Nvidia's two best gaming cards are gone from its US store, street prices are two to six times list, and the next GeForce generation may not land until 2028. Here's what's going on.
The RTX 5080 has vanished from Nvidia's US store, weeks after the RTX 5090. The best card Nvidia sells directly is now the 5070 Ti.
US street prices: 5080 over US$2,000 (list US$999), 5090 over US$6,000 (list US$1,999).
Nvidia says the 5090 isn't discontinued, but admits supply is "somewhat constrained" as AI eats the world's memory.
Gaming was just 7% of Nvidia's revenue last year, and it no longer reports gaming as its own line.
Leaks point to no RTX 60 cards until 2028. Nvidia is pushing AI frame generation, cloud gaming and RTX Spark laptops instead.
Want to buy a top-end GeForce card straight from Nvidia? Tough. The RTX 5080 has quietly vanished from Nvidia’s US online store, a couple of weeks after the RTX 5090 did the same disappearing act. The fastest card Nvidia will now sell you on its own marketplace is the RTX 5070 Ti.
IGN, VideoCardz and Wccftech all picked up the 5080’s exit this week, and partner cards from the likes of Asus and MSI are gone from the listings too. Nvidia hasn’t explained it. But when you line it up with everything else Nvidia has done this year, a pretty clear picture forms, and it isn’t a flattering one for PC gamers.
The RTX 5090 dropped off Nvidia’s US marketplace earlier in October. Reports date it to around the 15th. The 5080 followed this week. Before it went, the 5080 was selling for over US$2,000 at US retailers, more than double its US$999 list price. One MSI model was sitting on Amazon at US$2,422.
The 5090 is worse. It launched at US$1,999 in January 2025. Tom’s Hardware’s price tracker now has it above US$6,000, and it reported Amazon listings as high as US$8,500 during a panic over whether the card had been discontinued. Micro Center has marked a stack of partner 5090s as “not carried”, which is retailer speak for “we’re not getting more”.
Nvidia says no. At the RTX Spark launch event in Taipei on 16 October, Nvidia’s GeForce boss Jeff Fisher told Taiwanese media the 5090 is not discontinued and will keep shipping. He admitted supply is “indeed somewhat constrained” and said Nvidia is “going all out to bring products to market”.
Leakers reckon Nvidia has stopped making the 5090’s big GB202 chip for gaming cards, and is putting them into its RTX Pro workstation cards instead. Nvidia hasn’t confirmed that. But the maths is obvious: the same chip goes into the RTX Pro 6000, which reportedly sells for around US$16,000. If you were Nvidia, which card would you rather build?
As for the 5080, the rumour mill says a 24GB version is coming, possibly by the end of the year. Nobody credible has put their name to that yet, so file it under maybe.
It’s the same story as the RAM in your PC: AI is eating all the memory. A 5090 carries 32GB of GDDR7, and every chip that goes into a gaming card is a chip that could have gone into AI hardware worth far more. IDC estimates AI data centres will swallow about 70% of the world’s memory output this year. SK Hynix says it’s sold out for 2026, and Micron’s boss has said supply will stay tight through 2028.
Then there’s the money. In its last financial year, Nvidia made US$193.7 billion from data centres. Gaming brought in US$16 billion, about 7% of the total. That’s not nothing, but it’s a rounding error next to the AI money.
And here’s the tell. Since May, Nvidia doesn’t even report gaming as its own line any more. It’s been folded into a new “Edge Computing” bucket alongside PCs, robots, cars and telecom gear. Its last annual report also warned that supply constraints would hold back gaming “in the first quarter of fiscal 2027 and beyond”. When a company stops telling you how much it makes from something, it usually isn’t because the number looks great.
Not new graphics cards, by the look of it. Nvidia hasn’t launched a new desktop GeForce card all year, and it hasn’t announced the RTX 60 series. Two of the more reliable leakers, Kopite7kimi and Kepler_L2, both say the RTX 60 “Rubin” cards have slipped to 2028. Another, Moore’s Law Is Dead, claims the first half of 2027. The RTX 50 Super refresh is either dead or pushed to 2027, depending on who you ask. None of this is official.
What Nvidia is pushing instead tells you where its head is at:
Nvidia isn’t killing GeForce. But it’s very clearly not the priority any more, and it’s stopped pretending otherwise. Gamers built this company. Now we’re getting the leftovers: whatever chips and memory are left after the AI customers have eaten, at two to six times the sticker price.
The pitch for the next few years seems to be: buy less hardware, let AI make up the frames, or rent a GPU in the cloud. That might suit Nvidia’s spreadsheet. It doesn’t suit anyone who just wants to buy a decent graphics card at a sane price. If you’re in New Zealand or Australia and waiting for prices to come back down, we wouldn’t hold our breath. Grab a 5070 Ti or an AMD card if you need one now, or hang on to what you’ve got.
Are you still holding out for a 5090, or have you given up on Nvidia? Tell us in the GetRekt Discord.
Featured image photo: Geekerwan, CC BY 3.0, via Wikimedia Commons.
The OpenAI boss reckons memory sitting in your gaming PC is a waste of the world's limited DRAM supply. That would be the supply his company helped sq...
Epic's boss reckons Steam's "social walled garden" has held gaming back for years. This from the company that pulled Rocket League and Fall Guys off S...
According to The New York Times, US intelligence believes an executive at one of Nvidia's biggest Asian customers got close to Jensen Huang to feed in...
Comments 0
Quick check before we post that