Russia Eyes Modular Nuclear Reactor to Power AI Workloads

To meet rising energy demands from AI systems, Russian engineers are exploring small, scalable nuclear reactors and sodium-ion batteries for stable power delivery.

Artificial intelligence is not just transforming tech—it’s reshaping the energy sector. Russian experts are now discussing the creation of compact modular nuclear reactors tailored to the power needs of AI infrastructure.

According to Kristina Vishnevetskaya, development director at Ippon, large language models like ChatGPT and Gemini consume enormous amounts of electricity. Training a 200-billion-parameter model, for instance, can require up to 12 GWh—equivalent to the annual usage of thousands of homes. The International Energy Agency projects that by 2030, data centers worldwide will consume more power than the entire nation of Japan, with AI being the main driver.

To keep up with demand, major tech companies are reinvesting in nuclear and renewable energy. In Russia, this has led to proposals for rapidly deployable, small modular reactors designed to scale with AI-driven workloads.

Reliability is key: even a minor power disruption can wipe out hours of processing. Backup power systems must withstand voltage fluctuations and heat. Accelerators and cooling systems also rely heavily on consistent energy supply.

That’s where sodium-ion batteries come in. Operating in extreme temperatures—from -40°C to +50°C—they retain up to 80% of their capacity, resist overheating, and recover quickly after discharge. These features make them ideal for high-density computing environments where power reliability is non-negotiable.

What may sound futuristic today could become the backbone of AI infrastructure tomorrow.