Facing increasing restrictions and backlash against data centers, the United States may now start turning outward, with Armenia poised to serve as an enthusiastic pilot model for outsourcing them.
Backlash against AI data centers has taken place across the U.S. — a March Gallup poll found 71% of the public said they opposed the construction of a data center in their area, with just 7% saying they “strongly favor” as much. But popular backlash hasn’t put a dent in the demand for more computing power, putting tech leaders in a tight spot as they face waves of local legislation restricting new construction. One possible solution is emerging halfway across the world, in the small Caucasian country of Armenia.

Eleveight AI, an Armenian company founded in 2024, opened the Caucasian country’s first commercial AI data center in June, one it hopes is just the first of many. And critical to this optimism is the difference in public perception — Arman Aleksanyan, co-founder and CEO of Eleveight AI, told the Washington Examiner that the Armenian public’s reception has been “very positive.” No polling on Armenians’ perceptions of data centers exists because they are so new in the country, but the kind of popular discontent that has spread across the U.S. hasn’t surfaced publicly.
Armenia has been pursuing close AI cooperation with the U.S. for both economic and strategic reasons, seeking to deepen its ties with the U.S. to broaden its rolodex of international partners after losing a war to neighboring Azerbaijan and as relations with its traditional ally, Russia, worsen.
Eleveight’s data center and its future plans are fully reliant on the U.S. Eleveight Chief Business Officer Davit Arakelyan told the Washington Examiner in an interview that the project was a “practical implementation” of the Armenia-U.S. memorandum of understanding, signed by President Donald Trump and Armenian Prime Minister Nikol Pashinyan, on Aug. 8, 2025.
“We are Armenia, we have a border with Iran. If that memorandum would not have taken place, nobody would give us that technology,” he said.
The MOU helped Armenia become “one of the first countries in the world to deploy NVIDIA’s latest B300 GPUs,” Aleksanyan said. When asked to estimate how the project was divided by importance from a technological point of view, Arakelyan put it at “90% U.S., 10% other,” with the project constructed by locals but relying on U.S. technology from GPUs to Dell-built servers.
A State Department spokesperson welcomed Armenia’s efforts to increase their AI cooperation as part of its Pax Silica initiative, and said it was working with the country to “build a system that rewards growth and innovation while unleashing American startups, developers and companies to set the pace.”
“We encourage Armenia to select trusted cloud and data center providers as these technologies are the foundation of a prosperous economy and the SI revolution,” the spokesperson added.
A large part of the difference in public perception in the U.S. vs. Armenia has been the difference in design — Aleksanyan boasted that Eleveight’s data center, which is aiming for 40 megawatts, was built on a smaller, sustainable model, with a cooling system that doesn’t use any drinking water.
“We believe our approach offers a viable model for responsible AI infrastructure,” he said. “We have not faced the kind of resistance seen around some major U.S. projects, and our environmental choices help explain that.”
“In the end, a facility’s design, location, energy supply and operation make the difference,” Aleksanyan said. “Thoughtfully developed data centers can build public trust and spread the benefits of AI more widely.”
Aleksanyan is so confident in the viability of Eleveight’s design, in fact, that he spoke of his intention to “expand both in Armenia and internationally, carrying forward what we have learned in Gagarin.”
The AI industry has created an ever-growing demand for energy and computation, drastically changing what is considered a large data center. Bryan Marsh, a 17-year veteran in the data center industry and former CEO of StratCap’s data center division, told the Washington Examiner that his standard Digital Realty data center from around 2009 to 2017 had about 7 MW, making Eleveight’s data center colossal by comparison. But today’s U.S. hyperscale sites run 300MW to 1 gigawatt, dwarfing Eleveight.
Jonathan Koomey, an independent researcher formerly at Lawrence Berkeley National Laboratory who’s been studying data centers for more than a quarter century, told the Washington Examiner that whether Armenia’s model can be effective for AI data center development depends on what compute loads they are given.
If the centers are used for AI training, where they can be run in “batch mode” with little to no communication with the outside world, smaller data centers are reliable. But if the data centers are used to serve customer and agent queries, they become less viable due to latency problems — the lag between when a customer types in a prompt compared to when they get the reply.
Ivo Ivanov, CEO of DE-CIX, a major internet exchange operator, largely echoed this argument, putting its viability down to the tasks it is used for.
“For extremely low-latency, real-time applications, such as financial trading and physical AI, compute would need to be located in the same metro area as the users,” he said. “For less latency-sensitive use-cases, such as drug discovery or chatbots, compute may be located hundreds of kilometers from the users.”
Marsh, Ivanov, and Koomey all agreed, however, that smaller, dispersed data centers would play a key role in a diversified global AI infrastructure network, filling certain niches, but so far aren’t in a position to truly replace hyperscale facilities.
“Such facilities cannot replace the mega-scale campuses for AI training at this stage,” Ivanov said. “Both will remain essential to the further development of AI.”
TRUMP ADMINISTRATION HELPS END ARMENIA AND AZERBAIJAN’S FOREVER WAR
Marsh agreed that hyperscale sites are still needed, but outlined the need for smaller, distributed facilities such as Eleveight for less compute-heavy tasks like IT services. Eleveight’s focus on serving a U.S. market could help fill certain niches, establishing Armenia as an emerging AI partner.
“I am confident that Armenia will not only maintain but strengthen its position in AI factories, and that this will become one of the most important areas of cooperation between our two countries,” Aleksanyan said.
