Okay, so the NVIDIA RTX 3060 12GB is back on retail shelves, like, right now, as of July 5, 2026. This is a five-year-old card, right, it launched in 2021. And it’s showing up at prices around $329 in the US and €333 in Europe. That’s basically its original launch price. The big problem is the DRAM crisis, it’s a huge deal.

Global DRAM and GDDR supply is getting eaten up by AI infrastructure, like, at an insane pace. Memory manufacturers just can’t keep up with new fab capacity. Micron, SK Hynix, Samsung, they’re all locking in record-high memory prices with major customers for years. This isn’t a short-term thing.

Analysts don’t see any real relief until late 2027 or even 2028. So, what’s happening is that AI data centers need massive amounts of memory, and they can pay way more per gigabyte than a graphics card maker can. This means the memory goes to the AI guys, and GPU makers get whatever is left over. Video memory now accounts for a huge part of the cost to build a graphics card. On some cards, the memory itself costs more than the GPU chip did a year ago. NVIDIA actually cut its GeForce RTX 50 series production by like 30% to 40% in the first half of 2026.

They are redirecting memory supply to higher-margin products, you know, the AI stuff. This isn’t a rumor about scarcity, it’s a company choosing to build fewer gaming cards. The RTX 3060 12GB is an Ampere card, it’s older architecture. It lacks features like DLSS 4.5 and the newer Tensor cores that the Blackwell-powered RTX 50 series has. The newer RTX 5060 8GB, for example, is available for $349-$359, so it’s only a little more expensive than the old 3060.

The 5060 actually performs better in most games, unless you’re playing at 4K, which isn’t really the target for these cards. The 3060’s only real saving grace would be a price under $300, but it’s not there.AMD is also feeling the pinch. They told their board partners that GPU kit prices, which include the GPU die and GDDR6 memory, are going up by 10% in July. This is because GDDR memory is in short supply. The PC gaming market itself is still growing, though.

It’s projected to hit $96.69 billion in 2026, up from $86.12 billion in 2025. That’s an 8.52% CAGR through 2034. But the GPU side is just… it’s a mess. NVIDIA’s gaming revenue for Q2 fiscal 2026 was $4.3 billion, up 49% year-over-year, but that’s still a small piece of their overall revenue.

Their data center segment, that’s where the real money is, it was $41.1 billion in Q2 fiscal 2026, up 56% year-over-year. SK Hynix is planning an eight-fold increase in their 1c DRAM production for 2026. They are going from about 20,000 wafers per month to 160,000 to 190,000 wafers per month by the end of 2026. But even that huge increase won’t fix the current DRAM shortage, because the new production is already allocated, mostly for HBM for AI. Samsung was the only top-three memory maker to increase its DRAM market share in Q1 2026, going from 36.5% to 38.6%.

SK Hynix and Micron saw their shares drop. Total DRAM sales for Q1 2026 were $97.1 billion, up 85.8% from Q4 2025. I mean, the whole situation is just wild. It’s like, how long can this go on? I bought MU, Micron, back on January 15, 2024, at $88.50 a share.

I’m holding it until it hits $1500 or until the AI bubble bursts, whichever comes first. (I think it’s going to be the former, but who knows, right?) Micron’s 52-week high was $1,255.00, and their low was $103.38. It’s been a good ride so far. NVIDIA’s desktop GPU market share is still super dominant, like 90% in Q1 2026. AMD’s share actually decreased a tiny bit, and Intel gained a little. The RTX 50 Super refresh, which was supposed to fix VRAM capacity issues, slipped from a 2026 launch to a rumored CES 2027 debut.

NVIDIA’s Computex 2026 messaging was all about AI hardware, not new gaming silicon. Board partners are raising prices on existing RTX 50 cards by 10-15% because of rising GDDR6/GDDR7 costs. This is just the new normal, I guess.