Nvidia may have stopped supplying VRAM together with its GPUs to board partners, according to a new rumor. With the global memory shortage worsening due to manufacturers prioritizing AI-related demand, Nvidia is reportedly telling AIBs they must now source their own video memory instead of receiving bundled GPU-and-VRAM packages.
Nvidia doesn’t produce VRAM itself — memory chips come from manufacturers like Samsung, Micron, and SK Hynix. Normally, Nvidia either bundles this memory when selling GPUs to partners or AIBs procure it themselves. The claim suggests that the shortage has become severe enough that Nvidia no longer sees bundling VRAM as realistic.
For major AIBs, this shift likely isn’t a major obstacle, as they already have relationships with memory suppliers and can follow Nvidia’s reference requirements. But for smaller vendors, being forced to buy cutting-edge GDDR independently could further squeeze profits in an already difficult market. Some observers note that similar pressures contributed to EVGA’s exit from the GPU space, raising concerns that smaller manufacturers may struggle to survive the memory crunch.
However, all of this remains unconfirmed — the information comes from a single leak. If true, Nvidia could be losing potential revenue from selling memory at a markup, suggesting the shortage is so severe that bundling simply isn’t possible anymore. For now, the rumor reflects growing tension across an industry dominated by AI-driven memory demand.
