The GeForce RTX 5090 is about to be discontinued, at least according to a fresh rumor that hit the airwaves over the weekend. According to leaker Megaize GPU, who has been correct with some leaks in the past, Nvidia will no longer supply their GB202 GPU die for use in RTX 5090 graphics cards. Instead, that die will be allocated exclusively to the RTX Pro series for use in workstation cards such as the RTX Pro 6000. This leak sent shockwaves through the PC gaming scene, and I immediately saw a bunch of despair from gamers who are already unhappy with the current state of the hardware market. A lot of talk of this being the beginning of the end. Nvidia are about to shove gamers to the side and only service the AI market.
However, if we're being honest, the RTX 5090 has already been discontinued for gamers. For several months now, perhaps even all of 2026, it simply has not been a gaming GPU. And practically speaking, it hasn't been available for gamers to buy. So, if this leak turns out to be accurate, and we don't have information to confirm either way, it actually won't be a very significant change in some ways. There are some serious concerns here, which I'll get to, and I'll chat a bit about the rumored replacement for the RTX 5090, but this GPU has already been discontinued for gamers. Let me explain.
Today's video is brought to you by Sapphire and their premium Nitro Plus range of AMD Radeon 9000 series graphics cards and X870 motherboards. For enthusiasts seeking a no-compromise showpiece system, Sapphire's Nitro Plus Phantom Link graphics cards allow for cable-free power delivery when paired with Sapphire's Phantom Link X870 EA Nitro Plus motherboard. This is one of the cleanest high-end AM5 builds you can put together while also supporting PCIe 5.0, Wi-Fi 7, and a stacked I/O. The Nitro Plus Radeon RX 9070 XT Phantom Link GPU supports a large triple-fan cooler and plenty of raw compute for 4K gaming while providing 16 GB of GDDR6 memory and all the latest RDNA4 features. Sapphire's Phantom Link provides a serious step up in aesthetics without sacrificing performance. So, for more information, please check the links in the video description.
The GeForce RTX 5090 launched on January 30, 2025 at a supposed launch MSRP of $1,999 US. This was the second most expensive GeForce GPU launch price of all time, behind the dual-GPU GeForce GTX Titan Z from 2014 that cost $3,000 on day one. The RTX 5090 was a very expensive GPU from the beginning, but when you look at the MSRP and compare it to the prices of previous models like the RTX 4090 and RTX 3090, and then also look at general buying behavior for all these cards, it was a gaming GPU. High-end gamers with loads of cash to spend did indeed buy the 5090 for gaming, as they did the 4090 and 3090 for prices in excess of $1,500 US.
While the RTX 5090 did launch at $2,000 US, it was a very difficult card to obtain at the time. Back in March of 2025, so two months after launch, you couldn't really buy them at all. And when we asked retailers, we discovered that Nvidia didn't supply very many of these models. We concluded the 5090 had a fake MSRP at the time. But when supply did hit store shelves, typically for around $2,500 US, it was still in the realms of high-end gaming GPU pricing.
Eventually, the card came back into normal stock, and we regularly checked in with the market to see how things were going. In June 2025, the RTX 5090 cost around $2,800 in the US, and pricing slowly started to fall in the second half of the year, reaching $2,350 for the cheapest models in September based on our tracking data. Even in November of 2025, as memory prices started to rise, the RTX 5090 was on store shelves for around $2,500 US, best case scenario. So, across 2025, the RTX 5090 wasn't really a $2,000 US GPU, but it was available. Best case, it was only about 20% above where it should have been priced. Gamers were genuinely buying this model for the ultimate gaming experience, and the price was within reason for high-end shoppers. Expensive, but not ludicrous.
Then comes 2026. In the space of just three months from November to February, the RTX 5090 jumped in price by $1,000 to a whopping $3,500 US. Or should I say a modest $3,500 US considering the price increases to come. In April, the 5090 went up to $3,700, then $4,000 in July, and in August, we got another shock, a $900 price increase after a month. And then in our last update in September, probably the worst 30-day price increase for a GPU we have ever seen over many years of tracking GPU pricing. The RTX 5090 became $7,000 minimum, up over $2,000 from the prior month.
Today, the 5090 is even less available and will typically cost you over $7,000 from third-party resellers at most major retailers, if not closer to the $9,000 US mark. At $2,500 US, the RTX 5090 was an expensive high-end gaming GPU. At $3,700 or more for most of 2026, though, was the RTX 5090 really a graphics card that gamers were buying? That alone is more than the vast majority of gamers spend on their entire PC. At $5,000, the August price, we're in completely absurd territory. And then at the current $7,000 US, we're so far beyond what gamers spend on their entire PC that it's pointless to describe the RTX 5090 as a gaming product.
Realistically, for most of this year, the RTX 5090 has not been a gaming GPU. It has not been priced like a gaming GPU. People have not been purchasing this model for gaming, and no one in the industry has tried to make this card attainable for gamers. Not Nvidia, not the memory makers, not AIBs, no one. It simply has not been an option for 99.99% of PC gamers. This is why in effect the RTX 5090 was discontinued for gamers months ago. The fact that this may become official if the leaks are true doesn't change anything really.
Before the card was discontinued, gamers could not realistically purchase the RTX 5090 for gaming. And after it's discontinued, you'll still not be able to purchase it for gaming. In fact, we don't even need to figure out whether the leak is true or not. The market already tells us this GPU exited the gaming segment long ago. It's already an RTX Pro GPU in all but name.
I understand the panic and despair about this leak, though. Not everyone regularly checks in on the price of an RTX 5090. It's out of reach for most gamers at the $2,000 MSRP, let alone at current prices. So, why would you bother looking at what it costs each month? It's totally fair to not be aware that for the last few months you'd have to spend over $5,000 to buy one, making it effectively already discontinued for gamers. If you were going about your business happily, still thinking that the RTX 5090 was available at a high but still reasonable price, then you suddenly find out that this model is being discontinued, yeah, it's pretty reasonable that you would panic. But no, that is not actually what is happening. This card hasn't been available at a reasonable price for a long time.
So, what do I mean when I say it's already an RTX Pro GPU in all but name? Well, simply put, it's comparatively priced right now to other RTX Pro GPUs and has been for months. The RTX Pro Blackwell series includes five GPUs using both the GB203 die found in the RTX 5080 and the GB202 die found in the RTX 5090. All of these GPUs are much more expensive than GeForce cards despite using similar GPU cores.
The cheapest RTX Pro model is the RTX Pro 4000, which has a current retail price of $3,300 US, which is very cheap for a workstation card. This model has 8,960 shader cores, so it's effectively an RTX 5070 Ti in its GPU core, but it has 24 GB of GDDR7 instead of 16 GB. And it also has other professional features. The extra memory and pro support is how Nvidia justifies selling an RTX 5070 Ti die, which go for around $1,200 in a graphics card these days, for $3,300 US instead.
Then we've got the RTX Pro 4500, also based on GB203, but with 10,496 shader cores, slightly cut down from the 10,752 we see in the RTX 5080. This model has 32 GB of VRAM and retails for $4,700, but it has much lower compute capabilities than the RTX 5090.
Then there's three models that use GB202 from the RTX 5090. The RTX Pro 5000 comes in a 48 GB version for $8,000 US and a 72 GB version for $9,600 US. With 14,080 shader cores, you get roughly 65% of an RTX 5090's GPU core, but more memory. Then there's the flagship RTX Pro 6000, which costs $16,000, gives you more shader cores than the 5090, and also 96 GB of GDDR7.
AI workloads are highly dependent on memory capacity and bandwidth. So, the upper-tier models are desirable for those that want to run large models locally. But even 32 GB is enough for a lot of workloads. And this explains why the RTX 5090 quickly rose to RTX Pro-level pricing, along with supply issues, of course. If the RTX 5090 was still available for 2025 pricing and you needed 32 GB of VRAM, would you rather spend $4,700 to get that capability in the RTX Pro 4500 or $2,500 to get that in an RTX 5090, which is faster as well? Of course, you'd just get the 5090, so the price skyrockets to match the professional range.
Today, at $7,000 US, it makes perfect sense among those other Pro cards. It's much faster than the RTX Pro 4500 for the same amount of memory, but it doesn't have as much memory as the Pro 5000. So, it's a bit cheaper than the Pro 5000 and more expensive than the 4500. And if you want nearly the full GB202 die like the 5090, but 3x the memory, that is also available for around 2.5 times the price.
Nvidia doesn't need to discontinue the RTX 5090 for this model to already be effectively discontinued for gamers and for it to already be effectively an RTX Pro GPU. But there's a clear reason why Nvidia would want to actually and totally discontinue the model. $7,000 is cheap for GB202 silicon. They already sell binned, cut-down, defective dies for at least $8,000 with 48 GB of memory and as much as $16,000 with most of the die enabled. So instead of offering that GPU core for $7,000 US in the 5090, offer it for $16,000 in the RTX Pro 6000 and get rid of the 5090 entirely. Yeah, all these numbers sound absurd if you're used to gaming GPU pricing, but it's the reality of the workstation market at the moment.
Actually discontinuing the 5090 would also give Nvidia much greater control over pricing and would give Nvidia a better share of the retail price. Instead of selling the GPU die to OEMs at a lower price who can then go about scalping 5090s because AI pros are buying them for pro-level pricing, Nvidia can charge the maximum amount for use in actual RTX Pro 6000 cards. And Nvidia themselves make these cards, all to a reference spec. There's no AIB designs at all like you see for GeForce.
If Nvidia does actually discontinue the RTX 5090, not much is going to change in terms of the gaming range for gamers. All the GeForce GPUs in the price range gamers would actually consider are going to continue to be available, and the RTX 5080 would be the new flagship model. It's already the de facto gaming flagship given the 5090's price, but that would be made official with the move. What we're seeing now would be the effective situation for longer.
If anything, it will screw over AI professionals who wanted the GPU performance of the 5090, but only required 32 GB of VRAM for their workload. With no RTX 5090 on the market, Nvidia would bleed those customers dry. You'd need the $16,000 RTX Pro 6000 to match the compute capabilities of the 5090, far higher than what buyers are paying today, tiny violins and all that.
But there is a serious issue here for gamers. With no RTX 5090 on the market, AIBs would be unable to replace any broken RTX 5090s with brand new RTX 5090s. This will almost certainly lead to customers getting screwed over in warranty claims, and these customers could be genuine gamers who bought the card for gaming when prices were more reasonable, such as in the middle of last year.
If an RTX 5090 dies, the only recourse for AIBs would be refunding the purchase price, which could leave that buyer massively screwed over. If the buyer only spent $2,500 on that 5090, and they get $2,500 back as a refund, they wouldn't even get close to replacing that card with something equivalent in performance. They'd be forced into a downgrade to an RTX 5080 because no one is going to stump up an extra $13,500 to buy an RTX Pro 6000 for gaming. And an RTX 5080 is much slower than an RTX 5090, slower than an RTX 4090 even, so it's not even a comparable replacement.
Alternatively, when refunded, that buyer could try their luck on the used market, but good luck getting an RTX 5090 for anywhere near $2,500 US used. Even RTX 4090s go for around $3,000 used at the moment, and I'd expect that to rise if the RTX 5090 exits the market.
The right thing for Nvidia and AIBs to do would be to still allocate a small amount of GB202 supply to RTX 5090s to facilitate proper warranty replacements. We're just discussing a leak here. They of course could do that. We don't have the details, but considering the track record of some warranty claims we have seen, I wouldn't hold my breath for this situation eventuating.
Just recently, for example, PNY allegedly denied a gamer's warranty claim on their RTX 5090 that had its connector melt. PNY denied the warranty, according to the report, because the user connected the GPU using a third-party power cable, in this case a 12V-2x6 cable provided by the power supply. PNY allegedly claimed that only PNY-approved cables or adapters are covered under warranty, so damage due to the use of other cables is not on PNY. This user didn't even receive a refund, which wouldn't have covered a replacement, but would have been better than nothing, I guess.
And that case happened before the RTX 5090 has been maybe officially discontinued. Companies are already looking at ways to screw over genuine gamers with 5090 warranty claims, and I would be very surprised if anything changed if RTX 5090s get the axe. It's going to be some nervous times for existing 5090 owners, praying their 5090 doesn't die or have a melted connector, which is a real issue that can affect anyone.
That's the big issue with this supposed plan to discontinue the 5090. It's arguably Nvidia's least reliable GeForce GPU, the product that is most likely to fail. And without proper replacements available, anyone that needs to RMA their card, they're going to get screwed.
According to a separate rumor, this time from Honging 2020, Nvidia have a replacement for the RTX 5090 lined up. By replacement, I mean a replacement flagship model, not in any way an actual replacement for the performance of the RTX 5090. No, apparently the new flagship is supposed to be an RTX 5080 24 GB model, with this leaker claiming the RTX 5080 16 GB will be discontinued.
It's not clear whether the new 24 GB model will be an RTX 5080 Super or just an RTX 5080 card with more VRAM under the same name. Probably should be a Super given our usual issues with naming there, but yeah, we don't know. It's not even clear whether this rumor is true, to be fair.
A 24 GB version of the 5080 has been rumored though for some time, and at one point it was expected to hit the market as the RTX 5080 Super. That GPU was largely set to differentiate itself from the RTX 5080 through 8 GB of extra VRAM plus minor changes to the GPU's performance. The RTX 5080 already uses the full GB203 die, so there's no room to increase shader core count while using the same GPU silicon.
The way to make an RTX 5080 with 24 GB of VRAM is to use 3 GB GDDR7 modules instead of the current 2 GB modules. Using eight of these chips takes you to 24 gig instead of 16 gig. And Nvidia already uses the 3 gig modules for some laptop GPU and RTX Pro configurations. It's probably what the RTX 5080 should have used from the start.
The problem with creating an RTX 5080 24 GB in the current market is that increasing the VRAM amount will make the card substantially more expensive. Adding an extra 8 GB of GDDR7 will add several hundred to the final price of the GPU, accounting for the cost of the chips, Nvidia's margin, and the fact that 3 GB GDDR7 modules are more expensive than 2 GB modules per gigabyte. RTX 5080s currently cost at least $1,600 US, so I'd expect a 24 GB version to debut around $2,000 US in the current market.
The bad news there is that if the RTX 5080 16 GB is discontinued like the rumor suggests, that category of product will simply get more expensive. The new flagship GPU will be more than $1,600 US, and there will be a larger gap between it and the now $1,200 RTX 5070 Ti. Getting more VRAM in the flagship model is more palatable, but normally this would come at a fairly minimal cost, or at least a much lower cost than we see today.
The other issue is the potential for an RTX 5080 24 GB to be more usable for AI workloads, driving up the price like we see with the RTX 5090. The RTX Pro 4000 costs $3,300 US with 24 GB of VRAM, and used RTX 4090s go for around $3,000 US. You'd hate to see the RTX 5080 become a $3,000 product due to this supposed shift in VRAM capacity. But in the current market, anything is possible, and if there's a way to screw over gamers somehow, you just know that's what will happen.
Like I've said multiple times throughout this video though, this is just a leak/rumor at this stage. We don't actually know for sure whether the RTX 5090 will be discontinued or what is happening with the RTX 5080. We do know the RTX 5090 has been effectively discontinued for gamers, which sucks, and perhaps suggests it's only a matter of time before that becomes official.
We reached out to Nvidia for comment on the RTX 5090 being discontinued, but as of when we filmed this video, they are yet to respond. Probably not that unusual given the news broke over the weekend.
Anyway, that's it for this one. Just thought I'd come on here and have a chat about the RTX 5090 supposedly being discontinued and where it currently is positioned in the market. We've of course been tracking GPU pricing for a long time now, not just with the current series. We've been doing this for so long, way back into the crypto era, way back even before that. So yeah, things just, yep, they are what they are and they just continue to be what they are.
So anyway, that's nice. If you do want to support Hardware Unboxed and independent testing, as usual, we have Patreon links to those in the description below. If you sign up, you get access to all the usual benefits like our Discord community, monthly live streams, and more. So, thanks for watching, and I'll catch you in the next one.