Key insights
- FuriosaAI partners with Broadcom to develop a third-generation AI accelerator chip for data centers. The collaboration leverages FuriosaAI's architecture and Broadcom's networking technology. The new chip will feature a 2nm compute die and HBM4 memory. While positive for long-term AI infrastructure growth, the immediate impact on US equities is limited.
MOUNTAIN VIEW, Calif. - FuriosaAI announced a partnership with Broadcom (NASDAQ:AVGO) to develop its third-generation AI accelerator chip, according to a press release statement. The collaboration brings together FuriosaAI’s specialized AI architecture with Broadcom, a prominent player in the Semiconductors & Semiconductor Equipment industry with a market capitalization of $2 trillion.
The collaboration will combine FuriosaAI’s Tensor Contraction Processor architecture with Broadcom’s networking technology to create a multi-die chiplet system designed for AI inference workloads in data centers.
FuriosaAI’s current product, RNGD, is a 180-watt, PCIe-based accelerator manufactured using TSMC’s 5nm process. The chip is in mass production and has been deployed by Samsung SDS and LG AI Research.
"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. Broadcom’s financial strength supports such strategic partnerships, with revenue growing 25% over the last twelve months to $68.3 billion and gross profit margins of 77%. The stock has delivered an 80% return over the past year, though InvestingPro analysis suggests it may be overvalued at current levels. For investors tracking semiconductor valuations, InvestingPro’s Most Overvalued list provides comprehensive screening tools.
The third-generation chip will feature a 2nm compute die and HBM4/4E memory, using Broadcom’s packaging technology to integrate multiple silicon dies. The system will incorporate Broadcom’s Ethernet and PCIe technologies for networking across AI compute clusters.
"Having proven the performance and efficiency of our architecture with RNGD, our second-generation chip now in mass production with TSMC, we will deliver a third-generation inference solution," said June Paik, FuriosaAI cofounder and CEO.
FuriosaAI’s software stack includes a compiler that maps PyTorch code to its hardware and a Virtual ISA programming model for developers requiring more control.
Sampling of the third-generation chip is scheduled to begin in the first half of 2028.
FuriosaAI was founded in 2017 and has raised $250 million through Series C Bridge funding. The company operates offices in Korea and Silicon Valley.
In other recent news, Broadcom Inc. announced the launch of three new Wi-Fi 8 system-on-chip devices aimed at enhancing Ethernet routers and mesh networks. These chips, named BCM6772, BCM6774, and BCM6776, are designed to reduce component count and power consumption. Additionally, Broadcom introduced the BCM68850, a 50G ITU-PON home gateway chip with AI capabilities, featuring symmetric 50G throughput and advanced security algorithms. Evercore maintained an Outperform rating on Broadcom, citing strong demand in AI networking despite supply constraints. UBS recently raised its price target for Broadcom to $490, influenced by a revised Anthropic order that promises higher margins. TD Cowen also increased its price target to $500, highlighting Broadcom’s potential to meet or exceed a 30% growth target in the upcoming quarter. These developments underscore Broadcom’s strategic advancements in technology and market positioning.
This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.