The growing demand for high-performance computing solutions for artificial intelligence workloads continues to put pressure on the consumer GPU market. In a notable development that highlights this trend, a Vietnamese retailer has begun repurposing high-end gaming graphics cards for commercial AI applications, potentially exacerbating the already strained supply situation for gamers.
 |
|---|
| This display showcases high-end gaming graphics cards that are increasingly being repurposed for commercial AI applications, highlighting the impact on the consumer GPU market |
Vietnamese Retailer Repurposes Gaming GPUs for AI Workloads
Nguyencongpc, a Vietnamese PC retailer, has started building specialized AI server systems using multiple NVIDIA GeForce RTX 5090 graphics cards. These systems leverage the RTX 5090's substantial 32GB VRAM capacity, which makes it particularly valuable for memory-intensive AI tasks despite being marketed primarily as a gaming GPU. The retailer has showcased several configurations, including an impressive seven-GPU setup using Gigabyte RTX 5090 models, as well as more modest dual-GPU configurations featuring MSI Gaming Trio RTX 5090 cards.
 |
|---|
| The image shows high-end NVIDIA GeForce RTX graphics cards that Vietnamese retailer Nguyencongpc utilizes for building specialized AI server systems |
Technical Specifications of the AI Servers
The most powerful configuration features seven Gigabyte RTX 5090 GPUs in a single system, with an estimated power consumption approaching 4000W. This massive power requirement necessitates dual 2000W power supply units to operate reliably. For less demanding AI applications, Nguyencongpc also offers dual-GPU configurations built on AMD platforms, specifically using the ROG Crosshair X870E Hero motherboard paired with MSI Gaming Trio RTX 5090 cards. While the complete technical specifications of the seven-GPU systems remain undisclosed, these machines are clearly designed to handle substantial parallel processing workloads.
Pricing and Market Impact
The financial investment required for these AI servers is considerable. The dual RTX 5090 configurations are estimated to cost approximately US$10,000, while the seven-GPU servers likely command prices exceeding US$30,000. These estimates reflect the current market reality where individual RTX 5090 cards typically sell for US$2,500 or more, significantly above their US$2,000 MSRP. Premium models like the MSI Gaming Trio variant may fetch closer to US$3,000 per unit. The situation has been further complicated by supply constraints, recent US tariffs, and growing demand from both gaming and AI sectors.
Concerns About GPU Availability
This development has sparked concerns reminiscent of previous GPU shortages during cryptocurrency mining booms in 2018 and 2021. During those periods, consumer graphics cards became extremely difficult to obtain at reasonable prices as they were diverted to mining operations. The emergence of these AI server builds suggests a similar pattern may be developing, where high-end consumer GPUs are increasingly repurposed for commercial AI applications rather than their intended gaming use. This trend could further restrict availability for individual consumers, especially if more businesses follow Nguyencongpc's example.
Professional Alternatives
Industry observers note that dedicated professional GPUs with higher VRAM capacities already exist specifically for AI and computational workloads. Previous examples include specialized RTX 4090 variants with 48GB memory that were designed for such applications. However, the relative cost-effectiveness of consumer GPUs like the RTX 5090, despite their premium pricing, may continue to make them attractive alternatives for businesses building AI infrastructure, particularly for smaller operations that cannot afford enterprise-grade hardware.