Russia’s Nuclear “Batteries” Are Getting Smarter And a New Global Race Has Already Begun

As artificial intelligence reshapes the world’s energy demands, Russia’s top nuclear research university is quietly developing mini atomic reactors built specifically for data centers placing Moscow at the heart of a new technological arms race.

Russia’s Nuclear Scientists Are Building Atomic Power Units for AI Data Centers

Whether the world likes it or not, digitalization is here and it is only accelerating. Artificial intelligence sits right at the top of that wave. Russia as a major global player clearly understands this reality. Valery Romanyuk, vice-rector of MEPhI, Russia’s foremost nuclear research university, confirmed in a recent statement that the consortium known as “Big MEPhI” is already developing small and ultra-small nuclear reactors. These reactors range from five to fifty megawatts in capacity. The goal is direct power the data processing centers that the entire world is building at a breathtaking pace.

“Big MEPhI” Consortium Was Built to Move Fast and Cut Bureaucracy

The importance attached to linking nuclear energy with digital technology is no accident. The “Big MEPhI” consortium was deliberately set up two years ago and not as a standard legal entity. This was a calculated move to eliminate bureaucratic and management bottlenecks. The aim was to accelerate Russia’s scientific and technological development and achieve genuine technological sovereignty. Over time, it became clear just how urgent the challenge was. Creating ultra-small reactors was then carved out as a separate sub-program a clear sign that Moscow treats this as a priority.

Russia’s Rosatom the state nuclear giant often gets attention for laying concrete and loading uranium rods. But all of that is the final stage of a long, invisible, and painstaking scientific process. The “Big MEPhI” model allows partner companies to bring specific challenges to Russia’s nuclear specialists. Findings and practical results can either remain classified at various access levels or be shared openly across the full consortium. The second option lets teams combine efforts speeding up results significantly.

Russia’s Power Grid Is Running Out of Room for AI’s Hunger

Russia’s energy planners have kept close watch on every energy-intensive sector of the domestic economy. The reason they are pushing ultra-small reactors is plain the ever-growing needs of data centers are already straining the country’s thermal and other conventional power plants. There is no critical shortage yet. But the trajectory is clear: either data centers get their own dedicated “batteries,” or industrial and residential power consumers will face cuts.

That is not just a scare story for effect. Russia’s largest single customer for data center construction is the state itself. A comprehensive industry development plan already exists and it calls for eight large new data centers, above what already exists. These facilities are specifically oriented toward scaling artificial intelligence systems. Two problems stand in the way. First, much of the specialized equipment is expensive and imported including from countries not particularly friendly toward Russia. That can be addressed through technology purchases. But without reliable industrial power supply, all that hardware is just a mountain of dead metal.

Beyond the Urals and Still Not Enough Power

New data center projects are already drifting away from Russia’s central regions squeezed by the power deficit. Plans now point east of the Urals, toward Siberia, where electricity has traditionally been cheaper and more available. But even that reserve is nearly exhausted and again, AI is the reason. Increasingly complex AI systems have driven up the power demands of server racks. Just a couple of years ago, a standard rack drew no more than 20 kilowatts. Today that figure has jumped to 50–60 kilowatts. Industry estimates suggest that by 2030, operational data centers may each require between 50 and 100 megawatts.

Existing power plants cannot fill that gap. Falling behind in the technology race, meanwhile, is treated as an existential threat. Consider this: the United States alone, through the single Stargate project, plans to invest no less than $500 billion into building data centers equipped with supercomputers. Under that initiative, technology giants including OpenAI, SoftBank, Oracle, MGX, and Nvidia are all pulling in the same direction.

America Is Using Old Nuclear Sites to Power the AI Boom

American energy planners are just as pragmatic as their Russian counterparts. They warned the country’s leadership and major corporations early a machine of that scale can only be powered by new nuclear plants. To speed things up, the Department of Energy pushed a clever solution through Congress. Legislation now allows new nuclear plants to be built on the sites of decommissioned power stations and large industrial facilities specifically for data centers. This bypasses the usual rounds of technological, geological, and other licensing steps. A list of 16 qualifying locations has been approved. Among them the National Nuclear Laboratory in Idaho, the Oak Ridge site where the top-secret Manhattan Project was once conducted, the uranium gas diffusion enrichment plant in Paducah, Kentucky, and the Savannah River nuclear site in South Carolina. Savannah River is notable it produced weapons-grade uranium and tritium for the American nuclear triad from the 1950s onward.

China Is Also Quietly Tightening Its Grip on AI Talent

No discussion of this race is complete without mentioning China. Beijing has increasingly come to view AI developers as a strategic state resource in recent years. Regional media report that key specialists in the sector now require special government permission to travel abroad. Cases have been documented involving employees of Alibaba and DeepSeek. Authorities also banned the two founders of the high-profile AI project Manus from leaving the country.

In today’s world, the race for digital technology matters more than the race for weapons. And the winner will be whoever has enough “batteries” to keep running.