Nvidia has reportedly stopped assigning new GB202 silicon to the GeForce RTX 5090, redirecting the chip exclusively to its RTX Pro professional and AI-focused lineup. The claim, which surfaced on October 10, 2026, through hardware leakers and an alleged factory notice, has not been confirmed by Nvidia. But it lands at a moment when the RTX 5090 has already vanished from first-party US retail shelves and third-party sellers are asking as much as $9,500 for a card that launched at $1,999. For gamers who have spent a year watching the flagship GPU slip further out of reach, the question is no longer whether prices are high. It is whether the card is being phased out altogether.
The story traces back to posts from hardware leakers MEGAsizeGPU and hongxing2020, later picked up by outlets including HotHardware, Wccftech, and HWBusters. None of them have independent confirmation from Nvidia, and the company has issued no statement on an end-of-life date for the RTX 5090. What is verifiable, though, is the retail reality underneath the rumor: a GPU that is functionally unobtainable at its sticker price, a secondary market charging multiples of MSRP, and a professional GPU lineup that generates far more revenue per die than any GeForce card ever could.
What the Leaked Factory Notice Actually Claims
According to a notification from the factory, Nvidia will no longer manufacture any RTX 5090 models, including the D variants sold in China, as reported by leaker hongxing2020 and relayed by Shattered.io. The same notification reportedly states that going forward, the highest-end GeForce card will be a new RTX 5080 24GB variant, while the existing RTX 5080 16GB model gets discontinued. A separate claim from leaker MEGAsizeGPU put it more bluntly: “All future GB202 supplies are RTX Pro exclusive,” a statement reported by Shattered.io on October 10, 2026.
GB202 is the full-fat Blackwell die that powers the RTX 5090. It is the same silicon, binned differently, that goes into Nvidia’s RTX Pro 6000, a workstation card that ships with 96GB of memory and lists at roughly $16,000, according to pricing cited by TokenPost. Put the two side by side: a GeForce card that sold for $1,999 at launch and a professional card built from the same die that sells for eight times that amount. From a margin standpoint, every GB202 wafer Nvidia diverts away from gaming and into RTX Pro or AI infrastructure is a wafer that earns dramatically more per unit. That is the business logic underpinning the rumor, even without an official confirmation of a hard production stop.
It is worth being precise about what is and is not established here. No leaked document includes a quantified production-cut percentage or a unit count. No AIB partner, including ASUS, MSI, or Gigabyte, has issued a public statement confirming or denying the claim. Nvidia itself has stayed silent. The most defensible characterization, repeated across the outlets covering this story, is that this remains an unverified supply-allocation rumor layered on top of an already documented and severe shortage.
The RTX 5090 Has Already Disappeared From US Retail
Whatever Nvidia’s internal allocation decisions turn out to be, the visible supply chain backs up the premise that something has changed. Newegg, Amazon, Best Buy, and B&H currently show no first-party RTX 5090 stock in the US, according to Ground News. Third-party marketplace sellers are asking between $6,500 and $9,500 for the same card that carried a $1,999 MSRP at launch. A September 14 retail check found the cheapest available listing at $6,395 on Amazon’s marketplace, while Newegg listings reached $8,699.
Separately, an October 7 price tracker run by Capital and Compute put the RTX 5090’s marketplace floor at $7,199.99, which works out to roughly 260% above MSRP. The RTX 5080, the next card down the stack, isn’t faring much better: its lowest tracked price sits at $1,653.99 against a $999 MSRP, a 65.6% markup. Micro Center remains one of the few outlets with limited in-store stock, but normal online supply across the US has thinned to almost nothing.
This isn’t purely a gamer-versus-scalper story anymore, either. AI server builders have reportedly started buying retail RTX 5090 cards in bulk, according to Ground News reporting that cites the card’s strong price-to-performance ratio for local AI inference compared to data-center-grade hardware. That adds a second source of demand pressure on top of ordinary gaming demand and whatever supply constraints Nvidia is imposing from the factory side. Tech Insider covered the earlier phase of this squeeze in RTX 5090 Resale Prices Hit 2.4x MSRP in a Month, and the trend documented there has only accelerated in the weeks since.
GB202, RTX Pro, and the Economics of Die Allocation
To understand why Nvidia might make this call, it helps to look at where GB202 silicon actually earns the most money. The RTX 5090 carries 32GB of GDDR7 memory and launched at $1,999. The RTX Pro 6000, built from the same die family, ships with 96GB of memory and lists at approximately $16,000. Even accounting for the far larger memory footprint and workstation-grade certification, the per-die revenue gap is enormous, and that gap only widens when the comparison shifts to Nvidia’s broader AI infrastructure business, where data-center GPUs routinely sell in volume at prices gaming hardware has never approached.
Memory costs add another layer to the calculation. Market data cited by German outlet PC Games Hardware estimates that the RTX 5090’s 32GB of GDDR7 costs Nvidia roughly $320 at current component pricing, or about $20 per memory IC, citing analysis from TrendForce. That figure represents around 16% of the card’s $1,999 MSRP, according to a separate breakdown from Capital and Compute’s VRAM cost tracker. With GDDR6 spot pricing also climbing (a September reading put it at $11.84 per GB, per the same tracker), the input costs for high-VRAM gaming cards are rising at the same time AI customers are willing to pay far more for the exact same silicon. That combination makes redirecting GB202 away from GeForce and toward RTX Pro and AI products look, at minimum, financially rational even if Nvidia never confirms a formal policy.
None of the current reporting verifies that GB202 dies are specifically flowing into Nvidia’s data-center families such as Blackwell Ultra, B300, GB300, or the upcoming Vera Rubin architecture. The documented claim is narrower: GB202 supply is being steered toward the RTX Pro professional graphics line, which sits between consumer GeForce cards and full data-center accelerators. Readers should treat any broader claim connecting this story to Vera Rubin or Blackwell Ultra as speculation rather than confirmed fact.
RTX 5090 Pricing: MSRP vs. Street Price, October 2026
| Card | VRAM | Launch MSRP | Lowest Tracked Price (Oct. 2026) | Markup vs. MSRP |
|---|---|---|---|---|
| RTX 5090 | 32GB GDDR7 | $1,999 | $7,199.99 (marketplace floor) | +260.2% |
| RTX 5090 (peak asking price) | 32GB GDDR7 | $1,999 | $9,500 | ~375% |
| RTX 5080 | 16GB GDDR7 | $999 | $1,653.99 | +65.6% |
| RTX Pro 6000 | 96GB GDDR7 | ~$16,000 | $16,000 | N/A (workstation tier) |
| GDDR6 spot price (Sept. 2026) | per GB | N/A | $11.84/GB | Rising |
Those figures, sourced from Capital and Compute’s live tracker and Ground News retail checks, describe asking prices on marketplaces rather than confirmed completed-sale averages. Still, the direction is unambiguous: every tier of Nvidia’s current-generation lineup is trading well above its original price, and the flagship card is trading at a multiple that puts it closer to workstation territory than consumer gaming hardware.
Historical Context: This Isn’t Nvidia’s First Supply Squeeze
Gamers who lived through the 2017-2018 and 2020-2021 cryptocurrency mining booms will recognize the shape of this story, even if the mechanism is different. During those cycles, GeForce cards were diverted toward miners through sheer market demand: miners simply outbid gamers for the same retail inventory, and Nvidia’s own production volumes didn’t necessarily change. Scarcity was a demand-side phenomenon layered on top of normal supply.
What’s being alleged now is structurally different. Rather than gamers and a third party competing for identical retail units, the claim is that Nvidia itself is choosing, at the factory level, to stop making new RTX 5090 units and instead bin that same GB202 silicon exclusively for RTX Pro cards. That is a supply-side allocation decision, not a demand-side bidding war, and it would mark one of the more direct moves away from a flagship consumer SKU in the company’s recent history. Nvidia has increasingly prioritized AI and data-center products in its earnings commentary over the past two years, and the margins on that business dwarf anything the GeForce line generates, so the strategic direction is consistent with the company’s recent financial priorities even where the specific claim remains unconfirmed.
Competitive Landscape: AMD and Intel’s Opening
An RTX 5090 shortage, confirmed or not, creates room for competitors, though neither AMD nor Intel currently fields a direct flagship rival. AMD’s RDNA-based Radeon lineup has focused on the mid-range and upper-mid-range segments, leaving the extreme high end largely uncontested by a true competing SKU. Intel’s Arc lineup, including the Arc B390 found in handhelds like Acer’s Predator Atlas 8, has made inroads in the gaming handheld and mid-range desktop markets but hasn’t attempted to challenge Nvidia at the $2,000-plus tier.
That leaves buyers who need top-tier gaming or local AI inference performance with few alternatives beyond the used market for prior-generation Nvidia hardware. One local-AI-focused outlet, RunAIHome, has already started recommending a used RTX 4090 in the $2,300-$2,800 range as a faster path to a working setup than waiting on RTX 5090 restocks. That dynamic also pushes up demand, and likely price, for the previous generation. Tech Insider’s earlier generation-over-generation comparison, RTX 5090 vs 4090 for 3D Rendering, remains a useful reference point for buyers weighing whether the previous flagship is still worth sourcing secondhand.
Market Impact: Who Gets Hurt and Who Benefits
The most immediate losers in this scenario are PC gamers and small-scale AI builders who were hoping to buy an RTX 5090 at anything close to its original price. If GB202 truly stops flowing to GeForce SKUs, the existing retail channel will run dry as remaining AIB inventory sells through, and prices on the secondary market will likely climb further rather than stabilize. Content creators and 3D artists who rely on the RTX 5090’s 32GB frame buffer for large scene and video workloads face the same squeeze, with fewer alternatives below the professional tier.
AIB partners sit in an odd middle position. Board makers like ASUS, MSI, and Gigabyte earn margin on assembling and selling finished RTX 5090 cards, so a supply cutoff reduces their available product to sell at a moment when demand, and achievable pricing, is at its highest. At the same time, none of these partners have publicly confirmed the production change, which suggests either that Nvidia hasn’t formally communicated an end date to its partners, or that partners are choosing not to get ahead of an official announcement.
Nvidia itself is the clear financial beneficiary if the rumor holds. Diverting GB202 wafers to RTX Pro cards selling near $16,000, instead of GeForce cards selling (in theory) at $1,999, multiplies the revenue generated per wafer many times over. Given that memory and die costs are also climbing, as reflected in the roughly $320 GDDR7 bill-of-materials estimate for the RTX 5090’s 32GB configuration, the margin case for prioritizing RTX Pro only strengthens as component costs rise.
What a Rumored RTX 5080 24GB Would Change
Part of the leaked claim includes a successor product: an RTX 5080 24GB variant that would become the new ceiling of the GeForce lineup if the RTX 5090 is indeed discontinued. Wccftech’s reporting frames this as Nvidia’s planned consolation for gamers losing access to the 5090, doubling the existing RTX 5080’s 16GB frame buffer without introducing an entirely new die configuration. If accurate, this would represent a meaningful spec bump for creators and gamers running memory-hungry workloads, even if it falls well short of the RTX 5090’s 32GB capacity.
It’s worth treating this detail with the same caution as the rest of the story. No roadmap, launch date, or pricing for an RTX 5080 24GB has been confirmed by Nvidia, and leaks about memory-configuration changes to existing dies have a mixed track record of panning out exactly as described. Tech Insider’s earlier coverage in RTX 6080 Launch Order Hinges on $15 GDDR7 Chips and RTX 6080 Could Beat RTX 6090 to Shelves showed a similar pattern of GDDR7 supply constraints shaping Nvidia’s next-generation launch sequencing, which lends some circumstantial credibility to the idea that memory availability, not just AI demand, is steering product decisions across the whole stack.
The Broader DRAM and Memory Squeeze Behind the Story
The RTX 5090 situation doesn’t exist in isolation. It’s unfolding against a backdrop of a global memory shortage that has already pushed PC shipments down 20.1% in the third quarter of 2026, as Tech Insider reported in PC Shipments Drop 20.1% in Q3 2026 as DRAM Soars, and driven gaming hardware sales to a 13-year low, detailed in Game Hardware Sales Hit 13-Year Low as DRAM Soars. Rising GDDR6 and GDDR7 spot prices, cited above at $11.84 per GB for GDDR6, aren’t unique to Nvidia’s flagship. They’re squeezing margins across every GPU maker and every memory-dependent product category simultaneously, from gaming laptops to enterprise servers.
That context matters for interpreting the RTX 5090 story correctly. Even if Nvidia never confirms an explicit decision to end RTX 5090 production, rising memory costs alone create pressure to prioritize whichever products generate the best return per gigabyte of GDDR7 consumed. RTX Pro and data-center products clear that bar by a wide margin over consumer GeForce cards, which means the economic forces described in this story would exist in some form regardless of whether a specific factory notice is real.
Expert and Industry Voices
Much of what’s known about this story comes directly from the hardware leaker community rather than named Nvidia executives or analysts, since the company has not issued an official statement. Leaker MEGAsizeGPU wrote that “All future GB202 supplies are RTX Pro exclusive,” a claim reported by Shattered.io on October 10, 2026. That single-line claim has become the anchor point for nearly every subsequent report on the alleged production shift, despite the absence of corroborating documentation from Nvidia or its manufacturing partners.
Separately, leaker hongxing2020 described the contents of an alleged factory notification stating that Nvidia would stop manufacturing RTX 5090 models entirely, including the China-market D variant, with the RTX 5080 24GB becoming the new top-tier GeForce card. That claim was relayed through outlets including TweakTown and Shattered.io, though it similarly lacks direct Nvidia confirmation. Readers should weigh both claims as sourced from anonymous or pseudonymous hardware leakers rather than verified corporate disclosures, a distinction every outlet covering the story, including HotHardware, Wccftech, and HWBusters, has been careful to flag in their own reporting.
Predictions: Where This Goes From Here
- Nvidia will eventually confirm some version of this shift, even if quietly. Given the financial incentives detailed above, a formal end-of-life notice for the RTX 5090, timed around a product refresh announcement, looks more likely than an outright denial.
- RTX 5090 secondary market prices will keep climbing before they fall. With no confirmed restock timeline and AI builders now competing with gamers for existing units, the current $6,500-$9,500 range is more likely to drift higher than lower through the rest of 2026.
- An RTX 5080 24GB variant will materialize within six to nine months. The GDDR7 supply constraints already documented around the RTX 6080 and RTX 6090 launch sequencing suggest Nvidia is actively managing memory allocation across its entire stack, and a memory-boosted RTX 5080 is a logical stopgap.
- AMD and Intel will not launch a direct flagship competitor in this window. Both companies remain focused on mid-range and handheld-class silicon, leaving Nvidia’s RTX Pro strategy largely unchallenged at the extreme high end through at least mid-2027.
- Used RTX 4090 pricing will firm up rather than soften. As local AI builders and gamers alike treat the previous generation as the practical ceiling for new purchases, expect secondhand RTX 4090 prices to resist the usual depreciation curve for an aging flagship.
RTX 5090 vs. RTX Pro 6000: Same Silicon, Different Business
| Spec | GeForce RTX 5090 | RTX Pro 6000 |
|---|---|---|
| Die | GB202 | GB202 |
| Memory | 32GB GDDR7 | 96GB GDDR7 |
| Launch/List Price | $1,999 | ~$16,000 |
| Target Market | Consumer gaming | Professional/workstation, AI |
| Revenue per Die (approx.) | Baseline | ~8x RTX 5090 price |
Laid out this way, the economic logic of the rumor requires no inside knowledge of Nvidia’s manufacturing plans. Even a modest reallocation of GB202 wafer starts away from GeForce and toward RTX Pro translates into a dramatic revenue swing, which is exactly the kind of decision a public company facing both surging AI demand and rising memory input costs would be expected to make, confirmed or not.
What Buyers Should Actually Do Right Now
For anyone currently shopping for a high-end GPU, the practical guidance hasn’t changed much even with this new rumor in play. Paying $6,500 or more for an RTX 5090 at today’s marketplace prices rarely makes sense for a gaming build, since even a modest cooling of the shortage could leave early buyers well above a fair market price. Buyers prioritizing gaming performance over maximum VRAM should look at the RTX 5080, despite its own 65.6% markup over MSRP, or wait for clarity on the rumored RTX 5080 24GB variant.
For local AI inference workloads specifically, a used RTX 4090 in the $2,300-$2,800 range, as currently recommended by RunAIHome, offers a more defensible price-to-performance tradeoff than chasing RTX 5090 listings north of $7,000. Whatever Nvidia ultimately confirms about RTX 5090 production, the near-term reality for buyers is the same: supply is scarce, prices are inflated well beyond MSRP, and waiting for a dramatic price correction in the next few months looks like the less likely outcome.
Frequently Asked Questions
Has Nvidia officially confirmed it is ending RTX 5090 production?
No. The claim originates from hardware leakers MEGAsizeGPU and hongxing2020 and an alleged factory notification reported by outlets including HotHardware, Wccftech, and Shattered.io. Nvidia has not issued a statement confirming or denying the report as of October 11, 2026.
What is GB202 and why does it matter?
GB202 is the Blackwell architecture die used in the GeForce RTX 5090. The same die, differently configured, also powers Nvidia’s RTX Pro 6000 professional card. The rumor claims Nvidia is redirecting all future GB202 output to RTX Pro products instead of GeForce.
How much does an RTX 5090 cost right now?
Marketplace prices as of early October 2026 range from roughly $6,500 to $9,500, against a $1,999 launch MSRP, according to tracking from Ground News and Capital and Compute. First-party retail stock is largely unavailable in the US.
Is there a replacement card coming for gamers?
Leaked reports describe a potential RTX 5080 24GB variant that would become the top GeForce SKU if the RTX 5090 is discontinued, with the existing RTX 5080 16GB reportedly being phased out. No official specs, pricing, or launch date have been confirmed by Nvidia.
Why would Nvidia prioritize RTX Pro over GeForce?
The RTX Pro 6000 lists at roughly $16,000 compared to the RTX 5090’s $1,999 launch price, despite sharing the same GB202 die family. With rising GDDR7 memory costs and surging AI/professional demand, each GB202 wafer generates substantially more revenue when built into RTX Pro cards.
Are AMD or Intel positioned to fill the gap?
Not directly. AMD’s Radeon lineup and Intel’s Arc lineup both compete primarily in the mid-range and handheld GPU segments, and neither currently offers a card targeting the RTX 5090’s extreme high-end tier.
Should I buy an RTX 5090 right now?
At current marketplace prices of $6,500 and up, most analysts and trackers covering the story suggest waiting, since prices well above MSRP carry real risk of a correction if supply normalizes or demand cools.
Is this similar to the cryptocurrency mining GPU shortages of 2017-2021?
Not exactly. Those shortages were driven by demand-side competition for identical retail stock. This rumor describes a supply-side allocation decision at the factory level, with Nvidia itself reportedly choosing to stop producing new RTX 5090 units rather than simply losing a bidding war to another buyer category.