When OpenAi blocked China’s access to its advanced artificial intelligence systems in July last year, Chinese developers did not shake. They preferred to bet on open source systems to continue their projects.
At the time, this meant resorting to the open source product of the goal, an American company.
But the following year, the Global Race by Advanced Changed significantly. Chinese companies like DeepSek and Alibaba have launched Open Source Systems that are among the best in the world.
China is rapidly reducing the distance to the United States in the dispute for technologies that mimic the human brain – and that didn’t happen by chance. The Chinese government has channeled resources for a decade to turn into AI power, using the same strategy it has made the country leading in electric vehicles and solar energy.
“China is applying state support throughout the AI technology chain, from chips and data centers to energy,” says Kyle Chan, an associate researcher at Rand Corp., a Think Tank.
In the last 10 years, Beijing has encouraged local companies to develop manufacturing capabilities in high -tech sectors previously dependent on imports. The strategy helped China stand out in global production – including batteries, solar panels and electric vehicles – and now extends to the fundamental blocks of AI: computational power, qualified engineers and data.
Continues after advertising
This industrial policy occurs while three US administrations tried to contain Chinese technological advancement, restricting the sale of Nvidia chips, the American AI’s main manufacturer.
On Monday, a. Nevertheless, companies like Huawei accelerate the development of alternatives with state support.
Unlike the US, where companies such as Google and Meta invest billions in data centers, in China the government finances much of the infrastructure and semiconductors necessary for AI.
Continues after advertising
Beijing also created a network of laboratories – many in partnership with giants such as Alibaba and Bytedance – to concentrate talents on advanced AI research.
In addition, local governments and banks were told to release credit for hundreds of startups. Since 2014, the government has invested about $ 100 billion in the semiconductor industry, and in April announced $ 8.5 billion in AI startups.
In the US, systems like ChatgPT were trained with data from websites such as Reddit and Wikipedia, inaccessible on the censored Chinese internet. For this, China has created official databases – such as the “state media values corpus” – ensuring that the contents follow the government’s censorship.
Continues after advertising
Companies such as Tiktok’s Bytedance have access to large -scale user data, which drives the development of their AI systems.
But this top -down approach is inefficient. Many startups dispute space in a fierce market, offering low prices models to attract engineers.
This model makes it difficult to redistribution of resources in the face of rapid technological changes. China has invested heavily in facial recognition and was surprised by the evolution of the Generative AI, such as the one used in the chatgPT.
Continues after advertising
“It’s hard to know where to invest and allocate resources. It is not stable like steel industry or shipbuilding,” says Chan.
Most of the public resources were destined to the Semiconductor Manufacturing International Corp. (SMIC), largest local chip manufacturer, which produces semiconductors for Huawei and Qualcomm. SMIC is developing AI chips to compete with those of Nvidia.
Although Huawei chips can meet some demands, they are still behind Nvidia’s. Large -scale production is also a challenge not yet exceeded by SMIC.
“The idea is: if we are blocked, there is a viable – yet less powerful – alternative for the sector to continue advancing, without stopping,” Chan explains.
c.2025 The New York Times Company
