이미지: NVIDIA
Summary
- NVIDIA announced that AI cloud company Firebird has brought online the largest AI factory in the CIS region, located in Hrazdan, Armenia.
- Firebird plans to deploy more than 70,000 NVIDIA Rubin and Blackwell GPUs and build out 300 megawatts of infrastructure by the end of 2027.
- NVIDIA, which invested in CoreWeave earlier this year, has expressed intent to invest in Firebird as well, with Perplexity cited as an early customer.
- 운영 주체
- AI 클라우드 기업 Firebird, 아르메니아 흐라즈단(Hrazdan) 소재
- 구축 기간
- 6개월 남짓 만에 가동 개시
- 2027년 목표
- NVIDIA Rubin·Blackwell GPU 7만 장 이상, 300MW 인프라
- 플랫폼
- NVIDIA DSX — 동일 면적에서 최대 40% 더 많은 GPU 운용
- 장기 로드맵
- 아르메니아·카자흐스탄 등 약 2기가와트(GW) 규모
- 투자
- NVIDIA 투자 의향 표명, 올해 초 CoreWeave가 선행 투자
- 주요 파트너
- Dell Technologies(서버), Schneider Electric(전력), Vertiv(냉각)
- 초기 고객
- Perplexity가 에이전트 플랫폼·답변 엔진용 인프라 확보 협력
The Largest AI Factory in the CIS, Built in Hrazdan
AI cloud company Firebird has brought online the largest AI factory in the CIS (Commonwealth of Independent States) region in Armenia, NVIDIA announced on its blog. Built on NVIDIA accelerated computing and Dell Technologies' high-performance AI infrastructure, the facility reflects an ambition to establish Armenia as a new hub for AI compute.
The opening ceremony was attended by Armenian Prime Minister Nikol Pashinyan, Kazakh Deputy Prime Minister Zhaslan Madiyev, and David Allen, Chargé d'Affaires at the U.S. Embassy in Armenia — underscoring the project's character as a national-level infrastructure initiative.
What stands out is the speed. According to NVIDIA, the facility was built and moved into actual operation in a little over six months — an unusually short timeline for an AI data center, which typically requires large-scale power and cooling systems.
70,000 GPUs and 300 Megawatts by 2027
Firebird has laid out plans to deploy more than 70,000 NVIDIA Rubin and Blackwell GPUs in Armenia and to build out 300 megawatts of AI infrastructure capacity by the end of 2027.
| Item | Detail |
|---|---|
| Target date | End of 2027 |
| GPU volume | 70,000+ NVIDIA Rubin/Blackwell units |
| Power capacity | 300MW |
| Build time (phase 1) | A little over 6 months |
| Long-term roadmap | ~2GW across Armenia, Kazakhstan, and other locations |
At this scale, power efficiency becomes a key variable. Firebird's AI factory is built on the NVIDIA DSX platform, which integrates accelerated computing, networking, power, and cooling into a single co-designed system. NVIDIA said DSX enables up to 40% more GPUs to be deployed within the same footprint, increasing tokens produced per dollar and value delivered per megawatt. This figure is based on NVIDIA's own materials and has not been independently verified.
Power and Cooling Partners
Power infrastructure at the Hrazdan site was handled by Schneider Electric, which supplied key equipment — including medium- and low-voltage switchgear, three-phase uninterruptible power supplies (UPS), and rack enclosures — quickly enough to meet the compressed timeline, according to NVIDIA.
Cooling was handled by Vertiv, combining chilled water technology with advanced control systems and Vertiv TrimCooler technology. The iCOM CWM chilled water management system centrally coordinates cooling resources, improving responsiveness to fluctuations in AI workloads.
| Partner | Role |
|---|---|
| NVIDIA | Accelerated computing, networking, AI software, DSX reference architecture |
| Dell Technologies | High-performance AI server infrastructure |
| Schneider Electric | Switchgear, UPS, rack enclosures, and other power equipment |
| Vertiv | Chilled-water-based cooling architecture and controls |
Capital Following NVIDIA and CoreWeave
Firebird also announced that NVIDIA has expressed intent to invest in the company. Earlier this year, CoreWeave reportedly made an investment first. Specific investment size or terms were not disclosed.
Firebird's plans extend beyond a single site. With NVIDIA's backing, the company is pursuing an AI infrastructure roadmap of roughly 2 gigawatts spanning Armenia, Kazakhstan, and additional markets — a strategy aimed at securing large-scale compute clusters in frontier markets.
The shift of AI infrastructure investment toward a national-level competitive axis is a trend worth following alongside metallab.ai's coverage of AI infrastructure and semiconductors.
Sovereign AI Demand and Early Customers
NVIDIA cited, as background for the project, the idea that even as AI services have become globally available, individual nations still need their own capacity to develop and operate AI tailored to their languages, industries, and national priorities. Firebird's AI factory is positioned to provide compute resources to local developers, startups, enterprises, universities, and public institutions.
Early demand is coming from AI-native companies. NVIDIA said Perplexity is working with Firebird to secure high-performance infrastructure for running its AI agent platform and answer engine. Details such as contract size and duration were not disclosed.
This report is based on a single source — NVIDIA's blog — and the deployment volumes and power capacity figures cited should be read as future targets rather than current figures.
