RTX 5090 Vanishes From Nvidia's Store as AI Eats the GPU

The RTX 5090 is disappearing, and Nvidia isn't saying why

There has been no announcement, no end-of-life notice, no farewell blog post. The RTX 5090 is simply becoming harder to find — including on Nvidia's own store.

The clearest signal comes from Nvidia itself. Search the company's GPU marketplace for the RTX 5090 and you won't find it; the highest-end card listed is now the RTX 5080. The 5090 still appears elsewhere on Nvidia's site, so this isn't a site-wide deletion, but the marketplace is where Nvidia sells and points buyers toward cards. Removing the flagship from that page is about as close to a quiet discontinuation as a company gets without saying the word.

Retailers are following. Micro Center reportedly lists the RTX 5090 as 'not carried', and in the US you can't buy one from a major retailer at anything resembling list price. Third-party sellers are asking close to $10,000. That's not a supply hiccup; that's a card that has left the building.

The rumor filling the gap, first floated by a leaker, is that Nvidia is diverting all GB202 chips — the GPU inside the RTX 5090 — to RTX Pro AI graphics cards. Those boards sell for far more than a gaming card, and demand from AI customers shows no sign of letting up. If you're Nvidia, putting a $2,000 gaming GPU's silicon into a five-figure AI board is not a hard call.

What happens to GeForce prices when the RTX 5080 is the flagship?

The RTX 5080 becomes the top GeForce card by default, and its price is already well above where a xx80 card used to sit.

On Newegg, the RTX 5080 starts around $1,800 in the US. In the UK, Overclockers lists it from about £1,450. Those are the cheapest listings, not the average. For a card that launched as the second tier of the Blackwell generation, that's a lot of money — and it's the price before the 5090's absence starts pushing buyers down onto it.

The Reddit reaction captured the mood bluntly: if you want a 5080, buy one while you can still afford one. That's gallows humor, but it reflects a real dynamic. When the only high-end option left is the 5080, Nvidia and its partners have little reason to discount it. Scarcity at the top tends to lift prices one rung down.

There's a second pressure point: the 5080 is also useful for AI work. The same buyers who made the 5090 impossible to find at list price can absorb 5080 stock too. Gaming cards have become a side market for silicon that AI customers want.

The RTX 5080 Super rumor and the memory problem

Talk of a 24GB RTX 5080 Super suggests Nvidia may consolidate its high-end lineup into a single card — and that the next generation is further away than expected.

The rumor going around is that Nvidia could launch an RTX 5080 Super with 24GB of memory and retire both the RTX 5090 and the standard 16GB RTX 5080. That would leave one high-end GeForce card sitting well above the RTX 5070 Ti. It's a cleaner lineup on paper, and a much easier one to price aggressively upward.

A Super card arriving in early 2027 would also imply the RTX 6000 generation is slipping. A stopgap refresh keeps Blackwell ticking over while Nvidia waits out the memory market. And memory is the elephant here: DRAM prices are predicted to get worse next year, with the crunch potentially lasting into 2028. Launching a new GPU generation into that is not attractive.

None of this is confirmed. Nvidia hasn't announced a 5080 Super, hasn't confirmed the 5090's end, and hasn't explained the marketplace removal. Treat the roadmap as rumor until there's a product page.

Is Nvidia replacing GeForce with cloud gaming?

Not exactly — but the company's software push and GeForce Now give it options that don't require selling you a graphics card.

The counterargument is DLSS. Nvidia is pouring resources into neural rendering and frame generation, and DLSS 5 has already arrived with significant performance gains. That work benefits PC gamers directly, and it would be strange to invest so heavily in a platform you plan to abandon.

But DLSS also benefits GeForce Now. This week Nvidia added DLSS 4.5 override controls to the streaming service, with RTX 5080-tier Ultimate subscribers getting Dynamic Multi Frame Generation up to 4x and Ray Reconstruction improvements. Every DLSS advance makes the cloud version of the pitch stronger, and every strong cloud pitch makes a $2,000 local GPU easier to skip.

That's the uncomfortable symmetry. Nvidia can keep improving gaming performance while making the hardware that delivers it scarcer and more expensive. The software keeps gamers inside Nvidia's ecosystem; the cloud keeps them paying monthly instead of once. Neither requires the company to prioritize a consumer graphics card with thin margins next to an AI accelerator.

Availability and pricing: what this means for US, UK and EU buyers

In the US, the RTX 5090 is effectively unavailable from major retailers, and third-party listings run near $10,000. The RTX 5080 starts around $1,800 on Newegg. In the UK, Overclockers lists the 5080 from roughly £1,450, with prices rising from there. EU availability follows a similar pattern, with the 5090 scarce and marked up.

If you're building a high-end PC now, the practical advice is to buy the card you can find at a price you can stomach rather than waiting for a 5090 restock that may never come. Watch for an official RTX 5080 Super announcement before committing to a 16GB 5080 if 24GB matters to your workload. Prices and stock are moving; check retailer listings before buying.

The RTX 5090's quiet exit is a small event with a large implication: when AI silicon pays better, gaming silicon waits. Nvidia hasn't abandoned GeForce, but it is clearly rationing it. For PC builders, that means higher prices, fewer choices at the top, and a growing case for watching GeForce Now as closely as the GPU shelf.