TLDR

  • FuriosaAI has initiated the commercial production of its new RNGD AI inference chip.
  • The company asserts that RNGD delivers competitive performance while consuming less power compared to Nvidia’s GPUs.
  • OpenAI presented the chip in Seoul, and LG reported robust real – world testing outcomes.
  • CEO June Paik established FuriosaAI after leaving Samsung to pursue AI innovation.
  • Following its latest funding round, the company is currently valued at nearly $700 million.

FuriosaAI, a South Korean semiconductor company, commences the commercial production of its latest inference chip this month, encroaching on Nvidia’s domain. The RNGD chip claims to offer strong performance with lower power consumption, targeting a market led by Nvidia. OpenAI, LG, and Meta have all shown interest as FuriosaAI positions itself against Nvidia’s dominance.

FuriosaAI Launches RNGD to Rival Nvidia

FuriosaAI was launched in 2017 under the leadership of CEO June Paik, a former Samsung memory – chip engineer and AMD GPU designer. The company is now gearing up to mass – produce its RNGD chip, concentrating on AI inference workloads.

OpenAI recently demonstrated RNGD at an event in Seoul, where LG also reported “excellent real – world performance” during its testing phase. RNGD, which stands for “renegade,” supports high – performance inference with enhanced energy efficiency, a crucial challenge in AI hardware.

Nvidia dominates the AI hardware market, especially for training models, but FuriosaAI sees an opportunity in inference, where energy consumption is of greater significance. FuriosaAI states that RNGD can run large AI models while using less power than Nvidia’s top – tier GPUs.

CEO Paik presented RNGD at Stanford’s Hot Chips conference, showcasing the Llama model’s performance with better power efficiency. “It was a moment when we felt we could truly move forward with our chip with confidence,” Paik said.

Company Roots in Samsung, Vision from Stanford

June Paik founded FuriosaAI after recovering from an Achilles injury he sustained at a Samsung soccer match. During his recovery, he studied AI through Stanford’s online courses.

After returning to Samsung, he decided to leave and venture into the AI field. He then founded FuriosaAI with former colleagues in hardware and algorithms.

Hanjoon Kim, the current CTO, worked with Paik and co – founded the startup. He says Paik focused on making fast decisions and formulating long – term strategies.

The startup adopted strategies from the book “Blitzscaling,” advocating speed and bold action in early market entry. Paik emphasized this during the company’s initial development stages.

Market Position Strengthens as Talks Advance

FuriosaAI completed a new funding round, valuing the company at nearly $700 million. Meta showed interest in acquiring the company last year, but no deal was finalized.

FuriosaAI claims it is now in discussions with more potential clients and industry players. These talks follow the successful demonstrations of RNGD’s AI inference capabilities.

According to Paik, RNGD rivals Nvidia’s GPUs while using less energy, which reduces operating costs. “A market dominated by a single player is not a healthy ecosystem, is it?” he said.

supports AI development through both private and government initiatives. Nvidia has already finalized GPU supply deals with South Korea, arranged by the government.

FuriosaAI continues to build commercial partnerships as it moves into production. The company remains focused on AI inference chips with performance and power advantages.