/ Armenia just became an AI compute powerhouse.
Liam Ortiz is a tech journalist who covers AI related big tech and breaking news at TheTweaks. Before joining TheTweaks he worked for almost four years in corporate and national tech news in different companies. Few are quick but Liam is quicker, he breaks news before anyone else and that makes her special. His passion is somewhere connected with profession as his hobby is watching documentary movies.
The world just saw a big change in the evolution of AI infrastructure and it wasn’t in the place that anyone would’ve expected. Firebird, a startup firm focused on the provision of cloud AI services, and NVIDIA, Dell Technologies, have just turned on the biggest AI factory in the CIS region in Armenia.
It is not some sort of small-scale project under a fancy name. It captured the interest of Armenia’s prime minister Nikol Pashinyan, Kazakhstan’s deputy prime minister Zhaslan Madiyev and a representative of the US government, who took part in the opening ceremony – proof that this piece of infrastructure is important from a government perspective.
Massive compute power, ideal for training, fine-tuning and running AI models is the backbone of the AI era, and you can get that power with an AI factory. While many AI services are globally accessible, every single country has its own computational power needed to develop and operate AI systems tailored to its language, industry needs and priorities, says Firebird. This type of capabilities will be offered by this factory in Armenia.
To support this endeavor, by the end of 2027, Firebird will roll out more than 70,000 NVIDIA Rubin and Blackwell GPUs and 300 MW of capacity for AI infrastructure in Armenia. To give you some perspective – these are the numbers for hyperscale facilities in the USA or Europe – now it’s in the South Caucasus.
This facility is based on NVIDIA DSX platform – the integration of accelerated computing, networking, power and cooling into one code-signed solution. That is, everything is designed from the ground up rather than hacking GPUs into network switches, or cooling systems separately.
The advantage of DSX is that the factory will be able to run up to 40% more GPUs in the same space, providing a higher number of tokens per dollar and more value per megawatt consumed, NVIDIA says. It is important for such density increase where power is the real constraint (and not the chips).
Firebird however is not stopping here, the company is creating an approximate 2-gigawatt AI infrastructure roadmap that will cover not just Armenia but also Kazakhstan and other regions. Its strategy puts the company in the leading position in the CIS region in the field of provision of AI infrastructure.
It seems that the ambitions are really serious, NVIDIA intends to make investments in Firebird after its previous investment made by CoreWeave earlier this year. This is an obvious indicator that NVIDIA is not just selling hardware to this company, it’s investing in the company, as well.
More telling is the construction speed, the entire facility became fully operational in just over six months, which is fairly quick even for the AI-infrastructure standards where everyone claims to be rapidly building.
The lesser sexy but just as important partners under the GPU headlines are the ones who are providing power literally. To keep up with the rate of compute deployment, Schneider Electric supplied the medium and low-voltage switchgear, as well as three-phase UPS systems and rack enclosures for Hrazdan.
In terms of the thermal aspect, Vertiv delivered the cooling infrastructure based on their chilled-water technology, advanced cooling controls, as well as their TrimCooler system, which can be remotely controlled from an iCOM CWM Chilled Water Manager that coordinates with AI workloads.
The workloads are already in place too. Perplexity is one of the first AI native companies that has integrated with Firebird and is taking advantage of its infrastructure to run its AI agent platform and answer engine, the first tangible indication that the factory isn’t just a photo opportunity for the ribbon-cutting ceremony.
Stripping off all the marketing jargon, what is going on is a geopolitical and economic gamble – AI compute is becoming the form of national infrastructure, comparable to power grids or telecom infrastructure. Armenia, along with Kazakhstan, positions itself as a compute hub in the region, which has so far been mostly ignored by the buildout of AI infrastructure – dominated by the US, China and Gulf regions.
It is either the flourishing AI ecosystem in this region, or lots of compute that generate cash for the providers. However, the scale, speed and lineup of partners (NVIDIA, Dell, Schneider Electric, Vertiv, CoreWeave, Perplexity) indicate that this is taken rather seriously as a long-term approach.
This is not a product launch. It’s an infrastructure land grab and a very clever one. NVIDIA is able to bring its DSX platform and Rubin/Blackwell silicon to a vital but underdeveloped geographic area, Firebird will be the de facto AI cloud for the CIS and Armenia will be able to bypass building the AI ecosystem altogether and enter the AI economy.
The most tangible and reliable number in the entirety of the news release is the 40% density boost provided by DSX, everything else is goal setting and future planning. What will be interesting to see over the next 18 months is if 70,000 GPUs for target and 2 GW regional rollout actually happen, as in this industry, timelines can be tricky.
Tim Cook’s 15 years as Apple’s chief executive officially ended on September 1 and now the torch has been passed on to John Ternus, who is the company’s long-time hardware. It’s Apple’s first leadership change…
Be respectful and constructive. Have a question or feedback? We’d love to hear from you. Contact us at contact@thetweaks.com