Accelerators for model inference
FuriosaAI develops specialized AI chips and software, including the RNGD accelerator. Its company timeline distinguishes earlier computer-vision hardware from the later data-center inference platform, and separates sampling from mass production. These stages describe different levels of availability. [1]
An actual EXAONE deployment
In the July 2025 LG AI Research adoption, FuriosaAI described RNGD hardware installed at LG AI Research’s data center. The teams optimized EXAONE across multiple cards and server configurations. This was a concrete accelerator integration, rather than a borrowed benchmark or a generic supplier mention. [2]
Configuration-specific evidence
The announcement reported performance and power comparisons from the vendor/customer evaluation. Those results apply to the stated models and configurations. Plans to expand enterprise server offerings are separate from that installed deployment, and do not establish adoption by every LG business. [2]
Sources
- About — FuriosaAI
FuriosaAI · retrieved Oct 23, 2026
- LG AI Research taps FuriosaAI to achieve 2.25x better LLM inference performance vs. GPUs
FuriosaAI · Jul 22, 2025