Nvidia spent the back half of October 2026 doing something it rarely has to do: publicly denying that one of its own graphics cards is dead. The GeForce RTX 5090, the company’s $1,999 flagship, vanished from Nvidia’s own online marketplace by October 15, disappeared from Best Buy’s shelves, and got reclassified as “Not Carried” at Micro Center, according to Tom’s Hardware. Resale listings on Amazon climbed as high as $8,500, more than four times the card’s launch price, while a Brazilian distributor separately reported that two pallets of Zotac RTX 5090 cards worth roughly $470,000 went missing at a São Paulo airport.
The trigger for the panic was a leaked factory notice, amplified by hardware leaker accounts, claiming Nvidia had stopped allocating its GB202 silicon to consumer GeForce cards altogether and was routing every die to RTX Pro workstation and data center products instead. Nvidia pushed back within days, but the episode has become a case study in how the AI boom is quietly rewriting who gets to buy the best gaming GPU on the market.
What actually happened to the RTX 5090 in October 2026
The story started with a factory-floor leak. On October 10, 2026, Tom’s Hardware reported on a circulating notice, attributed to hardware leaker accounts including one going by @hongxing2020 and reinforced by a second account, @Zed__Wang, claiming Nvidia had halted manufacturing of GeForce RTX 5090 cards, including the China-specific RTX 5090D variant. The claim went further than a simple shortage: it alleged that Nvidia’s GB202 die, the largest chip in the Blackwell gaming lineup, would no longer be fabricated for consumer add-in boards at all. A separate leaker account, MEGAsizeGPU, put it bluntly: “no more GB202 for 5090, confirmed,” followed by a second post stating “All future GB202 supplies are RTX Pro exclusive,” as cited by Ground News.
Tom’s Hardware was careful to flag the notice as unverified and not an official Nvidia statement, but the retail evidence lined up with the rumor almost immediately. The RTX 5090 disappeared from Nvidia’s own marketplace by October 15, even though the storefront had kept a “starting at $1,999” price tag visible on a card nobody could actually buy at that price. Micro Center moved the RTX 5090 from “out of stock” to the more permanent-sounding “Not Carried” label. By October 17, Best Buy effectively had zero units available anywhere in its network.
Pricing on the secondary market told the rest of the story. Third-party Amazon listings were asking around $8,500 for a card that launched at $1,999, a roughly 4.25x markup. Newegg listings sat closer to $7,300, with shipping details suggesting a chunk of that inventory was coming from overseas resellers rather than normal US retail channels. None of this, on its own, proved Nvidia had killed the card. But it was enough to turn a leaker rumor into a mainstream tech story within a week.
Nvidia’s on-record response: “not been discontinued”
Nvidia’s answer came from Jeff Fisher, the company’s senior vice president overseeing the GeForce and PC business, who addressed the reports in a Taipei interview around October 17, 2026. “Although the RTX 5090 has other variants, it has definitely not been discontinued,” Fisher said, according to Dexerto. He added that “Nvidia will continue shipping, and this is very important,” while also acknowledging that “the supply chain is indeed a bit constrained, but Nvidia is going all out to get products to market.”
Those two statements, read together, are the whole story in miniature. Nvidia is not confirming a formal end-of-life for the RTX 5090, and it is not confirming the leaker claim that GB202 is now off-limits to GeForce entirely. But it is confirming that the chip is scarce, that the scarcity is a supply-chain capacity problem rather than a retail-channel hiccup, and that the company is not promising the shortage ends soon. Hardware Busters summed up the gap between message and market reality directly, noting that Nvidia said production and shipments would continue even as the card had vanished from its own store and was selling for roughly four times its list price on the open market.
That is a meaningfully different story than the RTX 5090 simply being hard to find because of strong demand. A GPU that is merely popular gets restocked in waves and sells through fast, with prices settling back toward MSRP within a quarter or two. A GPU whose core silicon is being systematically redirected toward a more profitable product line does not come back when demand cools, because the scarcity was never about demand in the first place. Nvidia’s own statements leave both readings technically open, which is exactly why the denial didn’t end the controversy.
Why GB202 is suddenly the most valuable die at Nvidia
GB202 is the largest chip built on Nvidia’s Blackwell architecture, and it is not exclusive to gaming. The same die, in slightly different configurations and binning tiers, underpins the RTX 5090 gaming flagship and the RTX Pro professional workstation lineup that feeds engineering firms, AI developers, and render farms. When Nvidia has a fixed number of GB202 wafers coming out of TSMC each quarter, every die that goes into an RTX Pro workstation card is a die that does not go into a GeForce box.
The economics of that tradeoff are not close. An RTX Pro workstation card built on the same silicon commands a price many multiples higher than a GeForce RTX 5090’s $1,999 MSRP, and demand for AI-capable workstation and data center silicon has been running far ahead of supply throughout 2026. Nvidia does not have to formally cancel the RTX 5090 to deprioritize it: it only has to keep allocating marginal wafer capacity toward the higher-margin professional and data-center line every time a scheduling decision comes up. That is a quieter, harder-to-fact-check version of discontinuation, and it is consistent with everything Nvidia has said on the record. Fisher never denied that GB202 allocation had shifted; he denied that the RTX 5090 itself had been discontinued. Those are not the same claim.
This is also where the leaked factory notice and the retail evidence stop contradicting each other and start reinforcing each other. A card that disappears from a manufacturer’s own storefront while the company insists shipments are continuing looks exactly like what you’d expect if GB202 wafers were being quietly re-sorted toward RTX Pro bins before GeForce ever got first pick.
The $470,000 Brazil disappearance
While the production-halt rumor was playing out in English-language tech press, a separate and unrelated supply-chain story broke in Brazil. Distributor Oderço, also rendered Orderco in some reporting, said two pallets of Zotac-branded RTX 5090 cards arrived at Guarulhos International Airport in São Paulo and then disappeared before they could be collected, according to Mezha.net. The distributor valued the missing shipment at approximately $470,000, or about R$2.5 million.
The timing is what made the story land. The disappearance surfaced amid an international shortage in which the RTX 5090 had already vanished from online retailers by mid-September 2026, weeks before the GB202 production-halt rumor even started circulating. A six-figure pallet of flagship GPUs going missing at an international airport is newsworthy on its own, but in the context of a card already selling for 4x MSRP and disappearing from retailer shelves, it read to a lot of PC builders as one more data point confirming that something structural, not just seasonal, was happening to RTX 5090 supply.
The rumored replacement: an RTX 5080 with 24GB
Alongside the GB202 reallocation claim, the same wave of reporting floated a specific successor for gamers left without a flagship: a revised RTX 5080 carrying 24GB of VRAM, up from the 16GB on the card currently shipping. Outlets covering the leak framed it as the GPU gamers would get instead of a steady RTX 5090 supply, built on a smaller, cheaper-to-allocate die than GB202. Nvidia has not confirmed specifications, pricing, or a launch window for any revised RTX 5080, and this detail remains at the rumor stage rather than a verified product. It is worth tracking precisely because it would be the cleanest signal yet that Nvidia is comfortable repositioning its gaming lineup around whatever silicon AI and workstation demand leaves on the table, rather than the other way around.
How this fits the rest of 2026’s GPU market
The RTX 5090 situation did not happen in isolation. Gaming GPU prices across the board have been climbing for months as AI infrastructure spending pulls fabrication, packaging, and memory capacity away from consumer parts. The same high-bandwidth memory and advanced packaging capacity that gaming flagships compete for is the capacity data centers need for AI accelerators, and in 2026 the data-center side of that fight has far deeper pockets. That dynamic has already shown up elsewhere in Nvidia’s lineup, and in competing products from AMD and Intel, where consumer GPU pricing has been drifting upward even on mid-range cards that have nothing to do with AI workloads directly.
What makes the RTX 5090 case distinct is the specificity of the claim. This is not a generic “prices are up because of AI demand” story; it is a claim that a single die, GB202, has a documented dual role in two product lines with wildly different margins, and that Nvidia is allegedly resolving the capacity conflict by favoring the higher-margin line. If true even partially, it is a preview of how every future Nvidia gaming flagship built on shared silicon with a professional or data-center sibling could get squeezed the same way.
RTX 5090 pricing and availability timeline
| Date (2026) | Event | Detail |
|---|---|---|
| Mid-September | Retail vanishing act begins | RTX 5090 reported gone from most online retailers |
| October 10 | Factory notice leak | Leaker claims Nvidia halted RTX 5090/5090D manufacturing, GB202 redirected to RTX Pro |
| October 13 | Regional price spike | AMD RX 9070 XT hits €819.90 in German price tracking, up €52 from September 23 |
| October 15 | Nvidia store listing disappears | RTX 5090 drops off Nvidia’s own marketplace despite showing “starting at $1,999” |
| October 16-17 | Retailer reclassification | Micro Center flags RTX 5090 “Not Carried”; Best Buy shows ~zero units |
| October 17 | Nvidia responds | SVP Jeff Fisher denies discontinuation, confirms constrained supply chain |
| October 18 | Brazil shipment disappears | Two Zotac RTX 5090 pallets worth ~$470,000 go missing at Guarulhos airport |
RTX 5090 price distortion by retail channel
| Channel | Listed price (approx.) | Multiple of $1,999 MSRP | Availability status |
|---|---|---|---|
| Nvidia Marketplace (official) | $1,999 (listed, unavailable) | 1x | Delisted as of Oct. 15 |
| Best Buy | N/A | N/A | Effectively zero stock, Oct. 17 |
| Micro Center | N/A | N/A | Reclassified “Not Carried” |
| Newegg (third-party) | ~$7,300 | ~3.65x | Limited, overseas-linked sellers |
| Amazon (third-party) | ~$8,500 | ~4.25x | Sporadic, high markup |
Historical context: this isn’t Nvidia’s first gaming-GPU squeeze
Nvidia has lived through supply crunches before, most visibly during the 2021 cryptocurrency mining boom, when RTX 30-series cards sold for double and triple MSRP because miners and gamers were bidding for the same limited wafer allocation. That crunch ended the way most demand-side shortages end: mining profitability collapsed, secondary-market prices fell back toward MSRP within roughly a year, and supply normalized once the competing buyer lost its reason to compete.
The 2026 RTX 5090 situation looks superficially similar but has a structurally different driver. Crypto miners were buying the same finished GeForce cards gamers wanted; AI and workstation buyers are not competing for finished RTX 5090 boards at all, they are competing for the GB202 die itself, upstream of the point where a card even gets branded GeForce or RTX Pro. That is a harder shortage to wait out, because it does not resolve when a speculative bubble pops. It resolves only if Nvidia adds enough fabrication and packaging capacity to satisfy both product lines simultaneously, or if the company makes a deliberate allocation decision to protect GeForce volume, something it has given no public indication of doing as of this reporting.
What this means for PC builders and gamers right now
For anyone currently planning a high-end gaming PC build, the practical takeaway is straightforward: budgeting around RTX 5090 MSRP is not realistic in the current market, regardless of what Nvidia’s own storefront says. Builders who need a 24GB-plus-class gaming GPU this quarter are facing either secondary-market prices north of $7,000 or a wait with no committed end date from Nvidia. The rumored 24GB RTX 5080 variant, if it materializes with real pricing and a real launch date, could become the more realistic high-VRAM option for gamers who are priced out of the 5090 tier entirely, but it remains unconfirmed and should not be treated as a purchasing plan.
Our earlier coverage of the RTX 5090 shortage and its 5x MSRP pricing tracked the consumer-side price shock as it unfolded in September. This story picks up where that one left off: not just that prices spiked, but a specific, named mechanism, GB202 die reallocation, that would explain why the spike has not corrected itself.
Competitive landscape: AMD and Intel’s opening
Nvidia’s flagship squeeze is, in theory, the kind of opening AMD and Intel would want. AMD’s Radeon RX 9070 XT has already seen European pricing climb to €819.90 in mid-October tracking, up €52 in roughly three weeks, which suggests AMD is not immune to the same packaging and memory cost pressure, even without a direct GB202-style dual-use die conflict. Intel, meanwhile, has leaned into the midrange with its Arc B580, a card that remains positioned as a value alternative rather than a flagship competitor. Neither company currently ships anything that directly replaces a 32GB-class RTX 5090 at the enthusiast tier, which means Nvidia’s supply problem is not translating into an AMD or Intel sales surge at the high end; it is simply leaving a gap at the top of the market that nobody is filling.
That gap matters more in 2026 than it would have a few years ago, because the same AI-driven memory and packaging shortage squeezing Nvidia’s gaming lineup is also pushing up DDR5 system memory prices, meaning builders chasing a high-end gaming PC are getting squeezed from two directions on the same bill of materials at once.
Market impact: what Nvidia’s silicon math says about its priorities
Nvidia’s gaming segment still matters to the company’s revenue, but it is no longer the business investors are pricing Nvidia on. The company’s data-center and AI-accelerator business has become the dominant driver of Nvidia’s valuation, and that dynamic has been visible in how Nvidia’s own AI infrastructure commitments have scaled through 2026. Against that backdrop, a leaked report that GB202 dies are being funneled toward RTX Pro and data-center products is entirely consistent with rational capital allocation, even if it is uncomfortable news for anyone trying to buy a GeForce card at MSRP. A wafer that becomes an RTX Pro card commands a dramatically higher sale price than the same wafer turned into a $1,999 GeForce board, and when fabrication capacity is the binding constraint, that pricing gap is the only variable that matters to a public company’s allocation decisions.
None of this requires Nvidia to lie about discontinuing the RTX 5090. The company can tell the truth, that shipments continue, that the card has not been formally discontinued, and still be allocating an ever-shrinking share of GB202 output to GeForce every quarter. That is the uncomfortable middle ground this story has settled into, and it is why Fisher’s denial, accurate as it may be on its own narrow terms, did not put the story to rest.
Predictions: where the RTX 5090 story goes next
- Nvidia will not issue a formal end-of-life notice for the RTX 5090 in the near term, since doing so would confirm the leaker narrative it has already denied; expect continued “limited but shipping” language instead.
- Secondary-market prices will stay elevated through Q4 2026 and likely into early 2027, since the underlying constraint is fabrication and packaging capacity, not a temporary retail restocking gap.
- If the rumored 24GB RTX 5080 is real, Nvidia will position it as the practical high-VRAM gaming option for the next model year, effectively demoting the RTX 5090 to a trickle-availability halo product rather than a mainstream flagship.
- AMD and Intel will not gain meaningful enthusiast-tier market share from this specific shortage, because neither currently ships a direct 32GB-class competitor, though both will likely use the moment in marketing around value and availability.
- Expect more shared-die supply stories across the GPU industry as AI accelerator demand keeps competing with gaming silicon for the same fabrication and packaging lines, making the RTX 5090 episode a preview rather than an isolated incident.
Frequently asked questions
Has Nvidia officially discontinued the RTX 5090?
No. Nvidia SVP Jeff Fisher said in a Taipei interview around October 17, 2026, that the RTX 5090 “has definitely not been discontinued” and that shipments would continue, while acknowledging constrained supply-chain capacity.
Why is the RTX 5090 so hard to find in October 2026?
Retail listings disappeared from Nvidia’s own marketplace, Best Buy, and Micro Center between mid-September and mid-October 2026. Leaked factory notices claim Nvidia’s GB202 die is being redirected away from consumer GeForce cards and toward RTX Pro workstation and data-center products, though Nvidia has not confirmed that specific allocation claim.
What is GB202 and why does it matter?
GB202 is the largest die in Nvidia’s Blackwell gaming architecture, used in both the GeForce RTX 5090 and higher-margin RTX Pro workstation cards. Because the same silicon serves both product lines, any capacity squeeze creates a direct allocation conflict between gaming and professional or AI-focused products.
How much is the RTX 5090 reselling for right now?
Third-party Amazon listings reached roughly $8,500, about 4.25 times the card’s $1,999 MSRP, while Newegg listings were closer to $7,300, according to Tom’s Hardware’s October 2026 reporting.
What happened with the missing RTX 5090 shipment in Brazil?
Brazilian distributor Oderço reported that two pallets of Zotac RTX 5090 cards, valued at approximately $470,000, disappeared after arriving at Guarulhos International Airport in São Paulo. The incident is a separate supply-chain story but surfaced amid the same broader RTX 5090 shortage.
Is there a replacement coming for gamers who can’t get an RTX 5090?
Unverified reports point to a revised RTX 5080 with 24GB of VRAM, up from the current 16GB, as a possible alternative. Nvidia has not confirmed specifications, pricing, or a launch date for this variant.
Does this affect other Nvidia GPUs besides the RTX 5090?
The reallocation claims are specific to the GB202 die used in the RTX 5090 and RTX Pro cards built on the same silicon. Other GPUs in Nvidia’s lineup, including the RTX 5080, use different dies, though the broader AI-driven memory and packaging shortage has pushed up prices across the gaming GPU market more generally.
Why would Nvidia prioritize AI chips over gaming GPUs built on the same silicon?
RTX Pro workstation and data-center cards built on GB202 sell for far more than a $1,999 GeForce card. When fabrication and advanced packaging capacity is the binding constraint, routing a limited number of dies toward the higher-margin product line is a straightforward capital-allocation decision, even though Nvidia has not confirmed it is doing so at the expense of GeForce production.