Nvidia Challenger FuriosaAI Secures First Major Customer: LG AI Research.
FuriosaAI Inc., the Seoul-based startup aiming to disrupt Nvidia Corp.'s dominance in AI chips, has landed its first significant contract with LG AI Research. This landmark deal comes just months after FuriosaAI reportedly turned down an $800 million acquisition offer from Meta Platforms Inc., Bloomberg reported, July 22, 2025 underscoring the growing competition in the AI chip market.
Adding to the recent financial activity surrounding the industry leader, Nvidia CEO Jensen Huang offloaded an additional $12.94 million worth of shares on Friday, July 18th, as part of a pre-arranged trading plan. (Huang remains among the world's wealthiest individuals, currently holding a real-time net worth of over $146 billion, according to Finance Monthly.)
The breakthrough follows a rigorous seven-month evaluation by LG AI Research, which ultimately granted final approval for FuriosaAI's AI chip, RNGD (pronounced “Renegade”). LG will integrate the RNGD chip into its Exaone large-language models, marking a crucial step for the burgeoning Korean chip designer.
Back in February 2025, Meta was in talks to buy the Korean AI Chip Startup founded by June Paik (a former Samsung Engineer) according to Forbes.
"For the last eight years, we worked very hard from R&D to product phases and finally this commercialization phase," said June Paik, CEO of FuriosaAI, to Bloomberg News. "This signals that our product is ready for enterprise adoption."
LG's validation of the RNGD chip is a major boost for FuriosaAI, positioning it as a key player among Korean chip designers vying to capitalize on the post-ChatGPT boom in AI infrastructure. The RNGD chip was designed not only to compete with industry leader Nvidia but also with other promising startups like Groq Inc., SambaNova Systems Inc., and Cerebras Systems Inc.
Founded in 2017 by Paik, an alumnus of Samsung Electronics Co. and Advanced Micro Devices Inc., FuriosaAI specializes in semiconductors for AI inferencing and services. The company boasts that its RNGD chip delivers 2.25 times better inference performance per watt compared to traditional graphics processing units.
FuriosaAI is part of a growing wave of Korean semiconductor innovators, including Rebellions Inc. and Semifive Inc., that are leveraging the nation's robust ecosystem of talent, suppliers, and government incentives, which has flourished around giants like Samsung and SK Hynix Inc.
Under their new partnership, FuriosaAI and LG plan to deploy RNGD servers powered by Exaone across diverse sectors, from electronics to finance. The collaboration will also extend to LG’s in-house enterprise AI agent, ChatExaone, with future plans for external client expansion.
Looking ahead, FuriosaAI is actively pursuing new customers in the US, the Middle East, and Southeast Asia, with expectations to finalize additional agreements in the latter half of this year. The startup garnered significant public attention in March when news broke of its decision to maintain independence by rejecting Meta’s acquisition overtures. FuriosaAI reportedly intends to raise further capital before eventually pursuing an initial public offering.
FuriosaAI Frequently Asked Questions (FAQ)
Q1: What is FuriosaAI and why is it considered a challenger to Nvidia?
A1: FuriosaAI Inc. is a Seoul-based startup that designs AI chips. It's considered a challenger to Nvidia because its RNGD chip aims to compete in performance and efficiency with Nvidia's dominant GPUs, particularly for AI inferencing workloads. The company claims 2.25 times better inference performance per watt compared to GPUs.
Q2: What is "AI inferencing" and why is it important?
A2: AI inferencing is the process where a trained AI model (like a large language model) makes predictions or draws conclusions from new, real-world data. Unlike AI training which is about teaching the model, inferencing is about putting that learned knowledge to use. It's crucial for real-time AI applications such as powering chatbots, autonomous vehicles, or recommendation systems, as it determines how quickly and efficiently an AI can respond.
Q3: What is LG's Exaone large-language model?
A3: Exaone is LG AI Research's large-language model (LLM). It's a powerful AI model designed to understand and generate human-like text, and can be applied across various industries, from electronics to finance, and for internal enterprise AI agents like ChatExaone. LG has been continuously developing Exaone, with versions like Exaone Deep (reasoning AI) and Exaone 4.0 (hybrid AI) released.
Q4: What are FuriosaAI's plans for funding and an IPO?
A4: FuriosaAI recently rejected an $800 million acquisition offer from Meta Platforms Inc., opting to remain independent. The company plans to raise additional capital through a new funding round before eventually pursuing an initial public offering (IPO). This strategy indicates their confidence in long-term independent growth in the AI chip market.
Q5: What is the significance of this deal for the Korean semiconductor industry?
A5: This contract with LG AI Research is a major validation for FuriosaAI and the broader Korean AI chip design ecosystem. It demonstrates that Korean startups can develop competitive AI hardware to challenge global leaders like Nvidia, leveraging the strong domestic semiconductor talent pool, supply chains, and government incentives that have supported giants like Samsung and SK Hynix. This positions South Korea as an emerging hub in the global AI hardware race.
Q6. Who is June Paik, the CEO of FuriosaAI, and what is his background?
The CEO of FuriosaAI is June Paik, who also founded the company in 2017. He specializes in AI accelerator design for data centers, with operations in Santa Clara and Seoul. With over 20 years of experience in both academia and the industry, June worked on the design and engineering of CPUs, GPUs, and memory systems at AMD and Samsung. He holds a master’s degree in Electrical Engineering from the Georgia Institute of Technology. At FuriosaAI, June leads the company in innovating its core AI accelerator product and technology, aiming to enhance data center performance and efficiency.
