Firebird launches AI factory in Armenia with 70,000 accelerators and 300 MW by end-2027

Firebird, an emerging AI cloud provider, announced the opening of the largest AI factory in the CIS region. The AI factory, located in Armenia, is built on NVIDIA-accelerated computing infrastructure and Dell Technologies hyper-converged equipment, and is intended to provide foundational infrastructure for training, fine-tuning and deployment of large-scale models. The inauguration was attended by Armenian Prime Minister Nikol Pashinyan, Kazakhstan Deputy Prime Minister Zhaslan Madiyev, and David Allen, head of the U.S. temporary diplomatic mission in Armenia.

Firebird plans to deploy more than 70 thousand Rubin and Blackwell accelerators from NVIDIA in Armenia by the end of 2027, drawing on a total power capacity of 300 MW. The AI factory runs on the NVIDIA DSX platform, which integrates accelerated compute, networking, power and cooling into a single, jointly designed system. According to the company, the architecture can run up to 40 percent more accelerators in the same floor space, generate more tokens per dollar and deliver higher value per megawatt. The AI factory was built at the Hrazdan site and completed in just over half a year; Schneider Electric supplies the electrical infrastructure, including three-phase switchgear, while Vertiv provides the cooling architecture, combining chilled-water technology with iCOM CWM controllers for dynamic AI-load response.

Supported by NVIDIA, Firebird is developing a roadmap for roughly 2 GW of AI infrastructure across Armenia, Kazakhstan and additional markets. The company also disclosed that NVIDIA intends to invest in Firebird, following a prior investment by CoreWeave earlier this year. The capital will be used to expand advanced compute clusters in frontier markets. Initial demand for the compute resources comes from AI-native firms such as Perplexity, which uses Firebird’s infrastructure for its AI-agent platform and answer engine. The AI factory is expected to enable developers, startups, universities and public institutions to build and run local models for languages, industries and national priorities, and not rely solely on global AI services.