China AI projects worth 4.6 trillion won rely on Nvidia chips
Despite U.S. export controls, Chinese infrastructure projects totaling over 23 billion yuan are incorporating Nvidia hardware.
The short version
- Total project costs for China's Nvidia-reliant AI infrastructure exceed 23 billion yuan (4.6 trillion won).
- The figure includes data center construction, power systems, and infrastructure, not just the price of the chips.
- Projects involve H100, H200, and H800 chips across tech giants, local governments, and universities.
- The data covers completed operations as well as projects currently in the planning or contracting stages.
Why it matters
This highlights how deeply Nvidia technology is embedded in China's broader AI infrastructure despite trade restrictions. It shows that the impact of export controls is complicated by the massive costs of the surrounding physical data centers.
The scale of Chinese AI investment
Recent findings indicate that artificial intelligence projects and equipment contracts in China relying on Nvidia high-performance chips exceed 23 billion yuan. This amount is roughly equivalent to 4.6 trillion won. These figures represent a significant portion of the country's technological expansion.
The scope of these projects is broad, covering both active operations and those that are still in the planning or contracting phases. This suggests that China is continuing to integrate high-end hardware into its long-term infrastructure goals. The data reveals a persistent demand for specific computing power despite existing trade barriers.
As reported by 재경일보, these projects are not limited to a few large corporations. They involve a wide variety of entities, including venture businesses and local government initiatives. This diversity indicates that the reliance on Nvidia technology is spread across many different sectors of the Chinese economy.
What the 23 billion yuan figure includes
It is important to distinguish between the price of the chips themselves and the total project costs. The 23 billion yuan figure does not represent actual sales of Nvidia chips or the purchase prices of hardware brought into the country. Instead, it accounts for the entire ecosystem required to run them.
The costs include the construction of data center buildings and the installation of power systems. Because high-performance AI chips require massive amounts of electricity and specialized cooling, the physical infrastructure is a major component of the budget. These expenses are bundled into the total project value for the computing centers.
By including infrastructure, the numbers provide a clearer picture of the total investment required to build an AI hub. It shows that the hardware is just one piece of a much larger industrial puzzle. The investment covers the physical space and energy needs necessary to support advanced machine learning tasks.
Specific projects and chip models
Several specific projects have been identified that utilize Nvidia's high-end processors, such as the H100, H200, and H800 models. For instance, a project in Xianyou County, Fujian Province, has a total cost of 6.2 billion yuan. The project lead specifically requested the use of H100 chips while recruiting partners for the data center.
Another computing center project in Shanxi Province lists H100 and H200 processors in its requirements. This project has a total budget of 3 billion yuan. These examples demonstrate how specific high-performance models are being targeted by regional developers for their computing capabilities.
In addition to new projects, some clusters are already operational. The Wuxi municipal government announced in March of last year that an H800-based computing cluster had begun operations. These instances show a mix of ongoing construction and active usage of Nvidia technology across different provinces.
Impact of export regulations
The presence of these chips in Chinese projects is notable because many of them are subject to U.S. export controls. These regulations are intended to limit the flow of high-performance computing hardware to certain regions. However, the data shows that Nvidia technology continues to appear in official documents and contracts across the country.
The persistence of these projects suggests that China is finding ways to incorporate these specific chips into its national AI strategy. Whether through various procurement channels or existing stock, the hardware remains a central component of the infrastructure. The documents reveal a significant role for Nvidia in the expansion of China's AI capabilities.
The data highlights a gap between the intent of export regulations and the reality of infrastructure development. While the U.S. seeks to restrict access, the Chinese government and local entities are still planning for systems that rely on these specific high-performance components. This creates a complex landscape for international trade and technology policy.
Broad reach across institutions
The reliance on Nvidia technology is not confined to the country's largest technology giants. The data shows that the chips are being integrated into projects involving local governments and universities. This suggests that academic and public sector research is also heavily dependent on these specific hardware models.
Computing centers under local governments are also part of this trend, indicating a decentralized approach to AI infrastructure. By distributing the technology across universities and regional centers, the country is building a wide network of AI capabilities. This broad distribution makes the infrastructure more resilient and accessible to different types of users.
The continued appearance of Nvidia chips in these diverse documents reveals the depth of the technology's integration. It shows that the hardware is a foundational element for both commercial and public-sector AI development. This trend will likely continue as China seeks to build out its domestic computing power.
What to watch
- How Chinese firms will navigate future U.S. export restrictions on H100 and H200 chips.
- Whether local Chinese chip manufacturers can provide a viable alternative to Nvidia for these large-scale projects.
Sources and links
- Reporting재경일보: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by 재경일보 · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
