Armenia Just Opened the Biggest AI Computing Hub in the Former Soviet Region

Firebird has switched on a large AI factory in Armenia, backed by Nvidia and CoreWeave, with plans to expand across Central Asia and beyond.

AI2Day NewsdeskAI-assistedPublished Updated Editor: Lee Brown3 min read
Illustration: a dense pile of bundled cash stacks filling the entire frame, shot from slightly above at a slight angle
Illustration made with AI. Not a photograph of the events described.
Share

Key points

  • Firebird launched what it says is the largest AI factory in the CIS region, a group of countries that includes Armenia and Kazakhstan, in the Armenian city of Hrazdan.
  • Nvidia plans to invest in Firebird, following an earlier investment by cloud computing company CoreWeave in 2025.
  • Firebird plans to deploy more than 70,000 specialised AI chips and 300 megawatts of computing capacity in Armenia by the end of 2027.
  • AI search company Perplexity is among the first customers confirmed for the new facility.
  • The factory was built in just over six months and forms part of a roughly 2-gigawatt infrastructure plan spanning Armenia, Kazakhstan and other markets.

A new kind of power plant opened this week in Armenia. It runs on data.

Firebird, an emerging AI cloud company, has launched what it describes as the CIS region's largest AI factory in Hrazdan. CIS stands for Commonwealth of Independent States, the grouping of countries that were part of the former Soviet Union. The opening ceremony drew Armenia's Prime Minister Nikol Pashinyan, Kazakhstan's Deputy Prime Minister Zhaslan Madiyev and David Allen, the acting U.S. Ambassador to Armenia.

What is an AI factory, and why does it matter?

Think of an AI factory as a very large, very power-hungry building full of specialised computers that companies rent to build and run AI systems. Without access to this kind of computing power, countries can't train their own AI models or run AI services in their own languages.

That's the gap Firebird wants to close. Armenia and its neighbours have relied on AI infrastructure based elsewhere, mostly in the United States and Europe. A local facility means local developers and businesses can build AI products suited to their own needs without shipping data abroad or paying for transatlantic connections.

Firebird plans to fill the Hrazdan site with more than 70,000 Nvidia Rubin and Blackwell GPUs, the specialised chips that do the heavy number-crunching AI needs, reaching 300 megawatts of capacity by the end of 2027. For scale, a typical hospital uses around one to two megawatts.

Who is paying for this?

Nvidia announced it intends to invest directly in Firebird. CoreWeave, a major U.S. AI cloud provider, invested earlier in 2025. Both have a clear interest: more AI factories mean more demand for their chips and services.

The hardware sits on Dell PowerEdge servers, with power infrastructure from Schneider Electric and cooling systems from Vertiv. Cooling matters more than it sounds. AI chips generate enormous heat, and keeping them cold enough to run continuously is one of the hardest engineering problems in building this kind of facility. Firebird says its design lets it run up to 40% more chips in the same physical footprint compared with older approaches.

What happens next?

Firebird's ambitions reach well beyond Armenia. The company is working toward a roughly 2-gigawatt infrastructure plan across Armenia, Kazakhstan and additional markets. Two gigawatts is enough to power a city of a million people.

Perplexity, an AI-powered search service, is already signed up as a customer for its AI agent platform. An AI agent is software that can carry out multi-step tasks on its own, going further than a simple chatbot.

First reported by Nvidia, the Hrazdan factory went from planning to operational in just over six months. That speed is the real story here. Larger cloud providers have largely ignored these markets, and Firebird is betting that moving fast in underserved regions is a better strategy than competing head-on in the United States. Our coverage of data center politics in the US shows why building somewhere that wants you is increasingly worth something.

© 2026 AI2Day