South Korean AI chipmaker FuriosaAI has turned down an $800 million acquisition offer from Meta, opting to continue developing its RNGD AI accelerator rather than becoming part of the social media giant’s AI hardware division.
The decision reflects FuriosaAI’s confidence in its technology and its strategic plan to expand independently. According to a report by Bloomberg, negotiations broke down due to disagreements over post-acquisition business strategy and structure, rather than price.
Why Meta Wanted FuriosaAI
Meta has been actively developing custom AI chips to reduce its reliance on Nvidia, which dominates the AI accelerator market. As part of this strategy, the company recently started testing its Meta Training and Inference Accelerator (MTIA), a proprietary AI chip designed to support its Llama AI models.
However, Meta’s previous AI hardware projects have struggled to compete with Nvidia’s cutting-edge GPUs, prompting an aggressive push for in-house semiconductor expertise.
NVIDIA CEO Jensen Huang just introduced the company’s latest AI infrastructure product, the Blackwell Ultra AI Factory Platform, designed to significantly enhance AI reasoning, agentic AI, and physical AI workloads.
Acquiring FuriosaAI would have given Meta immediate access to an established AI accelerator design, allowing it to compete with Nvidia, AMD, and emerging players like Groq.
However, FuriosaAI’s leadership saw Meta’s interest as a potential constraint on its long-term plans. A source cited by a Korean media outlet stated that “disagreements over post-acquisition business strategy and organizational structure, rather than price issues, caused the negotiations to break down.”
FuriosaAI’s Competitive Edge in AI Chips
By rejecting Meta’s offer, FuriosaAI is betting that its RNGD (Renegade) chip can carve out a niche in the AI hardware market. The RNGD chip, set to launch later this year, is designed for high-performance AI inference workloads and is already being tested by LG AI Research for integration into its enterprise AI systems.
FuriosaAI claims the RNGD chip delivers three times better performance per watt than Nvidia’s H100 GPU. This is largely due to its Tensor Contraction Processor architecture, which optimizes large-scale matrix computations. The chip features:
- 48GB of HBM3 memory with 1.5TB/s memory bandwidth
- 150W thermal design power (TDP), significantly lower than the H100’s 350W TDP
While FuriosaAI has provided these internal benchmarks, independent performance tests are still pending, and industry analysts caution that real-world deployment will be the ultimate test of its efficiency.
Funding Instead of Acquisition
Rather than selling to Meta, FuriosaAI is securing additional funding to scale production and expand its global footprint. The company is close to finalizing a Series C funding round, which is expected to exceed its initial target.
FuriosaAI has already raised approximately $115 million across previous investment rounds and is currently seeking an additional $48 million (70 billion won) to fuel its expansion.
The startup is also gaining traction with major corporate clients. Saudi Aramco has received sample RNGD chips for evaluation, indicating potential adoption within the energy sector’s AI-driven analytics and automation workflows.
How FuriosaAI Fits into the AI Chip Wars
The AI chip industry is in the midst of a battle for dominance. Tech giants like Meta, Microsoft, and OpenAI are investing heavily in custom AI hardware to avoid relying on Nvidia’s GPUs. OpenAI, for instance, has partnered with TSMC to develop its own AI chips, expected to enter production in 2026.
Nvidia is countering this competition with massive investments in domestic semiconductor production. Nvidia is pouring hundreds of billions of dollars into expanding U.S. chip manufacturing, ensuring it retains leadership in AI accelerator technology.
Meanwhile, alternative players like Groq have secured $1.5 billion in funding from Saudi Arabia to develop specialized AI inference chips, presenting additional competition in the AI accelerator market.
By rejecting Meta’s offer, FuriosaAI is positioning itself as an independent competitor in this evolving market, betting that it can survive and thrive without merging with a big tech firm. However, its success will ultimately depend on the real-world performance of its RNGD chip and whether its independence will translate into long-term market viability.
