FuriosaAI announced a strategic partnership with Broadcom Inc. (NASDAQ: AVGO) to develop its third-generation AI accelerator chip. The collaboration will combine FuriosaAI's Tensor Contraction Processor architecture with Broadcom's networking solutions and Ethernet switches to create an inference platform for AI compute clusters.

The partnership builds on FuriosaAI's RNGD chip, which is currently in mass production using TSMC's 5nm process. RNGD is a 180W PCIe-based accelerator that has been deployed by Samsung SDS and LG AI Research for large language model workloads in data center environments.

The third-generation chip will feature a 2nm compute die and HBM4/4E memory, utilizing Broadcom's packaging capabilities to integrate multiple silicon dies. The system will incorporate Broadcom's Ethernet and PCIe technologies for rack-scale networking across AI compute clusters.

"Inference performance is no longer defined solely by raw compute. It is increasingly a function of data reuse and communication efficiency across servers and racks," said Charlie Kawwas, president of Broadcom's Semiconductor Solutions Group.

FuriosaAI CEO June Paik stated that the partnership will deliver "a third-generation inference solution that offers industry-leading performance per watt for even the largest, most complex frontier AI models and agentic workloads."

The company's software stack includes a compiler that automatically maps PyTorch code to silicon and a Virtual ISA programming model for developers requiring hardware control. Sampling of the new chip is scheduled to begin in the first half of 2028.

FuriosaAI, founded in 2017 by engineers from AMD, Qualcomm, and Samsung, has raised $250 million through Series C Bridge funding and operates offices in Korea and Silicon Valley.