Why Russia Is Turning to Chinese Chips to Power GigaChat

The move illustrates a broader geopolitical shift in the global AI industry: sanctions are reshaping supply chains, pushing Russia toward Chinese technology while China attempts to build alternatives to Western AI hardware.

Why Russia Is Turning to Chinese Chips

After Russia’s full-scale invasion of Ukraine in 2022, the United States, the European Union, the United Kingdom, and allies imposed sweeping sanctions and export controls targeting the Russian economy and high‑tech sectors.

These measures included restrictions on:

• advanced technologies and dual‑use goods with military or strategic value
• high‑tech components such as semiconductors and electronics
• financial access for major institutions including Sberbank

The EU’s sanctions specifically expanded export controls on dual‑use technologies to limit Russia’s access to critical advanced technologies used by its military and industrial sectors. The UK and other allies also banned exports of key technologies and sanctioned major Russian banks.

In parallel, export controls imposed by Western governments require licenses—often denied by default—for selling high‑tech components such as semiconductors to Russia.

As a result, Russia’s ability to legally purchase advanced GPUs from Western suppliers such as Nvidia has been severely restricted. For AI systems that require enormous computing power, this has created a major bottleneck.

China Becomes Russia’s Most Viable Hardware Partner

With Western supply chains largely closed, China is one of the few countries capable of providing AI‑focused hardware at scale. Russian officials have explicitly promoted deeper technological cooperation with China to reduce reliance on Western systems.

During a visit to China, Sberbank CEO German Gref said the bank hopes to gradually use Chinese microchips to run GigaChat, though the transition is still in progress.

The shift is strategic for both sides:

• Russia gains access to non‑Western AI infrastructure.
• China expands the reach of its semiconductor ecosystem and AI platforms.

However, the relationship is asymmetric. Russia depends heavily on external compute supply, while China still faces its own constraints in advanced chip manufacturing.

Which Chinese AI Chips Could Power GigaChat

Sberbank has not officially confirmed which processors it plans to use. But industry reporting and analysts frequently point to Huawei’s Ascend AI accelerator family as the most likely candidate.

Huawei has rapidly expanded production of these chips as part of China’s effort to build a domestic AI hardware stack. Production of the Ascend 910C accelerator is expected to reach roughly 600,000 units in 2026, with total Ascend chip output potentially rising to around 1.6 million dies across the product line.

The Ascend architecture is designed for AI training and inference workloads similar to Nvidia’s data‑center GPUs.

Still, demand inside China—from companies such as ByteDance and Alibaba—means supply could remain tight, making it difficult for foreign buyers like Russia to secure large volumes.

How Huawei’s Ascend Chips Compare With Nvidia’s H200

Even as Chinese chipmakers improve rapidly, most analysts say they still trail the leading Western accelerators.

For example:

• Nvidia’s H200 GPU offers about 141 GB of HBM3e memory and up to 4.8 TB/s of memory bandwidth, enabling high‑performance training and inference for large AI models.
• Huawei’s Ascend 910C delivers lower computing performance and memory bandwidth—about 12,032 total processing performance versus 15,840 for the H200, and 3.2 TB/s bandwidth compared with 4.8 TB/s.

That gap does not make Chinese chips unusable. Instead, organizations can compensate with larger clusters of GPUs or distributed computing techniques. But it means achieving the same performance often requires more hardware, energy, and engineering effort.

How Competitive GigaChat Is Globally

GigaChat itself is a technically capable model family, especially for Russian-language tasks. The system uses a mixture‑of‑experts architecture with 47 billion total parameters and about 13 billion active parameters, and internal benchmarks show strong results across tasks such as mathematics and coding.

However, independent comparisons suggest Russia’s models still trail the global frontier.

Some benchmarking analyses place leading Western models such as those from OpenAI and Anthropic at the top of global rankings, while Russian models—including GigaChat 2 Max—appear significantly lower in comparative evaluations.

Analysts have also noted that GigaChat ranks behind several Western and Chinese models on public evaluation platforms and has yet to match the capabilities of top systems like GPT‑series models, Claude, or DeepSeek’s latest releases.

In practice, GigaChat’s strength is localization rather than global leadership: it performs well for Russian-language applications and domestic enterprise use cases.

What This Means for the Russia–China Tech Partnership

Sberbank’s chip strategy reflects a broader geopolitical trend reshaping the AI industry.

Key dynamics include:

• Sanctions-driven realignment: Russia is forced to rely on non‑Western technology suppliers.
• China’s push for semiconductor independence: export controls have accelerated Chinese investment in domestic AI chips.
• Growing technological blocs: AI infrastructure is increasingly splitting into Western and China‑centered ecosystems.

Yet the partnership also reveals limitations. China’s most advanced chips still lag the top Nvidia hardware, and domestic demand already absorbs much of the available supply.

For Russia, that means Chinese technology is less a perfect replacement than a necessary workaround.

The Bottom Line

Sberbank’s plan to power GigaChat with Chinese chips highlights how geopolitics is reshaping AI development.

Sanctions have effectively cut Russia off from the Western semiconductor supply chain, pushing it toward Chinese hardware such as Huawei’s Ascend accelerators. While these chips enable continued AI development, they still trail the performance and ecosystem advantages of Nvidia’s leading GPUs.

The result is a deeper Russia–China technological alignment—but one defined as much by constraints as by cooperation.