RTX 5090 discontinued? The reported end of production, what is confirmed (or not) and why Nvidia would drop its flagship

No, Nvidia has not officially announced the end of the RTX 5090. Since October 10, 2026, several leakers have claimed that Nvidia no longer supplies GB202 chips to build new RTX 5090 cards, and that these chips are now reserved for RTX PRO professional cards. The story was picked up by Tom’s Hardware, TechPowerUp, VideoCardz and Digital Foundry, all of which present it as unconfirmed.

Here is what has been verified, what is still a leak, the reasons that make the scenario credible, and what you can buy right now if you were after a very high-end card.

Click to enlarge

On October 10, 2026, the leaker hongxing2020 posted a notice presented as coming from a factory: according to it, Nvidia would no longer make any RTX 5090 model, including the D variants for China, and an RTX 5080 24GB would take over as the most powerful GeForce. The same day, MEGAsizeGPU (@Zed__Wang), another leaker with a solid track record on Nvidia, wrote that there is no more GB202 for the 5090 and that all future supplies of the chip will be exclusive to RTX PRO.

Specialist outlets covered the story with caution: Tom’s Hardware points out that it is an unverified claim and that Nvidia did not respond to its request for comment. TechPowerUp and VideoCardz note that Nvidia has neither confirmed it nor announced an end-of-production date. Digital Foundry describes the RTX 5090 as “effectively” discontinued, based on these leaks.

Hardware Unboxed also released a video on the topic this week, with a blunt title: “The RTX 5090 Was Already Discontinued (For Gamers)”. The point: for gamers, the card was already almost impossible to find at a reasonable price well before these leaks.

Worth noting: this is not the first rumor of its kind. In November 2025, after a rumor about the end of the RTX 5090 Founders Edition, Nvidia told WCCFTech that all GeForce RTX 50 GPUs remained available (PCMag). This time, no denial has been published as far as we know.

The RTX 5090 uses the GB202 chip, the largest of the Blackwell generation. The same chip also powers the RTX PRO 6000 (24,064 CUDA cores, 96GB) and the recent RTX PRO 5500, which keeps the 5090’s 21,760 CUDA cores with 84GB of memory. Every GB202 sold in a pro card earns far more than in a GeForce: that is the explanation put forward by Tom’s Hardware and TechPowerUp.

With 32GB of GDDR7, the 5090 is in high demand for running AI models locally. Companies were reportedly buying them in bulk before they even reached stores, according to TechPowerUp. For Tom’s Hardware, ending the GeForce would also stop these buyers from sidestepping the much more expensive pro cards.

Memory has become the market’s bottleneck. In December 2025, Board Channels was already reporting a 30 to 40% cut in RTX 50 supply in the first half of 2026, tied to tight GDDR6, GDDR7 and DDR5 supply (TechPowerUp). In July 2026, VideoCardz reported that the RTX 50 SUPER cards were ready at at least one partner but held back by the price of 3GB GDDR7 modules, roughly three times as expensive as 2GB modules (VideoCardz; see also our article on the suspended RTX 50 SUPER cards).

Nvidia’s next cards (Rubin, RTX 60) are not expected before 2027, or even 2028, according to Digital Foundry and TechPowerUp. An RTX 5080 24GB could therefore remain the fastest GeForce still in production for a long time, if the rumor is confirmed (see what we know about Rubin).

• RTX 5090 cards won’t disappear overnight: partners, distributors and retailers still have stock, and chips already delivered can still be built into cards.
• Prices are likely to keep rising: the 5090 was already selling well above its launch price. Less new stock is rarely good news for prices.
• The RTX 5090 SE is off the table if the rumor is confirmed: with no GB202 for GeForce cards, a cut-down 5090 looks unlikely (our take on the RTX 5090 SE rumor).
• The RTX 50 SUPER cards are not affected by the end of production: none of the SUPER models mentioned in the leaks uses the GB202 chip. Their launch, however, is still blocked by memory prices.

If you were aiming for an RTX 5090 for gaming, it is probably no longer the right call. For 4K, the RTX 5080 remains the fastest GeForce still in production. For 1440p, an RTX 5070 Ti or an RX 9070 XT gives you much more performance for your money. Only go for the 5090 if you genuinely need its 32GB (local AI, 3D rendering) and accept its current price.

Before buying, check that your case can fit an RTX 5090 and that your power supply can keep up. And if you are still unsure whether to buy now, our guide on whether to buy or wait in the middle of the RAM and GPU crisis has you covered.

No. As of October 14, 2026, no announcement or statement from Nvidia confirms it. It is an end of production reported by leakers and picked up by the specialist press.

Yes, as long as manufacturers and retailers still have stock. But availability is already very low and prices very high.

Not officially. It only appears in the notice relayed by hongxing2020, and could be the RTX 5080 SUPER that has been rumored for months with 24GB of GDDR7.

Because the same chip sells for much more in a professional card for AI and rendering than in a GeForce, while production and memory are limited.

The RTX 5090 has not been officially discontinued, but everything suggests it will no longer really be produced for gamers: the GB202 chip earns more in pro cards, GDDR7 memory is scarce, and the next generation is far off. If you are looking for a very high-end gaming card, the RTX 5080 is the realistic target. We will update this article if Nvidia confirms or denies it.