FuriosaAI's AI inference accelerators will be deployed in a large-scale data centre in Europe. FuriosaAI said on Tuesday it is joining U.S. AI infrastructure company I/ONX HPC and data centre developer and operator Velox in a project to build a 15 MW AI data centre in Stockholm, Sweden.

The AI data centre will start operating its first-phase 2 MW facility in early 2027, and an additional 8 MW will be added within the year. Over the longer term, it will expand in stages to 15 MW. FuriosaAI will supply 1,800 of its inference accelerators, RNGD (Renegade), in the first phase, and more than 8,800 units in total are set to be introduced sequentially through the expansion stages.

RNGD will be installed in I/ONX HPC's high-density AI server platform, Symphony SixtyFour, to build a heterogeneous AI computing environment alongside GPUs. The structure uses RNGD for large language model and agentic AI inference, and GPUs for model training and other computations. It is designed to raise overall data centre performance and power efficiency at the same time. FuriosaAI will provide the RNGD hardware and software stack, Velox will handle data centre development and operations, and I/ONX HPC will be responsible for system integration and the AI platform.

Major European countries are pursuing AI sovereignty and the build-out of sustainable, power-efficient AI infrastructure as a core strategy, and demand is rising quickly for high-efficiency inference infrastructure. Sweden is emerging as a key hub for AI infrastructure in Europe, based on stable power infrastructure and a mature data centre ecosystem. The project is set to provide AI inference services to companies and cloud providers across Europe.

FuriosaAI CEO June-ho Baek (백준호) said, "AI data centres are now moving beyond a simple competition in computing performance, and power efficiency, total cost of ownership (TCO) and sustainability are becoming key competitiveness." He said, "This project is an important milestone as FuriosaAI's AI inference infrastructure is applied in earnest to large-scale data centres in Europe, and we will continue to accelerate the spread of next-generation AI inference infrastructure together with global partners."