Firebird launches the CIS region's largest AI factory in Armenia, backed by NVIDIA

Armenian AI cloud company Firebird has switched on what NVIDIA calls the largest AI factory in the CIS region, a data center in Hrazdan, Armenia built with NVIDIA hardware and support from Dell Technologies and CoreWeave.

What's new

NVIDIA said in a blog post that "Firebird plans to deploy more than 70,000 NVIDIA Rubin and Blackwell GPUs and 300 megawatts of AI infrastructure capacity in Armenia by the end of 2027." The facility runs on NVIDIA's DSX platform, a rack-to-data-center design NVIDIA says is built for energy efficiency and dense GPU deployment. According to NVIDIA, the DSX approach means the site "can run up to 40% more GPUs on the same footprint, producing more tokens per dollar and extracting more value" than conventional data center designs. NVIDIA also said the facility was completed in just over six months.

The buildout pairs NVIDIA's Rubin and Blackwell GPU generations in the same facility, giving Firebird a path to scale compute capacity through 2027 as it brings additional megawatts of power online. Dell Technologies and CoreWeave are named as infrastructure and delivery partners on the project, alongside unspecified regional government and business leaders.

Context

"AI factory" is the term NVIDIA uses for large, purpose-built data centers designed specifically to train and run AI models at scale, as opposed to general-purpose cloud data centers. NVIDIA has been promoting the AI factory model globally as a way for regional cloud providers and sovereign-AI initiatives to stand up large GPU capacity quickly, partnering with server makers like Dell and cloud operators like CoreWeave to handle deployment. The Armenia project is part of a broader pattern of AI infrastructure buildouts happening outside the traditional US hyperscaler footprint, in regions seeking their own large-scale compute capacity.

Why it matters

The race to build AI compute capacity has increasingly gone regional, with countries and smaller cloud operators racing to secure GPU supply and power for their own "sovereign AI" ambitions rather than relying solely on US hyperscalers. A 70,000-GPU, 300-megawatt buildout completed in roughly six months in a country the size of Armenia is a signal of how much capital and hardware supply is now available to move fast on AI infrastructure outside the biggest cloud providers — and how central NVIDIA's newest GPU generations and data-center reference designs have become to enabling that speed.