In a strategic partnership announced this week, South Korean AI chipmaker FuriosaAI will supply its high-performance RNGD accelerator to enterprises leveraging LG AI Research’s EXAONE 4.0—the latest evolution of LG’s sovereign hybrid large language model (LLM) platform.
FuriosaAI’s RNGD chip is purpose-built for AI workloads, optimized specifically for LLMs. This collaboration is expected to expand into multiple verticals including electronics, finance, telecommunications, and biotech—signaling a major push to position South Korea as a hub of AI innovation.
LG Picks a Homegrown Nvidia Alternative
The alliance is a notable shift in a market traditionally dominated by Nvidia. FuriosaAI’s CEO, June Paik, highlighted that LG’s public endorsement marks one of the few times a major corporation has chosen a domestic alternative over a U.S.-based GPU titan. The decision hinged not just on performance—but on value.
“We had to prove that our solution not only delivers strong performance but also lowers total cost of ownership,” said Paik in an interview with TechCrunch.
According to internal benchmarks, FuriosaAI’s RNGD chip delivered 2.25 times faster inference speeds when running LG’s EXAONE 4.0 model compared to conventional GPUs, while also offering superior energy efficiency.
A Missed Meta Merger and a New Global Focus
The announcement comes just months after FuriosaAI walked away from an $800 million acquisition offer from Meta. The proposed deal reportedly collapsed due to disagreements over business strategy and internal structure—not valuation. Meta’s attempt was part of a larger initiative to reduce dependence on Nvidia and vertically integrate its AI infrastructure.
“We want to continue our mission… It’s a meaningful opportunity to contribute to more sustainable AI computing,” said Paik.
Now, with M&A talks off the table, FuriosaAI is sharpening its focus on global expansion through strategic partnerships. While Paik did not confirm if the company is currently seeking new funding, he emphasized the international scope of the LG collaboration.
“EXAONE isn’t just for LG. It’s intended to serve the broader Korean AI ecosystem—and by extension, global enterprise users,” he said. “This is an opportunity to show that South Korea can offer not just sovereign AI models, but sovereign AI infrastructure.”
Built for AI, Not for Everything
Unlike traditional GPUs that handle a variety of computing tasks—including gaming and rendering—FuriosaAI’s RNGD chip is laser-focused on AI applications. Its architecture is designed from the ground up for deep learning and inference workloads.
“We don’t develop chips for rendering or mining,” said Paik. “Our architecture is natively optimized for AI computing and supports a wide range of AI models with high efficiency.”
Headquartered in Seoul with a satellite office in Santa Clara, California, FuriosaAI operates with a lean team of just 15 employees—underscoring the power of deep tech innovation on a global stage.
