FuriosaAI expanded availability of its RNGD AI inference accelerator across Europe, starting with the installation of RNGD servers at Equinix’s LS2 data center in Lisbon. The deployment will give European enterprises access to environments for evaluating FuriosaAI’s inference architecture and software stack on advanced AI models, while supporting the company’s recently established Lisbon office and compiler-focused R&D operations.
RNGD uses FuriosaAI’s proprietary Tensor Contraction Processor architecture and a 5nm semiconductor process. Each accelerator delivers 512 TFLOPS of FP8 compute performance within a 180W thermal design profile. FuriosaAI integrates up to eight RNGD accelerators into its NXT RNGD Server, a 3kW-class inference system designed for production AI workloads and higher inference compute density within power-constrained data center environments.
FuriosaAI positions RNGD for standard air-cooled data centers, allowing operators to deploy AI inference infrastructure without adding liquid cooling systems or major facility retrofits. The European expansion follows FuriosaAI’s recently announced collaboration with Broadcom to develop a third-generation inference accelerator targeting hyperscale inference workloads and frontier AI models with one trillion or more parameters. FuriosaAI also showcased its production hardware and software at the RAISE Summit in Paris.
• Initial European RNGD server deployment underway at Equinix’s LS2 data center in Lisbon.
• RNGD uses FuriosaAI’s proprietary Tensor Contraction Processor architecture and a 5nm manufacturing process.
• Each accelerator delivers 512 TFLOPS of FP8 compute performance at a 180W thermal design profile.
• The NXT RNGD Server integrates up to eight accelerators into a 3kW-class inference platform.
• RNGD targets deployment in standard air-cooled data centers without requiring liquid cooling or major facility upgrades.
• FuriosaAI operates a compiler-focused R&D lab and flagship European office in Lisbon.
• FuriosaAI said it raised more than $250 million through its Series C Bridge financing and works with TSMC, SK hynix, and Broadcom.
“We are pleased to be establishing an important new distribution channel in Europe with Equinix,” said FuriosaAI Co-Founder and CEO June Paik. “By pairing Equinix’s infrastructure footprint designed for efficiency and sustainability with our high-performance, energy-efficient RNGD architecture, we unlock the ability for enterprises to run inference sustainably and reliably.”
🌐 Analysis: Data center operators confront power availability, rack-density, and cooling constraints. The deployment gives FuriosaAI a European evaluation environment as established accelerator suppliers and AI silicon startups compete on performance per watt, software maturity, and the ability to deploy inference capacity within existing data center infrastructure.
🌐 We’re tracking the latest developments in networking silicon. Follow our ongoing coverage at: https://convergedigest.com/category/semiconductors/
|
FuriosaAI
HOT START-UP
South Korean AI semiconductor company developing energy-efficient inference accelerators for enterprise, cloud, and sovereign AI infrastructure.
|
|
| Overview | FuriosaAI designs AI inference accelerators and software for deploying large language models and multimodal AI workloads. Its RNGD accelerator targets high-throughput inference with a focus on tokens-per-watt efficiency and deployment in standard, air-cooled data centers. |
| Why It Matters | As AI infrastructure expands beyond a small number of hyperscale campuses, inference efficiency, power availability, and deployment flexibility are becoming central constraints. FuriosaAI is positioning RNGD for AI deployments where power density, cooling requirements, and sovereign infrastructure needs are key factors. |
| Founded | 2017 |
| Headquarters | Seoul, South Korea; U.S. presence in Santa Clara, California; European office in Lisbon, Portugal |
| CEO / Leadership | June Paik, Founder and CEO |
| Core Technologies | AI inference acceleration Energy-efficient silicon LLM inference Air-cooled deployment |
| Key Products / Platforms | RNGD, FuriosaAI’s second-generation AI inference accelerator, supported by the company’s software stack for model deployment and optimization. |
| Major Milestone | In July 2026, FuriosaAI announced expanded availability of RNGD across Europe in collaboration with Equinix, with an initial deployment underway at Equinix’s Lisbon LS2 data center. |
| Target Markets | AI Infrastructure Data Centers Enterprise AI Sovereign AI Inference Cloud |
| Editorial Coverage | Converge Digest is tracking FuriosaAI as part of its coverage of AI accelerator silicon, inference infrastructure, power-efficient data center architectures, and sovereign AI deployments. |
| Industry Context | FuriosaAI competes in the AI accelerator market as enterprises and cloud providers evaluate alternatives to GPU-only inference infrastructure. Its emphasis on power efficiency and standard data center deployment aligns with growing constraints around electricity availability, rack density, and cooling. |
| Profile Updated | July 2026 |
| Related Knowledge Hubs | AI Infrastructure | Data Centers | Semiconductors | Startups |