SMIC, China's largest chip manufacturer, has developed a 7nm process technology that represents a significant technological leap from its previous 14nm node. When compared to SMIC's 14nm technology, the N+1 process (their first-generation 7nm) reduces power consumption by 57%, increases performance by 20%, and reduces logic area by up to 63%1. Despite U.S. sanctions restricting access to advanced chipmaking equipment, particularly ASML's extreme ultraviolet (EUV) lithography machines, SMIC has managed to produce 7nm chips using deep ultraviolet (DUV) lithography with multi-patterning techniques23.

The technical specifications of SMIC's 7nm process show interesting design choices. While some key measurements like Fin Pitch are larger than TSMC's 10nm process, SMIC has implemented advanced Design Technology Co-Optimization (DTCO) features typically found in more advanced nodes, including a 6-track height cell design that TSMC didn't implement until their 7nm process45. However, SMIC's 7nm production faces challenges with yields reportedly below 50% (compared to industry norms of 90%) and significantly higher costs—approximately 10 times the price of equivalent TSMC 7nm chips67. Despite these limitations, SMIC's achievement has strategic importance for China's semiconductor self-sufficiency efforts, with the technology being used in Huawei's Kirin processors for devices like the Mate 60 Pro38.

The US government's latest export restrictions on Nvidia's H20 AI chips represent a significant escalation in semiconductor export controls targeting China. On April 9, 2025, US authorities informed Nvidia that it would require licenses to export H20 chips to China, citing concerns about their potential use in Chinese supercomputers1. This decision is expected to cost Nvidia approximately $5.5 billion in charges related to inventory and purchase commitments21, with analysts predicting the company's China business could "fall to nearly zero" from about 10% of total revenue last year3.

These restrictions are part of a broader pattern of US efforts to limit China's access to advanced AI technology, following earlier bans on Nvidia's more powerful A100, H100, A800, and H800 chips34. The controls have created an opportunity for Chinese firms like Huawei and domestic GPU startups to capture market share5, while simultaneously pushing Chinese tech giants including ByteDance and Tencent to accelerate their shift toward domestic semiconductor alternatives3. The restrictions highlight the ongoing technology competition between the US and China, with both nations viewing AI as critical for future economic and military power2.

The Ascend 910C represents Huawei's most ambitious AI chip to date, reportedly achieving 800 TFLOP/s at FP16 with approximately 3.2 TB/s memory bandwidth—roughly 80% equivalent to Nvidia's H100 performance1. DeepSeek's testing revealed the chip delivers about 60% of the H100's inference performance in practical applications234. This performance is particularly impressive considering Huawei's technological constraints.

The architecture employs a clever design approach, essentially combining two Ascend 910B chips interconnected via silicon interposers on an organic substrate15. Manufactured using 7nm process technology from both stockpiled TSMC chips and SMIC's domestic production capabilities15, the Ascend 910C consumes approximately 310 watts6 while supporting high bandwidth memory (HBM2e) and compatibility with frameworks including MindSpore, TensorFlow, and PyTorch64. While not yet matching Nvidia's dominance in AI training workloads, the 910C represents a significant step toward technological self-sufficiency for China's AI ambitions.