- Furiosaai’s new RNGD -Server delivers 4 Petaflops Calculation at 3 KW for Effective AI
- Businesses will be able to scale AI workloads without expensive infrastructure changes
- RNGD Server provides compatibility with Openais API along with a growing SDK Function Set
The South Korean Chipstart Furiosaia, which famously went away from Meta’s acquisition bid of $ 800 million, continues to push forward with new products as the demand for effective AI infrastructure Soars.
Start -up seeks to provide companies with hardware that can run LLMs without the expensive data center upgrades and heavy energy costs often associated with GPUs.
The latest product, the RNGD server, is a business-ready AI apparatus run by Furiosaia’s RNGD (pronounced “Renegade”) AI Inference Chips.
Scaling more effectively
Each system delivers 4 petaflops of FP8 calculation and 384 GB HBM3 memory while working at only 3 kW.
In comparison, NVIDIA’s DGX H100 servers can draw more than 10 kW. This means that a standard 15 kW data center stand can contain five RNGD servers, while the same rack would only fit a DGX H100.
Furiosaa says that most data centers are limited to 8 kW per day. Rack or less, its design addresses a key barrier to businesses.
Advanced AI models running in such environments typically require new cooling and power systems.
The company says that the companies, by adopting the RNGD server, will be able to scale more effectively, while compatibility maintains compatibility with Openais API.
The startup recently closed a $ 125 million Series C Bridge Round and expanded its partnership with LG AI Research.
LG uses RNGD -hardware to run its exaone models and says it gets more than twice as much inference performance per day. Watts compared to GPUs.
Furiosaa also recently collaborated with Openai, where the two companies demonstrated the open weight GPT-OSS 120B real-time Chatbot, running on only two of Furiosaia’s RNGD accelerators.
The new RNGD server receives continuous updates to Furiosai’s SDK, which recently introduced Inter-Chip Tensor parallelism, new compiler optimizations and expanded quantization formats.
RNGD server is currently being sampled with global customers and is expected to be available for orders by the beginning of 2026.
