Instead of selling to Meta, AI chip startup Furiosaai signed a huge customer

Korean AI chip startup You were angry On Tuesday, they announced a partnership offering AI chips. rngdFor the company that uses it LG AI ResearchThe recently announced EXAONE platform.

RNGD is optimized to run large-scale language models (LLM) Last week, South Korean technology giant LG announced it The next generation Hybrid AI model EXAONE 4.0. The collaboration covers key sectors such as electronics, finance, communications and biotechnology for a variety of applications.

This news comes about three months after Furiosaai Denied Meta’s $800 million acquisition offerchoose to stay independent.

According to local media, the transaction failed due to differences in opinion regarding business strategies and organizational structure after acquisition, not due to price issues. Meta’s interest in acquiring AI Chipmakers like Furiosaai reflects a broader strategy to reduce reliance on third-party suppliers such as Nvidia.

When asked why the Meta contract fell, Furiosaai June Paik CEO told TechCrunch, “I think it’s also an exciting opportunity. I think it’s a very impactful contribution, both personally and for the company, to make AI computing more sustainable.”

Paik refused to specify whether the startup is currently pursuing fresh funds as M&A (at least from the meta) went off the table.

However, Paik says the new partnership will lead to business possibilities that go far beyond Korea.

“LG AI’s EXAONE is considered the main sovereign AI model in Korea. It will not be used only by LG. It will become one of the main AI models used in the Korean AI ecosystem. By these customers, including global customers,” Pike said.

LG AI’s decision to employ Furiosa’s AI chip and accelerator is worth noting for another reason. This is one of the few public supporters of Nvidia’s rivals by major companies. One of the main reasons for victory is the low hardware costs of startups.

“We had to prove that the solution not only provided strong performance, but also lowered the total cost of ownership,” Paik said.

Furiosaai claims that the RNGD accelerator outperforms the competitive GPU on LG AI Research’s EXAONE model, providing 2.25 times better inference performance. According to Paik, LG has discovered that Furiosaai hardware is more energy efficient.

Furiosa’s chips were built specifically for AI rather than for a typical GPU. “It can efficiently support a variety of AI models. However, unlike GPUs, which are still fundamentally general purpose processors, the architecture is built natively for AI computing.

The Seoul-based startup, which also operates offices in Santa Clara, has a team of just 15 employees to just 15 people.