Russia's Sberbank says it hopes to use Chinese chips to power its flagship GigaChat AI model, as sanctions continue to block the country's access to hardware
Context & Ripple Effects
US restrictions have limited Russia’s access to major US AI accelerators since 2022, while China has been both Russia’s largest electronics supplier and a target of US pressure to align with sanctions.
Related coverage also shows Chinese firms adapting to export controls through domestic-chip efforts, modified product offerings, and workarounds for restricted Nvidia hardware. Sberbank’s position places a major Russian AI deployment inside that broader supply-chain realignment.
First-order effects
- Sberbank must assess Chinese chips as a practical compute base for GigaChat, including their availability and ability to support the model’s training and inference workloads.
- The bank’s AI hardware roadmap becomes more dependent on Chinese suppliers because sanctioned US-origin alternatives remain inaccessible.
Second-order effects
- A credible Sberbank deployment would create another demand signal for Chinese AI-chip vendors and the software, integration, and support layers around their hardware.
- US restrictions on Russia and pressure on Chinese suppliers make this channel vulnerable to tighter enforcement or supplier caution, potentially limiting the chips and configurations Russian buyers can obtain.
Third-order effects
- If major Russian AI users can run important models on Chinese hardware, sanctions may shift AI infrastructure toward a more regionally separated stack rather than halt deployment outright.
- The durable constraint may move from access to a particular accelerator to ecosystem maturity: whether non-US chips can sustain dependable model development, tooling, and scale across sanctioned markets.
The trend: This is one data point in the export-control-driven bifurcation of AI compute, with Chinese hardware becoming a more important fallback for markets cut off from leading US chips.