The Chinese chip designer Cambricon achieved a profit of 1 billion renminbi ($140 million) in the first half of the year and doubled its market capitalization to 580 billion RMB within a few weeks. The surge in stock price reflects investor hopes that the company could benefit from the state-supported push for technological self-sufficiency and establish itself as a key alternative to Huawei in the domestic AI chip market.

As of 2019, Cambricon was on the brink after losing Huawei as its main customer: 98% of its revenue at that time came from licensing smartphone technologies to the tech company. The breakup forced the company to shift to cloud accelerators and edge modules – a move that brought Cambricon directly into competition with Huawei.

Since then, the company has consistently focused on software. Between 2020 and 2024, 5.6 billion RMB were invested in research and development, particularly in interfaces that facilitate running models trained on Nvidia GPUs on Cambricon's Siyuan chips. This compatibility, confirmed among others by engineers at ByteDance, is considered a decisive advantage over Huawei's Ascend series, which has so far been difficult to integrate into existing systems due to software issues.

Goldman Sachs analysts expect Cambricon's revenue to rise to 13.8 billion RMB by 2026, and the market share to grow from today's 3% to 11% by 2028. Customers like Alibaba, Tencent, or Baidu – who themselves compete with Huawei in other sectors – have a strong interest in supporting the rise of a challenger.

The growth story, however, depends on production capacity. Cambricon relies on SMIC's manufacturing partners to deliver 7nm nodes. Beijing has instructed SMIC to allocate a larger share of this capacity to Cambricon, instead of almost exclusively serving Huawei. According to industry sources, SMIC plans to double its 7nm capacity next year.

Despite the tailwind from industry and politics, Huawei remains the top player in the market. The company is working on a more flexible chip architecture and, with NV-Link, is the only one in China offering a real alternative to Nvidia's high-performance technology for training and networking AI models.

Cambricon, on the other hand, is currently struggling with delivery bottlenecks: Large parts of production already went to ByteDance in the first quarter, according to industry insiders. Whether the company can maintain its position will depend significantly on how quickly SMIC expands capacities.