NVIDIA GeForce RTX 5090 128 GB GPU Leaked: Custom AI Memory Configuration Priced at $13,200 Each

NVIDIA’s GeForce RTX 5090 GPU is making headlines with custom memory implementations that boost capacity up to a remarkable 128 GB. This substantial expansion comes at a significant cost, attracting attention from AI professionals and enthusiasts alike.

Chinese AI Manufacturers Innovate with Custom NVIDIA GeForce RTX 5090 GPUs: 128 GB Memory Options

In response to the surging demand for AI-focused hardware, numerous factories across China have focused on producing tailored AI GPUs derived from NVIDIA’s GeForce RTX lineup. This trend emerged alongside the AI boom, paving the way for advancements in GPU technology. Previously, models such as the RTX 4090 offered impressive memory configurations, with options for 48 GB and even 96 GB, but now the RTX 5090 is stepping into the spotlight with customizable memory options.

These assembly facilities employ innovative methods to modify GPU and memory configurations, utilizing components from existing gaming graphics cards like the GeForce RTX 5090. By integrating these elements into customized printed circuit boards (PCBs), factories can achieve dual-sided memory setups, significantly expanding memory capabilitites. It’s important to note that these modifications aren’t trivial swaps; they require careful handling of firmware and BIOS to ensure full compatibility.

NVIDIA GeForce RTX 5090 128 GB GPU Spotted: Custom Memory, Designed For AI Workloads & Priced At $13, 200 Per Piece — Image Source: I_Leak_VN

According to insights shared by I_Leak_VN, the NVIDIA GeForce RTX 5090 is the latest model to undergo such custom memory enhancements. While the standard variant features a modest 32 GB of GDDR7 memory, the modified version offers an astonishing 128 GB frame buffer—quadrupling the original memory capacity. Although specific details about the PCB configuration remain scarce, projections suggest that the enhanced card employs higher-capacity GDDR7 memory chips that enable a single-sided 64 GB setup, which is then doubled for the 128 GB capacity with an innovative dual-sided PCB layout.

The notable increase in memory capacity significantly impacts the pricing structure. While the standard NVIDIA GeForce RTX 5090 typically retails for $1, 999 (MSRP) and is generally sold in the $2, 500 to $3, 000 range, the custom model with 128 GB of memory carries a price tag of approximately $13, 200. This marks a staggering 6.6 times increase in price for four times the memory.

Worker assembling a GPU with precision on a green workbench.

Despite the steep cost, the pricing aligns with industry standards. For comparison, the RTX PRO 6000 Blackwell with a 96 GB VRAM configuration is priced around $10, 000, making the RTX 5090’s 128 GB variant only 33% more expensive while offering a 33% greater memory capacity. It’s crucial to highlight that the RTX PRO 6000 is an officially sanctioned NVIDIA product, whereas the 128 GB GeForce RTX 5090 is a bespoke modification without NVIDIA’s endorsement. It primarily caters to AI specialists who require substantial VRAM for advanced computations, a necessity not currently justified in conventional gaming setups.

This increasing interest in custom VRAM configurations is especially pronounced in Asian markets, where access to high-quality AI GPUs is often limited due to elevated costs, tariffs, and regional trade restrictions. As the demand for powerful AI processing capabilities continues to grow, further insights into this custom model will be explored.

Source & Images