Upgraded AI Inference Chip Development Set to Expand with $125 Million Series C Funding
FuriosaAI Secures $125 Million Series C Funding for AI-Focused Semiconductor Solutions
FuriosaAI, a leading semiconductor company specializing in AI-native hardware, has announced the completion of a $125 million Series C bridge funding round. This new funding will propel the company towards its goal of delivering a more power-efficient and cost-effective solution for AI computing.
The RNGD AI accelerator, FuriosaAI's flagship product, offers between 256 and 512 teraFLOPS of floating point performance, depending on precision (16-bit or 8-bit). It boasts 48GB of memory via two HBM3 stacks, with about 1.5TB/s bandwidth. Despite modest raw specs compared to leading GPUs, RNGD excels in power efficiency, consuming about 150-180 watts, compared to Nvidia's H100 GPUs that draw 350W. In AI inference tasks, particularly for large language models (LLMs), RNGD achieves 2.25x to 3x better performance per watt.
LG AI Research has been a significant customer for FuriosaAI, adopting the RNGD AI accelerator to meet their low-latency, high-throughput needs for their EXAONE LLMs, and reducing total cost of ownership dramatically.
Since its founding in 2017, FuriosaAI has raised approximately $115 million across four funding rounds. Notably, the company turned down an $800 million acquisition offer from Meta earlier in 2025, choosing to remain independent to grow their business.
With the new funding, FuriosaAI aims to expand deployment of RNGD accelerators to enterprise customers beyond LG, supporting a wide range of AI models with energy-efficient inference hardware expressly designed for AI workloads rather than general GPU tasks. Their RNGD servers are scalable, with chassis and rack configurations to support data center needs, indicating plans for continued system-level solutions in AI computing.
Two world-class technical leaders, Professor Kang in compiler and software architecture, and Dr. Cho in silicon hardware, have been added to FuriosaAI's team. Jeehoon Kang has been appointed as Chief Research Officer, while Youngjin Cho will serve as Vice President of Hardware. Yoon D. Kang, CEO of PI Partners, believes Furiosa is one of the very few companies with the proven architecture innovation and software maturity to deliver.
FuriosaAI's funding will deliver a new foundation for AI that is powerful and efficient, addressing the issues of high power consumption and costs of legacy GPU hardware in AI computing. As AI today is dependent on a broken business model where the infrastructure costs and constraints of power-hungry GPUs are a critical, rapidly worsening roadblock, FuriosaAI's RNGD accelerator offers a promising solution.
The new funding will help accelerate the mass production of the RNGD AI accelerator and support the development of the next-generation chip. June Paik, Furiosa's co-founder and CEO, has demonstrated the ability to deliver on an audacious vision for transforming AI computing, and with the support of its investors and new hires, FuriosaAI is poised for continued growth and success.
Table:
| Aspect | Details | |------------------------|-----------------------------------------------------------------| | Performance | 256–512 teraFLOPS (FP16/FP8), 48GB HBM3, 1.5TB/s BW | | Power Efficiency | 2.25x–3x better perf/watt vs Nvidia H100 (150–180W vs 350W TDP) | | Key Customer | LG AI Research (EXAONE LLMs) | | Funding | $115 million raised, turned down $800M Meta acquisition offer | | Future Plans | Enterprise RNGD deployment, AI-only architecture, expand scale | | New Hires | Professor Kang (Compiler and Software Architecture), Dr. Cho (Silicon Hardware) | | Investors | PI Partners, Korea Development Bank, Kakao Investment | | Valuation | Currently valued at $735 million |
- FuriosaAI, with the newly secured $125 million Series C funding, plans to use the investment to broaden the deployment of their RNGD AI accelerators to enterprise customers, focusing on delivering energy-efficient inference hardware designed specifically for AI workloads, rather than general GPU tasks.
- In the realm of investing in technology, FuriosaAI's RNGD AI accelerator stands out as a promising solution for the finance sector, as it addresses the issues of high power consumption and costs of legacy GPU hardware in AI computing, leveraging AI-focused semiconductor solutions to offer better performance per watt.
