Key facts
- China's State Council reviewed artificial intelligence development, calling for accelerated breakthroughs and improved safety and ethical standards.
- Nearly 20 global semiconductor companies are expected to implement price increases starting July 1.
- The price hikes are driven by increased costs for wafer foundries and raw materials, coupled with high demand for AI data center chips.
- China's AI progress is built on a decade of coordinated national strategy, industry innovation, and investment in infrastructure and talent.
- Chinese AI models like DeepSeek-V3 and Qwen3 have demonstrated strong capabilities in language understanding and reasoning.
- Chinese AI firms are innovating despite hardware constraints, focusing on efficiency and open-source development.
China's State Council, led by Premier Li Qiang, has reviewed the nation's artificial intelligence development, emphasizing the urgent need for accelerated breakthroughs and the establishment of robust safety, ethical, and testing standards. This strategic push comes as global semiconductor companies are reportedly preparing for a new round of price increases on July 1. Nearly 20 firms are expected to raise prices, citing rising costs in wafer foundries and raw materials, alongside a surge in demand for power chips essential for AI data centers.
China's advancements in AI are not sudden but are the result of a decade-long coordinated national strategy, initiated with the 2017 "Next Generation Artificial Intelligence Development Plan." This directive fostered significant policy support and investment across various government levels, driving innovation in infrastructure and talent development. Universities have played a crucial role, with institutions like Tsinghua, Peking University, and Shanghai Jiaotong University serving as hubs for research and commercial spin-offs, nurturing companies such as Zhipu AI and Baichuan.
By 2022, China had become a global leader in AI patent filings, significantly closing the gap in top-tier research. In 2024 and early 2025, models like DeepSeek-V3 and Qwen3 have demonstrated impressive performance in language understanding, reasoning, and multilingual capabilities, partly due to architectural innovations. Despite limitations in chip access, Chinese AI firms have shown remarkable ingenuity. Developers have focused on efficiency and open-source collaboration, with DeepSeek notably training its R1 model on approximately 2,000 GPUs for a budget of $5.6 million. Companies like Huawei continue to develop domestic chip designs, such as the Ascend series, while SMIC progresses in semiconductor fabrication.