 2028. Through this, the two companies plan to build a hyperscale AI inference platform that differentiates itself from the existing graphics processing unit (GPU)-centric AI infrastructure established by Nvidia.
"By combining FuriosaAI's and Broadcom's technologies, we will build a platform that addresses key bottlenecks in large-scale agentic AI environments," said Charlie Kawwas, President of Broadcom's Semiconductor Solutions Group. June Paik, CEO of FuriosaAI, said, "With our next-generation products, we will deliver industry-leading performance per watt even in ultra-large AI models and hyperscale agentic AI environments."
