Russia Seeks Chinese AI Chips Amid Sanctions, Faces Supply Competition from China's Tech Giants
#Chips

Russia Seeks Chinese AI Chips Amid Sanctions, Faces Supply Competition from China's Tech Giants

Chips Reporter
3 min read

Russia's largest bank, Sberbank, is pursuing Chinese processors for its GigaChat AI model as Western sanctions restrict access to advanced hardware. However, the bank faces significant challenges in securing Huawei's Ascend 950 chips, which are in high demand among China's own tech giants.

Russia's largest financial institution, Sberbank, has announced its intention to utilize Chinese-made processors to power GigaChat, the country's flagship AI model. According to CEO German Gref, the bank is seeking alternatives to Western AI hardware amid ongoing international sanctions that have restricted Russia's access to advanced computing components.

Putin and president xi

During President Vladimir Putin's recent two-day visit to Beijing, Gref revealed Sberbank's plans on Russian state broadcaster Channel One. While he didn't specify which Chinese chips the bank is targeting, industry analysts point to Huawei's Ascend 950 family as the most likely candidate. These processors, however, face intense competition from China's domestic tech giants, creating a challenging supply landscape for Russian buyers.

The Ascend 950PR, Huawei's inference-optimized processor, sits between Nvidia's H100 and H200 in performance benchmarks. Huawei claims the chip outperforms the restricted H20 by a factor of 2.8 times, though this figure remains unverifiable as Hopper-era hardware lacks native FP4 support. The 950PR is specifically optimized for the prefill stage of serving large language models, making it suitable for inference workloads like those powering GigaChat.

Production constraints present significant hurdles for both Huawei and potential customers like Sberbank. Huawei is targeting 750,000 units of the 950PR in 2026, but manufacturing at SMIC is hampered by weak yields on the foundry's 7nm-class DUV process. Additionally, the production cycle extends approximately eight months from wafer start to finished processor, creating a bottleneck that limits available supply.

The competitive landscape for these chips is particularly challenging for Russian entities. ByteDance alone committed $5.6 billion in orders for the Ascend 950PR earlier this year, while other major Chinese tech companies like Alibaba and Tencent are also vying for limited production capacity. Every chip Huawei produces faces overwhelming domestic demand from companies that collectively represent the backbone of China's internet economy.

For training-focused AI workloads, Huawei's 950DT processor offers more substantial capabilities, featuring 144 GB of the company's proprietary HiZQ 2.0 memory with 4 TB/s bandwidth. However, this more powerful variant isn't expected to ship until Q4 2026, further complicating Sberbank's timeline for developing a fully Chinese-supplied AI compute stack for GigaChat.

Russia's current AI infrastructure relies on a combination of stockpiled Western GPUs, Chinese alternatives, and domestic production that hasn't yet reached competitive capability for frontier AI workloads. Sberbank's strategic acquisition of a 41.9% stake in Element, Russia's largest electronics producer, for 27 billion rubles ($356 million) signals the bank's commitment to developing domestic capabilities. Element manufactures integrated circuits and semiconductor devices that account for roughly half of Russia's microelectronics output, though its production focuses on defense and industrial applications rather than data-center AI accelerators.

Russia's most advanced domestic chipmaking roadmap targets 65nm lithography by 2030, a process node that is approximately 25 years behind the global leading edge. This technological gap underscores Russia's dependence on foreign semiconductor suppliers for advanced AI workloads in the near to medium term.

The joint declaration signed by Putin and Xi during the Beijing visit called for closer bilateral cooperation in AI and backed China's proposal for a global AI governance body. While this political alignment may create opportunities for increased semiconductor trade between the two nations, whether it translates into actual chip allocation for a sanctioned buyer vying for space on Huawei's already oversubscribed order book remains uncertain.

The situation highlights the increasingly complex semiconductor supply chain challenges emerging from geopolitical tensions. As Western sanctions continue to restrict Russia's access to advanced computing hardware, the country is forced to navigate an increasingly competitive landscape for alternative suppliers, particularly in China where domestic demand already outstrips available supply.

Comments

Loading comments...