Russian oil and gas buys AI hardware instead of subscriptions: spending on AI software fell 63%

This is not a story about one industry. Pavel Rastopshin, CEO of industrial integrator Ultimateg, says the situation is identical everywhere: metallurgy, energy, and construction are behaving the same way.

They ran the numbers. According to IT holding T1, in 2025 oil and gas invested 63% less in AI software and 84% more in hardware.

Foreign cloud AI services have become multiples more expensive, while in-house infrastructure pays for itself in 12–18 months under a stable workload. That is how Sergey Karpovich, deputy head of T1's AI division, explains the shift.

No one is giving up foreign models, though. They are moving them onto their own servers and running them within the corporate perimeter.

The law is pushing it along. The State Duma adopted the AI law (bill No. 1271570-8) on 8 July 2026; its main provisions have been in force since 1 September 2026. A model is considered "sovereign" if its data is stored only in Russian data centers. A "national" model may use foreign components under an open license.

The industry's IT budget is growing, meanwhile. In 2025, Russia's oil and gas sector spent 162 billion rubles on IT products, versus 135 billion a year earlier (estimate by the Sisoft Development analytics center). What is changing is not the size of the budget, but where it goes.

Karpovich calls demand for AI accelerators insane. He also says that without use cases with measurable production impact, the purchased hardware may fail to pay for itself.

By 1 January 2030, critical infrastructure facilities must switch to trusted software and hardware systems (government resolution No. 1912).