Western export controls on advanced semiconductors were designed to starve China and Russia of the chips they need for artificial intelligence and military applications. Instead, they are accelerating the creation of a parallel supply chain that could permanently reshape the global chip market.
The emerging dynamic was captured by a single remark in May 2026. German Gref, chief executive of Russia’s Sberbank, told state broadcaster Channel One that he hoped to run the country’s flagship AI model, GigaChat, on Chinese-made processors. The statement, made during President Vladimir Putin’s two-day state visit to Beijing, highlighted a structural shift: US sanctions are creating conditions for a Sino-Russian chip trade alliance.
China’s chip production surge
Far from collapsing under sanctions pressure, China’s semiconductor manufacturing has expanded dramatically. Tom’s Hardware reports that Chinese chip production jumped 40 percent in the first quarter of 2024 alone. Much of this growth is in legacy-node chips, 28 nm (28-nanometer) and larger, that are exempt from the most stringent export controls but are essential for everything from automobiles to industrial equipment.
However, China’s most advanced domestic foundry, Semiconductor Manufacturing International Corporation (SMIC), continues to struggle with production of 7 nm class chips on deep ultraviolet (DUV) lithography equipment. The yield and cycle-time constraints mean that every advanced chip SMIC can produce faces overwhelming domestic demand from Chinese tech giants.
Russia’s Chinese chip problem
This is where Russia’s predicament becomes acute. Sberbank wants Huawei’s Ascend 950PR AI accelerator for GigaChat, which sits between Nvidia’s H100 and H200 in inference performance and outpaces the restricted H20 by a claimed factor of 2.8. But every Ascend 950PR Huawei can produce, and the company is targeting 750,000 units in 2026 for approximately US$12 billion (roughly £9.3 billion) in AI chip revenue, is already spoken for. ByteDance alone committed US$5.6 billion in orders for the 950PR earlier this year, and Alibaba and Tencent have their own large allocations.
Sberbank acquired a 41.9 percent stake in Element, Russia’s largest electronics producer, in January 2026 for 27 billion roubles (US$356 million approximately £275 million). But Element’s production focuses on defense and industrial applications, not data-center AI accelerators. Russia’s most advanced domestic chipmaking targets 65 nm lithography by 2030, roughly 25 years behind the leading edge.
“The US has created a situation where China and Russia are natural chip-trade allies,” said one semiconductor analyst who tracks export controls. “But China’s own demand is so intense that there may not be much supply left for Russia.”
Policy whiplash
The Trump administration’s approach to chip sanctions has been anything but consistent. Policies have shifted between tightening and loosening restrictions, sometimes within the same month. In the first half of 2026, the US licensed Nvidia H200 AI chip exports to China while simultaneously expanding sanctions on Chinese equipment makers, creating confusion among both allies and adversaries about the stability of US semiconductor trade policy.
The joint declaration signed during Putin’s May 2026 visit to Beijing called for closer bilateral cooperation in AI and endorsed China’s proposal for a global AI governance framework. Whether that translates into actual chip allocation for a sanctioned Russian buyer competing against the backbone of China’s internet economy remains an open question.
For now, the sanctions regime has accomplished one thing that is hard to reverse: it has convinced both China and Russia that they need to build a semiconductor supply chain that cannot be cut off by Washington. And they are building it together.
Sources: While the U.S. flip-flops on chip sanctions, China is building its own chip supply market (Tom’s Hardware, July 9, 2026); Sberbank bets on Chinese chips for Russia’s AI race (Cryptonomist, May 20, 2026); Russia’s Sberbank wants Chinese chips for GigaChat AI (Tom’s Hardware/Yahoo Tech, May 20, 2026)