If you’ve priced a high-end graphics card lately, you already know something is off. The card that was supposed to anchor a nice gaming build now costs more than an entire decent PC did a couple of years ago. The RTX 5090 launched at $1,999. By mid-2026 it was a very different animal, and the reason has almost nothing to do with gamers.
The numbers are genuinely absurd
By July, the RTX 5090 was listing around $4,329 on Amazon, with premium partner cards running above $5,000. It’s gotten worse since. As of early August, South Korean retailers were pricing the card as high as roughly $5,112, up about $1,050 in a single month. This isn’t scalping or a launch-week blip. It’s a structural shift, and it stretches well beyond one flagship. Mid-range cards in the $300 to $800 range are projected to climb another 10 to 25% in 2026, with 12GB and 16GB models most exposed, precisely because those are the memory configurations AI buyers want too.
The real culprit is memory
Here’s the part most price complaints miss. The bottleneck isn’t the GPU chip, it’s the memory soldered around it. Industry trackers say VRAM now accounts for more than 80% of the total bill of materials for some high-end cards, with the cost of 16GB of GDDR7 climbing from roughly $65 to $80 in mid-2025 to more than $200 by year-end. When the most expensive ingredient triples, the finished product follows.
And that memory is being vacuumed up by AI. IDC projects that AI data centers will absorb roughly 70% of the world’s memory output in 2026, up from 20 to 30% in 2022, while DRAM supply grows just 16% year over year. Consumer devices are left fighting over the scraps. When demand accelerates and supply crawls, price becomes the only release valve.
Why this shortage is different, and who else it hits
The instinct is to compare this to the 2021 crypto-mining crunch, but that comparison undersells it. This time the issue isn’t logistics or availability so much as a deliberate reallocation, with AI data centers simply outbidding consumer hardware for the same DRAM and graphics memory. Nvidia even shelved its expected RTX 50 Super refresh, because a “Super” card typically adds VRAM, the exact component that’s scarce and expensive right now.
It’s not just GPUs, either. Nintendo raised the Switch 2, Sony bumped the PS5 a second time, and Valve hiked the Steam Deck by nearly 50%, all citing the same memory shortage, since consoles and handhelds pull from the same supply. So when does relief arrive? Honestly, not soon. As long as AI infrastructure spending keeps running this hot, the memory that would go into affordable consumer hardware has a higher bidder. The uncomfortable takeaway for gamers: you’re not competing with other gamers anymore. You’re competing with data centers, and they have deeper pockets.
This website uses cookies.