China is pushing hard to make its own advanced chips for artificial intelligence. This comes as top companies like Alibaba look for ways around rules from the United States. These rules limit access to powerful chips from companies like Nvidia. The goal is to build a strong local industry and compete in the fast-growing AI market.
Alibaba is leading this charge. The company recently showed off a new chip. This chip is designed for AI “inference” tasks, which means it helps AI models make decisions once they’ve been trained. It’s a key part of Alibaba’s plan to make up for not being able to buy as many Nvidia chips. Alibaba used to be a huge customer for Nvidia. Now, it heads a group of businesses trying to fill the gap left by trade restrictions. Alibaba’s cloud business grew by 26% in the second quarter, according to a report from The Wall Street Journal. This growth was boosted by more demand for AI services. Alibaba’s CEO, Eddie Wu, said that “AI plus cloud” is now one of the company’s two main growth engines, right alongside its e-commerce business.
Other Companies Step Up
Other Chinese tech companies are also making big moves. In July, Shanghai-based MetaX released a chip that could replace Nvidia’s H20 model. While MetaX’s chip uses more electricity, it has more memory. This helps it perform better for certain inference tasks. The company plans to start mass production in the coming months.
Beijing-based Cambricon Technologies is another important player. Its Siyuan 590 chip helped the company bring in $247 million between April and June. This success pushed Cambricon’s market value past $87 billion. Still, the company has asked investors to be careful and not get overly excited.
Huawei’s Role
Huawei has become a symbol of China’s effort to be self-sufficient in technology. Its Ascend chips are used in large systems of 384 units. Huawei promotes these as powerful alternatives, even though they use a lot of energy. Ren Zhengfei, Huawei’s founder, has stated that these chips perform as well as the best available, playing down any concerns about technical limits.
However, some private companies have not fully adopted Huawei’s chips. They argue that Huawei also offers cloud services, making them a direct competitor. Local engineers also point out problems with Huawei’s processors when training large AI models. Issues like overheating and shutdowns during long operations have been reported.
Facing Manufacturing Challenges
Making these advanced chips is China’s biggest challenge. Local factories still rely on older foreign equipment and less powerful homegrown machines. This limits how many chips they can produce. To get around this, companies like MetaX are using older technology and combining smaller chips to improve performance.
The central government is stepping up its support. Early in 2025, Beijing announced a new investment fund of $8.4 billion. This money is meant to strengthen the entire AI supply chain. The plan aims to cut down reliance on American chipmakers and help new startups create fresh AI models and hardware.
The Future Against Nvidia
Despite these steps, there’s still a big gap between China and companies like Nvidia, especially for training very large AI models. Reports show that Alibaba and other firms are focusing more on “inference” tasks. These tasks have lower technical demands than training new models from scratch.
Even so, there’s a lot of hope. DeepSeek, a startup creating AI models similar to OpenAI’s, has suggested its software could work well with local chips. This news has excited Chinese markets. Kevin Xu, founder of Interconnected Capital, believes these new approaches could help Chinese developers close the technology gap with the United States “sooner than many believe.” This could challenge Nvidia’s leading position both at home and worldwide.