In the aftermath of the cryptocurrency mining decline, there has been a notable shift in the use of Nvidia Geforce RTX 4090 graphics cards, as they are being repurposed for artificial intelligence (AI) infrastructure. Specialized firms are procuring these GPUs in large quantities to adapt them for use in AI farms and dedicated workstations.
This adaptation process is particularly prevalent among smaller Chinese manufacturers who are investing considerable effort into retrofitting these graphics cards, which were initially designed for gaming, for AI-specific applications. The conversion process involves disassembling the graphics cards, desoldering the GPU and memory components, and then reassembling them onto custom-designed circuit boards suited for AI operations. This indicates a strong market demand and profitability, even when sourcing the GPUs requires unconventional procurement methods.
These activities are taking place despite Nvidia's design specifications, which typically discourage the use of direct heat exhaust (DHE) coolers, commonly referred to as blower coolers, in high-performance gaming GPUs. These coolers are not generally favored for such GPUs due to potential incompatibilities with server casings and less optimal airflow management, which could result in heat recirculation issues when multiple GPUs are used concurrently in workstations.
Despite these challenges and the increased costs associated with the Nvidia Geforce RTX 4090 GPUs—exacerbated by the US ban on sales to China—the modification process for AI deployment remains economically feasible. Chinese circuit board and graphics card manufacturers continue this practice, focusing on the growing demand for AI-optimized graphics cards that can be integrated into larger systems. This trend, however, is causing an uptick in prices and affecting availability for the average consumer, who faces increased costs due to the surge in demand from these specialized applications.
