A rumour that Nvidia will cancel the RTX 50 Super cards because of RAM is almost certainly nonsense but everything won't be fine, either

If there's one thing that is a dead cert in the PC tech world, it's that in the build-up to an expected product launch, the interwebs will be full of leaks and rumours. In the case of the latter, these can be nothing more than a statement of the blindingly obvious, but occasionally, something gets dropped that's just wild. The recent claim that Nvidia will cancel its forthcoming GeForce RTX 50 Super series due to memory shortages is a prime example.

As reported by Uniko's Hardware on X, someone, somewhere, is claiming that the DDR4/DDR5 memory shortages will make the cost of 3 GB GDDR7 modules so high that Nvidia is going to cancel the refresh of RTX 50 graphics cards, which are expected to use the larger RAM chips. Oh, and the current cards will get a lot more expensive for good measure, too.

There's little point in asking Nvidia directly about this because it will always just say things like 'We don't comment on rumours' or 'We don't talk about future products', but if the GPU giant did respond, and although I obviously can't say for certain, I would expect them to deny the claim outright. That's because chips being in short supply has never stopped them from launching before.

Just recall what happened with the GeForce RTX 5090, 5080, 5070 Ti, 5070, and so on. The availability for every model was awful at launch, and it took many months for the situation to improve. Prices were also well over MSRP, too. And yet, PC gamers and hardware enthusiasts still bought them in droves, helping Nvidia to earn a record revenue from it all.

You might be wondering why GDDR7 would be in short supply anyway, as massive AI servers don't use it. Even though the ultra-fast graphics memory used on most RTX 50 cards is internally different to DDR5 and LPDDR5x, it's still fabricated in the same process lines.

So with the demand for the normal system RAM ploughing through the stratosphere, because of bloody AI, manufacturers such as Samsung and Micron will want to use as much production capacity as possible for DDR5, as well as AI-specific stuff like HBM.

What's likely to make matters worse is that, as things currently stand, only Samsung lists 3 GB GDDR7 as being in mass production. Neither Micron nor SK hynix even mention it, and just produce the standard 2 GB stuff that's currently used on RTX 50 cards. Given GDDR7 is only used by Nvidia at the moment, this is all perfectly understandable.

Graphics card partners don't source RAM modules directly: they purchase them as part of a package from Nvidia, which includes the GPU. It's unlikely Jensen will be telling his bean counters to absorb the rising cost of memory chips, so that means Asus, Gigabyte, MSI, and others will need to pay more for the packages, and it's obvious what that will mean for us.

So I am 99.99999% sure that Nvidia will launch RTX 50 Super graphics cards early-ish in 2026 (the 0.00001% is to account for the fact we live in bizarre times, so anything's possible). They might be delayed a touch, or Nvidia might space out the Super variations over a period of months, but they'll still turn up.

However, I'm 100% sure they're going to be more expensive than originally intended, perhaps no more than $30 to $50 extra for the RTX 5070 Super, but probably a lot more for the RTX 5070 Ti Super and 5080 Super, as they use 33% more RAM modules. I guess that's just another thing that's a dead cert in the PC world these days: Team Green GPUs priced to the heavens.

Best graphics card 2025

1. Best overall: AMD Radeon RX 9070

2. Best value: AMD Radeon RX 9060 XT 16 GB

3. Best budget: Intel Arc B570

4. Best mid-range: Nvidia GeForce RTX 5070 Ti

5. Best high-end: Nvidia GeForce RTX 5090

👉Check out our full graphics card guide👈