Chinese AI Models Utilize Restricted Nvidia Chips
Research indicates Chinese firms trained advanced systems using hardware banned by the U.S. government.
Artificial intelligence models developed by Chinese companies were trained using Nvidia chips subject to United States export restrictions. Technical analysis of these model structures revealed configurations compatible with the use of Blackwell processors in Chinese laboratories.
Access to these components persists despite American government efforts to limit China's technological advancement in the AI sector. Estimates suggest that approximately two million advanced chips entered Chinese territory between 2024 and 2025 through schemes that conceal the final destination of the products.
Some of the hardware may have been used in data centers located outside of China to bypass current prohibitions. Data points to Malaysia receiving a significant volume of Nvidia accelerators over the past year, making it a strategic hub for accessing computing power.
The situation raises concerns regarding the effectiveness of export controls and the corporate responsibility of the manufacturer. The U.S. Congress is evaluating stricter rules to prevent cutting-edge technologies from strengthening the strategic capabilities of geopolitical rivals.
The spread of this technology could accelerate innovation among software and infrastructure firms, increasing pressure on the global semiconductor supply chain.
Spotted an error? Write to suporte@criptoico.com.br.
