YSU’s supercomputer: When a wish becomes a necessity

There is a room in one of Yerevan State University’s buildings where it is always cold. Not because of winter, but because a system operates inside that generates heat from thinking.

The official opening of the supercomputer took place in this room on January 13, 2026. It is designed to solve scientific problems that are “unattainable” for conventional computers.

Acquired through government funding, the supercomputer is equipped with 64 NVIDIA H100 computing cards.

[[gallery1]]
Hrant Khachatryan leads a scientific group at YSU specializing in machine learning. In 2023, when the team began collaborating with colleagues in the United States, a reality quickly became evident: Armenian researchers were tackling the same scientific challenges as American laboratories, but with vastly fewer resources. 

[[gallery2]]
“When they saw the computers we were working with, they said, ‘This won’t work, we need more resources,’” Hrant recalls.

At that moment, the idea of acquiring a high-performance computer ceased to be a wish and became a necessity. Yet necessity alone was not a solution. Turning the idea into reality required partners, funding, and patience to navigate government procedures.

The substantive justification for the project was prepared by colleague Armen Aghajanyan, who compiled a detailed brochure explaining the purpose of such a system, the problems it could solve, and the opportunities it would open.

By the end of 2023, during budget discussions in the National Assembly, funding was proposed not only for the supercomputer itself but also for the establishment of a full data processing center. In 2024, the funds were officially allocated.

[[gallery3]]
“The computer physically arrived at the end of 2024, but we could not launch it because construction of the data center was delayed,” recalls Hrant Khachatryan.

By 2025, the hall was ready – equipped with power protection systems, backup generators, and advanced cooling infrastructure. The first thing one notices upon entering the data processing center is the cold: the temperature in the hall is consistently maintained at 16-18 degrees.

[[gallery4]]
Ruben Abrahamyan, Deputy Head of Yerevan State University’s Information Technologies Department, demonstrates the system’s layered infrastructure.

To ensure uninterrupted power supply, three groups of UPS units can sustain the full load for up to 20 minutes, providing sufficient time for the diesel generators to activate.

[[gallery5]]
The hall’s floor is raised, allowing hundreds of power and network cables to run beneath it. Overhead, gas-based fire suppression systems are installed to extinguish fires without damaging the equipment.

One of the most significant technological features is the liquid cooling system.

[[gallery6]]
“This cooling technology is rarely used not only in Armenia, but across the region. It reduces energy consumption and increases device efficiency,” says Ruben Abrahamyan.

[[gallery7]]
The supercomputer consists of eight computing nodes, each equipped with eight high-performance GPUs – a total of 64 graphics processors specifically designed for training artificial intelligence models.

[[gallery8]]
Such systems are called HPC (High Performance Computing) or super-powerful computing systems.

Hrant Khachatryan’s group is already working, for example, on an AI model that can be “asked” to propose new molecules with specific biological properties.

[[gallery9]]
“We are developing artificial intelligence that can propose molecules that may not yet exist, but could possess properties of scientific interest. Biologists can then test these proposals in the laboratory,” says Hrant Khachatryan. 

Another research direction focuses on creating AI components for robots and drones, enabling them to make decisions not only through pre-programmed instructions, but also through real-time perception of their environment. 

[[gallery10]]
For Rafayel Barkhudaryan, Vice-Rector for Academic Affairs of YSU, this system is not the project of a single group, but part of the university’s broader scientific strategy.

“We have several groups engaged in artificial intelligence, machine learning, materials science, optics, and chemistry. This platform is for everyone,” says Rafayel Barkhudaryan. 

Beyond AI research, the system is also designed to serve physicists, chemists, and biologists who require large-scale computational capacity in their work.

“A super-powerful computer does not make work easier, it makes it more challenging, because it expands your ambitions. You begin to formulate bigger problems,” Rafayel Barkhudaryan says frankly. 

[[gallery11]]
The system also serves as an educational tool. Master’s and PhD students learn to work with infrastructures that are typically available only in major international research centers. 

[[gallery12]]
For years, a common refrain in Armenia was, “We don’t know how to use such vast resources.” In 2019, it was noted that human training for operating large-scale computing systems remained insufficient.

Over the past five years, however, the situation has changed. Today, dozens of specialists in Armenia – both in universities and private companies – are capable of working with such infrastructures.

[[gallery13]]
“I was sure it would work. I just had to keep going,” says Hrant Khachatryan. 

[[gallery14]]
The machines in this hall are running now. But in reality, something else is at work here as well: the belief that science in Armenia can not only survive, but also generate and offer new ideas to the world.

Astghik Hovhannesov

Photos by David Ghahramanyan