FuriosaAI Rejects Meta's $800M Offer, Partners with LG AI Research Over RNGD Chip's 225% Inference Edge

South Korean chip startup FuriosaAI Inc. has rejected a $800 million acquisition offer from Meta Platforms Inc., opting instead to forge a strategic partnership with LG AI Research. The collaboration involves supplying FuriosaAI’s RNGD AI accelerator chips to power LG’s Exaone language models, a move that positions the startup as a key player in South Korea’s AI infrastructure landscape. The RNGD chip, designed specifically for large language model (LLM) workloads, promises 2.25 times the inference performance of competing GPUs while emphasizing energy efficiency and cost-effectiveness. FuriosaAI CEO June Paik highlighted the RNGD’s commercial readiness, stating the partnership signals growing demand for specialized AI hardware.

The decision to decline Meta’s offer, which collapsed due to disagreements over post-acquisition business strategies and organizational structures, underscores FuriosaAI’s commitment to maintaining independence. Paik emphasized the company’s focus on sustainable AI innovation, aligning with LG’s development of Exaone 4.0, South Korea’s first hybrid-reasoning AI model. The RNGD chip’s architecture, optimized exclusively for AI computing, differentiates it from general-purpose GPUs, which are often used for tasks like gaming and cryptocurrency mining. This specialization, combined with performance metrics, challenges the dominance of established players such as Nvidia and emerging competitors like Groq and Cerebras.

Under the partnership, LG will integrate FuriosaAI’s RNGD-powered servers into industries including electronics, finance, and biotechnology. The chips will also support LG’s internal AI agent, ChatExaone. Paik noted that LG’s selection of FuriosaAI’s hardware stems from its ability to deliver superior inference performance while reducing operational costs. The startup, which currently employs 15 staff, aims to expand its client base in the U.S., Middle East, and Southeast Asia, with plans to secure new partnerships in the second half of this year. FuriosaAI is simultaneously pursuing a $48 million funding round, expected to conclude by month’s end, to fuel its growth strategy.

The deal reflects broader industry trends as tech firms seek alternatives to Nvidia’s AI hardware. Meta, like other companies developing large language models, is investing heavily to reduce reliance on third-party suppliers. The tech giant has allocated $65 billion for AI initiatives in 2025, part of a global shift toward in-house AI infrastructure. By choosing FuriosaAI, LG reinforces its position as a regional leader in AI innovation, leveraging the RNGD’s capabilities to enhance Exaone 4.0’s performance and scalability. The collaboration also validates the growing appetite for purpose-built AI accelerators, particularly as energy efficiency and total cost of ownership become critical factors in hardware adoption.

FuriosaAI’s pivot from acquisition to partnership highlights its ambition to remain a standalone innovator in the AI chip market. The RNGD’s deployment with LG’s Exaone 4.0 serves as a benchmark for other firms evaluating non-Nvidia solutions, potentially accelerating the adoption of specialized AI hardware. While Meta’s offer represented a significant financial opportunity, FuriosaAI’s leadership prioritized long-term strategic alignment with LG over immediate consolidation. This decision aligns with industry dynamics where technological sovereignty and sustainability are increasingly valued. As the startup scales its international presence, the success of the LG partnership will be a key indicator of its ability to compete in a rapidly evolving AI infrastructure market.