Oops, sorry, my mistake! I accidentally overwrote my original text here, but I think everyone who was interested had already read it. Anyway, here’s an update:
Jepp Fisher, Senior Vice President of Nvidia said in an Interview, that the RTX 5090 still gets shipped (source) Should/could we believe? I don’t know…if this is true, the question is how many gpu units they deliver for RTX 5090 builts?
Here is some interesting fact I’ve found. On September 14, 2026, NVIDIA reportedly introduced the RTX PRO 5500 Blackwell, which has 21,760 CUDA cores, exactly the same as the RTX 5090 Chip “GB202” so they use the same Chip…hmmm (source)
I bought my first 5090 as soon as i could get hold of one for £2300, then a second one for £2700 and at that time, was pretty annoyed at the price increase. Looking at the same store I bought them from today, they’ve now just passed the £6000 mark each!
There’s a decent profit for me If I were to sell but what then would I buy so will be keeping.
My biggest worry is If one were to fail. There’s no way I’d get a replacement and would probably only get my initial purchase price refunded so would be hugely out of pocket If wanting to buy another.
Crazy times right now.
Lucky Guy, i also ordered a second 5090 for about CHF 2200 half a year ago and they had it “in stock” i already payed…but then they never shipped (they know why^^)
2 days after my last post, price is up by £500 and now £6500 with some others getting close to the £7k mark. I honestly think we’ll see a 5 figure priced 5090 soon if thing keep going the way they are.
That could happen if there’s no faster successor, especially once the 5090 reaches EOL now. There will always be people with money who are willing to pay outrageous prices for a top-end model, simply because it’s still by far the fastest GPU available.
On the other hand, the AI bubble could burst, flooding the market with Cards that becomes a second life. Or these cards could eventually be replaced in data centers with newer faster ones.
The economics are driving the production need for purpose built GPU cards to such a low volume that they will no longer be a thing you can buy at any price. I wonder if there will someday be a market for pulling chips off old obsolete data center cards for use in building “new” PC oriented GPU boards. Anyone out there with the right resources looking to launch a new business model?
This honestly reminds me of the RTX 1080 Ti and Titan era. Cards like the 5090 were perfect for communities like this one — people who do video restoration, AI generation, and production work as a hobby, but also game on the same machine.
Losing the true “mega‑die” tier hurts hybrid users the most. Those cards weren’t just for gaming or just for compute — they were the only GPUs that handled both workloads at once without compromise.
I bought a used RTX 4090 as well. It should arrive on Monday, the Asus TUF RTX 4090 (24GB)
The TUF has solid „die-cast metal shroud“, which helps improve cooling performance. For me it’s one of the best RTX 4090 out there. I’m not a millionaire, though, so I’m going to sell the 5090 and use the 4090 as its replacement.
(source)
