Rising memory prices are leading to a number of changes at NVIDIA. Up until now, the Greens have supplied both GPU and VRAM, but this is no longer the case. The Chameleon doesn’t seem to want to absorb memory costs. If this practice ensured a certain stability for smaller manufacturers, they are now exposed to greater risks.
No more GPU VRAM bundles from NVIDIA?
Once again, the artificial intelligence business is putting upward pressure on RAM and VRAM prices. All this would force NVIDIA to change its habits when it comes to delivering memory GPUs. Until now, GPUs and memory have been delivered in bundles from the Chameleon. There are several advantages to this: more controlled specifications for NVIDIA, and on-point negotiations for its partners. Finally, the rest of the board and its components could be purchased as normal from a variety of retailers.
However, with the change in policy that seems to be taking shape, the Greens will only be selling GPUs. As a result, large partners should have no trouble finding a source of memory. This is a little less true for secondary players in the sector, due to less established relationships with memory producers (Samsung, SK Hynix or Micron).
On the other hand, it could also become very interesting since, if the Chameleon no longer sells the memory with the GPU, we may end up with cards delivered with VRAM “out of spec”. Currently, the standard with an RTX 5070 is 28 Gbps memory, but potentially, a brand could release a card with 30 Gbps memory, as with an RTX 5080 for example. It remains to be seen how much control the Greens have over this.
