Russia Turns to Sberbank's GigaChat AI to Design Combat Jets

Using an LLM for early-stage aerodynamics

United Aircraft Corporation, the aerospace conglomerate behind the Su-57 and Su-35 combat jets, is deploying Sberbank's GigaChat AI to explore conceptual aircraft layouts before traditional computer-aided engineering begins.

Instead of drafting finished blueprints, the large language model is being evaluated to sift through preliminary configurations and speed up the early conceptual phase. UAC general designer Sergey Korotkov indicated that AI will only be introduced where tangible engineering requirements justify it, noting that digital systems allow engineers to handle larger data sets. More intensive physical simulations, such as aerodynamic drag computations and structural stress testing, remain outside the initial trial.

The rush to replace Western CAD suites

The experimental push into generative tools reflects necessity rather than luxury. When foreign sanctions hit in 2022, aerospace standard-bearers Autodesk, Dassault Systèmes, and Siemens halted Russian operations. That abruptly severed UAC from product lifecycle suites that global aviation relies on.

To adapt, UAC overhauled approximately 4,000 engineering workstations across its design bureaus, replacing nine critical software modules responsible for managing data throughout design and production. The corporation claims these localized workflows compressed certain calculations for cabin climate controls and hydraulic layouts from several months down to days. Independent verification of these claims is impossible under current reporting constraints.

The shift coincides with broader scrutiny over sovereign tech ecosystems, echoing debates seen when the Pegasus spyware founder warned governments to ditch US AI infrastructure to preserve strategic independence.

Severe domestic setbacks persist

Software migration is only half the battle. Moscow's broader aviation sector remains hobbled by supply-chain shortfalls as it attempts to replace thousands of foreign-made physical components.

The civilian MC-21 passenger jet serves as a prime example of the friction. Russian President Vladimir Putin acknowledged that the program fell roughly two years behind schedule, largely due to complications engineering an all-domestic wing structure. Certification for an import-free variant of the MC-21 has now been pushed back to late 2027, highlighting that software workarounds can only do so much when manufacturing pipelines face physical bottlenecks.

Availability and international impact

Western export controls strictly bar Russian defense contractors from accessing modern American and European computing architectures, CAD software, and enterprise AI tools. Consequently, the technologies powering this initiative—such as Sberbank's GigaChat—are developed exclusively for Russia's domestic market and are not commercially available in the US, UK, or European Union.

While Western aerospace contractors routinely integrate machine learning into conceptual airframe design, Russia's reliance on a banking sector language model highlights how isolated the country's military-industrial complex has become from specialized commercial engineering ecosystems.

Generative models can speed up early brain-storming in design labs, but translating concepts into flying hardware demands rigorous engineering tools that language models cannot replace. UAC's experiment demonstrates the lengths to which sanctioned defense firms must go to fill the void left by Western industry leaders.