Micron is killing its 2GB GDDR7 memory chips as booming AI demand makes higher-density products a much more attractive use of its wafers.
According to HWBuster, the memory maker has begun telling graphics card partners that it is discontinuing production of its 16Gb, or 2GB, GDDR7 dies, according to industry reports.
Instead, Micron is shifting production towards 24Gb GDDR7 dies, which provide 3GB of capacity per chip. The move comes as memory manufacturers wrestle with tight production capacity and ravenous AI demand.
Moving from 2GB to 3GB packages gives board makers much more flexibility. A graphics card using eight 2GB packages, for example, gets 16GB of memory. Swap those for eight 3GB packages and the same package count provides 24GB.
That does not mean the graphics card will suddenly receive a generous dollop of extra VRAM. GPU memory configurations depend on memory buses, package layouts, controller support, pricing and whatever segmentation wheeze the manufacturer has dreamt up.
Micron’s retreat from 2GB parts appears to be another symptom of memory manufacturers prioritising products that produce better returns from scarce fabrication capacity.
AI accelerators are swallowing vast quantities of high-bandwidth memory, while server DRAM demand has pushed memory suppliers toward the parts with the fattest margins.
The consequences have already spilt into consumer hardware. Memory pricing has climbed as manufacturers allocate capacity towards AI infrastructure rather than fighting to supply every last low-margin PC component.
For graphics card makers, the disappearance of Micron’s 2GB GDDR7 chips would not mean an immediate shortage. Samsung and SK Hynix also manufacture GDDR7, giving GPU vendors other places to shop.
However, removing one source of 16Gb GDDR7 reduces flexibility at a time when the graphics industry is moving deeper into the new memory standard. GDDR7 offers substantially more bandwidth than GDDR6 while improving signalling and power efficiency.
Micron has been pushing 24Gb GDDR7 as part of that transition, with the denser devices allowing higher VRAM capacities without forcing graphics card designers to plaster additional memory packages across their boards.
For gamers hoping that cheaper memory might eventually persuade GPU makers to stop being stingy with VRAM, the market is moving in the opposite direction. AI hardware remains where memory makers can smell the serious money.
