South Korean AI chip startup FuriosaAI (19459030) announced a partnership to supply its AI chip on Tuesday. RNGD – to enterprises using The recently launched EXAONE platform by LG AI Research.
RNGD has been optimized for large language models. Just last week, the Korean tech titan LGunveiled its next-generation hybrid AI model EXAONE 4.0. The partnership targets key sectors including electronics, finance and telecommunications for a variety of applications.
This announcement comes about three months after FuriosaAI rejected Meta’s $800,000,000 acquisition offer and chose to remain independent. According to local media, the deal failed due to disagreements about post-acquisition strategy and organizational structure rather than price issues. Meta’s interest to acquire AI chipmakers such as FuriosaAI is part of its strategy to reduce its dependence on third-party vendors, such Nvidia.
When TechCrunch asked June Paik, CEO of FuriosaAI, why the deal fell through, she replied: “We want our mission to continue, and I believe it’s a great opportunity at the same. I think it’s a great contribution, both personally for the company and for me, to make AI computing sustainable.
With M&A off the table (at least with Meta), Paik declined to say if the startup was now looking to raise new funding.
Techcrunch event
San FranciscoOctober 27-29, 2025
“LG AI’s EXAONE is regarded as the leading sovereign AI model in South Korea. It will not be used only within LG. It will be used as a main AI model in the Korean AI ecosystem. We expect that there will be a lot of demand for this EXAONE as well as our chip solutions, not just in Korea but also in South Korea. The LG team also partners with global customers and does business with them. We also expect that these customers, including those from around the world, will use this,” Paik said.
LG AI’s decision to adopt Furiosa’s AI chip is notable for another factor: It’s a rare public endorsement of a competitor to Nvidia from a major company, Paik said. The startup’s hardware was cheaper, which is one of the main reasons for its success.
Paik said, “We had prove that our solution delivers not only strong performance but lowers the total cost of ownership.”
FuriosaAI claims its RNGD acceleration outperforms competitive GPUs, including LG AI Research’s EXAONE model, delivering 2,25 times better performance in inference. Paik says that LG also found the FuriosaAI to be more energy efficient.
Furiosa’s chip was not designed as a general GPU, but exclusively for AI. “We are able to support a variety of AI models with efficiency. Our architecture is built specifically for AI computing, unlike GPUs which are still primarily general-purpose processors. Paik stated that the chip was not designed for mining or rendering.
A global team of only 15 employees makes up the Seoul-based startup that also has an office in Santa Clara.
Kate Park, a reporter for TechCrunch with a focus in Asia on technology, startups, and venture capital, is a reporter. She was previously a financial reporter at Mergermarket, covering M&As, private equity and venture capitalism.
View Bio
