The RTX 5090's street price has detached from its $1,999 MSRP — illustrative concept of the $10,000 GPU market. (AI-generated illustration)
The Nvidia GeForce RTX 5090 is the most powerful consumer graphics card ever made — and in October 2026, it has become something stranger: a ghost. Walk into a store, open a retailer website, refresh a price tracker, and you will find the same story everywhere. The shelves are empty. The listing pages are empty. And the few cards that do surface change hands at prices that would have seemed absurd a year ago.
Let us be precise about those numbers, because the precision is the story. Street prices for the RTX 5090 currently run from $7,488 to $10,000 — 3.75 times to 5 times the card's $1,999 launch MSRP. Amazon listings hover near $8,500. Newegg listings push as high as $9,600 to $10,000. And on Tom's Hardware's US price tracker, checked on October 16, 2026, the single best available price anywhere in the United States was $7,488. Not the best price at a good retailer — the best price anywhere. In normal times, a flagship GPU sells at or near its list price. Right now, the cheapest RTX 5090 on the American internet costs more than most complete gaming PCs.
This is not a normal GPU cycle. It is not a normal shortage, either. The RTX 5090 is not discontinued — Nvidia has said so directly. There is no official price increase, no new tariff, no single dramatic event to point at. Instead, the world's fastest gaming graphics card has simply… evaporated from the shelves, and the dribble of units that remains is priced like a luxury good. Enthusiasts have started asking the same question with real anxiety: where did the RTX 5090 go?
This article is a full investigation into that mystery, built only from verified reporting. We will walk the price ladder retailer by retailer, decode what Micro Center's blunt "Not Carried" label actually signals, examine the leading theories for the shortage — each one grounded in a reported fact — and give you an honest answer to the question that matters most: what should you actually buy instead of paying $10,000 for a graphics card?
Key Takeaways
- Street prices run $7,488 to $10,000. That is 3.75x to 5x the $1,999 MSRP — the worst price-to-MSRP ratio ever seen on a flagship Nvidia card.
- The cheapest available price in the US, per Tom's Hardware's tracker on October 16, was $7,488. There is no hidden retailer selling them cheap; that is the floor.
- Micro Center has tagged its RTX 5090 listings "Not Carried," meaning the chain is not even attempting to stock the card — the strongest retail-level signal of how thin supply has become.
- Nvidia says the RTX 5090 is NOT discontinued. The card is still an active product; this is a supply problem, not an end-of-life.
- ZDNet Korea reports Nvidia is raising GDDR memory kit prices to board partners, which directly increases the cost of building every card — a concrete, reported cost pressure behind the shortage.
- If you need high-end performance now, buy an RTX 5080 or 5070 Ti class card at a sane price and wait. Paying scalper prices for a 5090 is the single worst move a buyer can make.
- Nobody can honestly promise when stock will return. The signals to watch are partner pricing, retailer listing statuses, and price-tracker trends — we will show you exactly what to monitor.
The Price Ladder: What an RTX 5090 Actually Costs in October 2026
Forget the $1,999 MSRP printed on the box. That number has not reflected reality for months, and in October 2026 the gap between list price and street price has become a canyon. Here is the RTX 5090 price ladder as it actually stands, retailer by retailer — the landscape every buyer is facing right now.
| Retailer / Source | Observed Price | Multiple of $1,999 MSRP | Status |
|---|
| Tom's Hardware US tracker (Oct 16, 2026) | $7,488 | 3.75x | Best available price anywhere in the US |
| Amazon listings | ~$8,500 | ~4.25x | Third-party marketplace pricing |
| Newegg listings | ~$9,600–$10,000 | ~4.8x–5x | Highest observed street pricing |
| Micro Center | — | — | Tagged "Not Carried" |
Read that table slowly. The cheapest RTX 5090 you can buy in America costs $7,488 — and the spread between the cheapest and most expensive listings is only about $2,500, which tells you this is not one greedy seller: it is a market-wide reality. When the floor is 3.75 times MSRP, you are not looking at normal scalping. You are looking at a product whose entire supply chain has repriced itself.
For context on how extraordinary this is: flagship GPUs have always sold above MSRP during launch crunches — the RTX 3090 and RTX 4090 both had their scalper eras. But those spikes were launch phenomena that faded within months as supply caught up. The RTX 5090's situation is different in kind. We are not in a launch window. This is a mature product in the second half of its generation, and prices are still climbing to levels nobody has seen before. A $10,000 graphics card is not a gaming component; it is a luxury purchase, and the fact that cards still sell at these prices tells you everything about how desperate the remaining demand is.
It is also worth understanding who sets these prices. At $8,500 on Amazon and $9,600-plus on Newegg, you are mostly looking at marketplace sellers and smaller resellers, not the big first-party inventory of a year ago. When first-party stock exists, it sells at or near MSRP — which is exactly why its absence is so loud. The price ladder above is essentially a map of the vacuum: where Nvidia's own supply chain ends, the gray market begins, and the gray market sets its own price.
From $1,999 to $10,000: How the RTX 5090 Became a Luxury Good
To understand the mystery, you need to appreciate how far this card has traveled. The RTX 5090 launched with an official price of $1,999 — already a breathtaking number for a consumer graphics card, and already controversial. Reviewers noted it was the most expensive xx90-class card Nvidia had ever released, and buyers swallowed it because the performance was genuinely generational: the first GPU to make native 4K at high refresh rates feel effortless, with headroom for the heaviest ray tracing workloads and AI-accelerated creative work.
The card's journey from a $1,999 flagship to a $7,488–$10,000 gray-market trophy is the defining GPU story of 2026, and it happened without a single official price increase. Nvidia never raised the MSRP. No announcement, no press release, no new model. The repricing happened entirely in the shadows — in distributor contracts, in board partner cost sheets, and in the marketplace listings where actual transactions occur. That is what makes this shortage so disorienting: officially, nothing changed. Unofficially, everything did.
Compare the trajectory to the RTX 4090's infamous 2023–2024 era, when that card's street price drifted to roughly double its MSRP at the worst moments. The 5090 has blown past that benchmark and kept going — 3.75x to 5x is not a drift, it is a detachment. And unlike the 4090's spikes, which were tied to identifiable demand surges, the 5090's climb has been slow, grinding, and relentless, the kind of price action you see when supply is quietly shrinking month after month while demand never gets the memo.
This is why the shortage-mystery framing matters more than the price alone. A $10,000 price tag on a listing is a symptom. The disease is the vanishing supply — and supply does not vanish without a cause. Something upstream changed. The rest of this article is about figuring out what.
"Not Carried": What Micro Center Is Really Telling You
Of all the data points in this story, one of the most eloquent is also the shortest: Micro Center, America's most beloved brick-and-mortar PC hardware chain, has tagged its RTX 5090 listings "Not Carried."
That phrasing deserves a close read, because it is deliberately different from the labels retailers normally use. "Out of Stock" means: we sell this, and we will have more — check back, join the waitlist, set a notification. "Discontinued" means: this product is dead, do not expect it again. "Not Carried" is the third category, and it is the most damning of all. It means: Micro Center is not even attempting to stock the RTX 5090 right now. The product is absent from its active merchandising plan.
Why does one chain's label matter so much? Because Micro Center is not a random corner shop. It is a retailer famous for aggressive GPU stocking — during previous shortages it ran in-store-only policies, anti-bot purchase limits, and bundle deals specifically to get cards into gamers' hands. It has relationships with every major board partner and distributor. If Micro Center cannot — or will not — carry the RTX 5090, that is not a logistics hiccup. It is a signal that the wholesale channel itself has dried up to the point where a major retailer sees no point in maintaining the listing as an active product.
There is a second, subtler reading of "Not Carried" worth considering. Retailers carry products they can sell at competitive prices with healthy margins and reliable replenishment. When wholesale costs spiral — when board partners themselves are paying more for memory kits and passing those costs down — a retailer's math breaks. It cannot advertise a $1,999 MSRP product it would have to sell at $4,000 just to break even, and it cannot stock a product whose replenishment dates are fiction. "Not Carried" may be Micro Center's honest way of saying: at current wholesale economics, this product does not work for us.
Either way, the message to buyers is the same. The most gamer-friendly retailer in the country has effectively stepped out of the RTX 5090 market. When the store that built its reputation on getting GPUs to real people gives up on a card, you are not looking at a temporary blip — you are looking at a structural supply failure.
Is the RTX 5090 Discontinued? No — and That Distinction Matters
Whenever a flagship product vanishes from shelves, the first rumor to spread is always the same: it must be discontinued. Maybe Nvidia is clearing the decks for a 60-series. Maybe the 5090 had a hardware flaw. Maybe the whole Blackwell consumer line is being quietly wound down. These rumors have circulated about the RTX 5090 for months, and they deserve a direct answer.
Nvidia says the RTX 5090 is NOT discontinued. The card remains an active, current product in Nvidia's lineup. That is the company's stated position, and it is an important anchor in a story full of speculation — because it rules out the simplest explanation for the empty shelves and forces us to look at the harder ones.
Think about what "not discontinued" actually implies. It means Nvidia still intends to make and sell this card. It means board partners still have the product in their catalogs. It means the architecture, the drivers, the warranty support — the entire product infrastructure — is still live. And yet the cards are not reaching shelves. That combination — active product, absent inventory — is precisely the signature of a supply-side constraint rather than a demand-side or lifecycle event. The factory lines and the store shelves have become disconnected, and something in between them is broken or deliberately throttled.
The distinction also matters for buyers, because it changes the strategy completely. If the card were discontinued, the rational move would be to buy whatever you can find now, because prices would only rise as remaining stock drains away. But an active product with a supply problem is a different animal: supply problems can resolve, and when they do, prices can fall hard and fast. Paying $8,500 for a card that is still in production is a bet that supply will never recover — and Nvidia itself is telling you that bet is wrong.
There is, of course, a cynical reading: companies sometimes deny discontinuations right up until they announce them. But nothing in the reporting supports that reading here. No successor product has been announced, no end-of-life notices have leaked from board partners, and the memory-cost story (which we will get to) suggests partners are being squeezed on costs, not winding down production. The honest interpretation of Nvidia's statement is the straightforward one: the 5090 is alive, and its absence from shelves is a supply story, not a eulogy.
Empty shelves and no restock in sight: the hunt for an RTX 5090 at a fair price continues. (AI-generated illustration)
The Hunt: Why You Cannot Find One Anywhere
Let us now do what every frustrated buyer has done: actually try to buy an RTX 5090 in October 2026. The experience is instructive, because the pattern of failure tells you as much as any press release.
Start with the big online retailers. Amazon: listings exist, but they are marketplace listings at roughly $8,500 — not Amazon-direct stock at MSRP. That distinction matters enormously. Amazon-direct inventory means the manufacturer or an authorized distributor supplied the product through normal channels. Marketplace listings mean third parties are reselling units they acquired elsewhere. When a product exists only as marketplace listings, the official supply chain has effectively stopped delivering it.
Newegg: same story, steeper prices — up to roughly $9,600 to $10,000. Again, these are the prices of resellers operating in a vacuum, not of a healthy retail channel. Best Buy, historically one of the more reliable sources for Founders Edition cards: nothing meaningful. And Micro Center, as we have seen, has stepped out entirely with its "Not Carried" tag.
The pattern is total: no major retailer has healthy, replenishing, first-party RTX 5090 inventory in October 2026. Every available unit is a resale, a leftover, or a gray-market import. This is what makes the shortage a mystery rather than a mere inconvenience. In a normal shortage — a launch crunch, a demand spike — you can still see the shape of the supply chain. Restocks happen. Waitlists move. Some region, some retailer, some bundle has cards. Here, the supply chain has no visible shape at all. It is as if the cards stopped being made, even though Nvidia insists they have not been discontinued.
Price trackers confirm the bleakness. Tom's Hardware's US GPU price tracker, checked October 16, 2026, found a best available price of $7,488 — and that "best" price would have been a scandalous scalper price in any previous generation. The tracker exists precisely to find the cheapest legitimate listing; when its answer is $7,488, the answer to "where can I buy one at a fair price" is: nowhere.
So where did the stock go? The cards did not evaporate — silicon does not do that. Every missing RTX 5090 is somewhere: in a warehouse, on a pallet, in a mining-adjacent operation, in a workstation, in a scalper's storage unit, or simply never manufactured because the components to build it were allocated elsewhere. The theories below are the serious candidates, each grounded in a reported fact. None of them alone explains everything. Together, they sketch the outline of the answer.
Theory One: The Memory Squeeze
The strongest lead in this mystery comes from ZDNet Korea, which reports that Nvidia is raising GDDR memory kit prices to its board partners. This single reported fact, if you understand how graphics cards are built, explains an enormous amount.
Here is the mechanics. Nvidia does not manufacture graphics cards itself — not the ones you buy in stores. It designs the GPU chip and sells it, along with memory kits, to board partners: the companies whose names are on the boxes. Those partners assemble the finished card — GPU, memory, power delivery, cooler, packaging — and sell it to distributors and retailers. The partner's cost structure is therefore Nvidia's pricing plus memory plus everything else. When Nvidia raises the price of the GDDR memory kits it supplies, every partner's cost per card rises by definition.
Now follow the money downstream. A board partner facing higher kit costs has exactly three options: absorb the cost and lose margin, raise its wholesale price to distributors, or build fewer cards. In practice, partners do a mix of the second and third — and both push street prices up. Higher wholesale prices flow straight into retail pricing. Fewer cards built means thinner supply, which hands pricing power to whoever holds inventory. Either way, the buyer pays more.
Why would memory costs be rising in the first place? This is where the theory connects to the broader 2026 semiconductor landscape. GDDR memory — the specialized high-bandwidth memory that makes a card like the 5090 possible — is produced by a tiny handful of suppliers, and those suppliers are simultaneously feeding the insatiable memory appetite of the AI data center boom. When data center demand soaks up memory fab capacity, consumer GPU memory gets more expensive. It is the same silicon, the same factories, the same physics — just a bidding war that gamers cannot win.
The memory-squeeze theory is the only explanation grounded in direct reporting about Nvidia's own pricing behavior — and it predicts exactly what we observe: rising costs, thinning partner supply, and street prices detached from MSRP. It also explains why the shortage feels so quiet. There is no dramatic event, no factory fire, no export ban headline. Just a cost sheet that got worse, quarter after quarter, until building a 5090 at anything near $1,999 stopped making sense for anyone in the chain.
A note on honesty: ZDNet Korea's report tells us memory kit prices to partners are rising. It does not tell us by how much, and it does not by itself prove that memory costs are the sole or even primary cause of the shortage. Treat this theory as the leading hypothesis, not the verdict. But it is a hypothesis with a reported fact at its core, which is more than can be said for most GPU shortage folklore.
Theory Two: Allocation — When AI Eats First
The second theory is about priorities, and it starts with a simple question: if Nvidia is still making the silicon that goes into the RTX 5090, where is that silicon going, if not into graphics cards on store shelves?
The honest answer is that we do not know for certain — but the economics of 2026 point in one direction. The same GPU architecture that powers the RTX 5090 shares DNA, manufacturing capacity, and engineering resources with the data center products that are currently the most profitable silicon on earth. Every wafer of advanced process capacity is finite. Every unit of that capacity allocated to a $1,999 consumer card is capacity not allocated to products that sell for many times more to customers with effectively unlimited budgets.
This theory holds that the RTX 5090 shortage is, at least in part, an allocation decision: finite manufacturing capacity flowing toward the highest-margin customers first, with consumer flagships getting whatever is left. Note carefully what this theory is and is not. It is not a claim that Nvidia "abandoned gamers" out of malice, and it is not based on any leaked internal memo. It is an inference from the structure of the market: when one customer segment will pay almost any price and another is price-sensitive at $1,999, a rational supplier under capacity constraints serves the first segment first. The empty shelves are the visible residue of that invisible triage.
There is supporting texture for this view, though it stays in the realm of informed reasoning rather than reported fact. Throughout 2026, every signal from the semiconductor industry has pointed to data center demand consuming an ever-larger share of advanced-node capacity. Consumer GPU launches have felt thinner. Restocks have been slower across the board, not just for the 5090. When the whole consumer stack feels the pinch simultaneously, allocation — not any single product's problem — is the natural explanation.
If this theory is right, it also tells you something important about when the shortage ends: it ends when capacity expands enough that serving data center demand no longer requires starving the consumer channel, or when data center demand cools. Neither is something a buyer can control or precisely predict — which is why the "what to buy instead" section of this article is not a consolation prize but the main event.
Theory Three: The Scalper Flywheel
The third theory is the one every gamer already believes, and it deserves a fair hearing: scalpers and the resale ecosystem are amplifying the shortage into something far worse than the underlying supply gap.
The mechanism is straightforward and well-documented from previous GPU crunches. When legitimate supply is thin, every unit that does reach retail is instantly bought — often by automated purchasing tools — and relisted at a markup. Those markups then become the reference price: the next seller lists at the going resale rate, not at MSRP. Buyers who need the card urgently pay the resale price, which validates it, which encourages more resellers to enter, which ensures that legitimate buyers never see a fair price. It is a flywheel, and once it spins up, it sustains itself even if underlying supply improves modestly.
The $7,488-to-$10,000 spread we observe is the signature of a mature resale market: stable, liquid, and completely detached from the manufacturer's price. In a healthy market, the spread between the cheapest and most expensive listing of an identical product is small, because competition compresses it. A $2,500 spread on a single SKU means the market has no anchor — no steady MSRP inventory forcing resellers to compete downward. The flywheel is spinning freely.
But here is the crucial nuance that separates serious analysis from internet folklore: scalpers do not create shortages, they exploit them. A scalper can only resell a card that was manufactured and sold once already. If supply were healthy, resellers would be undercut by legitimate retail within weeks — the flywheel would never get traction. The fact that the flywheel has been spinning for months tells you the underlying supply really is broken. Scalpers are the amplifiers, not the source. Blaming them exclusively is emotionally satisfying and analytically incomplete.
This matters for strategy. Anti-scalper measures — purchase limits, queues, lotteries — help at the margins, but they cannot fix a supply problem. Only more cards fix a supply problem. Which brings us to the question every reader is actually asking.
When Will RTX 5090 Stock Return? An Honest Timeline
Let us be blunt: anyone who gives you a confident date for the RTX 5090's return to normal pricing is guessing. The honest answer is that nobody outside Nvidia's supply chain knows when meaningful stock will return, and the signals worth watching are specific and checkable. Here is how to read them like an analyst instead of a hopeful forum poster.
Signal one: board partner pricing behavior. The ZDNet Korea report on rising GDDR kit costs is the fundamental variable. If memory costs stabilize or fall, partners can rebuild margin at lower street prices, and supply should thicken. Watch for reporting — from VideoCardz, Tom's Hardware, or the Korean and Taiwanese press — on memory pricing trends and partner cost sheets. A reversal in that story is the earliest fundamental indicator that relief is coming.
Signal two: retailer listing statuses. Micro Center's "Not Carried" tag is a canary. When it flips back to an active listing — even at an elevated price — it means the wholesale channel has reopened enough for a major retailer to re-engage. Similarly, watch for the return of first-party listings on Amazon and Best Buy at or near MSRP. The return of legitimate retail is the moment the scalper flywheel loses its anchor-less freedom.
Signal three: the price-tracker floor. Tom's Hardware's tracker put the US floor at $7,488 on October 16. That number is your dashboard. If it drifts down toward $5,000, then $4,000, then $3,000 over successive checks, supply is healing — even before any official announcement. If it climbs toward five figures and stays there, the shortage is deepening. Check it the way you would check a stock price: regularly, without emotion, and with a plan for what each level means for your buying decision.
Signal four: Nvidia's own communications. The company has said the card is not discontinued. If that changes — if a successor is announced or an end-of-life notice appears — the strategy flips entirely, because remaining stock becomes a wasting asset and prices may spike further before collapsing. Conversely, any official acknowledgment of supply improvement would be a strong buy-the-dip signal for patient buyers.
What should you actually expect? Reason with base rates. Severe GPU shortages historically resolve on timelines of months, not weeks — and this one has structural drivers (memory costs, allocation pressure) rather than a single transient cause. A reasonable posture is to plan as if fair-price 5090s are many months away, while monitoring the four signals above for early evidence of improvement. Hope is not a strategy; a watchlist is.
And critically: do not let the timeline question paralyze you. The "what to buy instead" section below is not about settling — it is about recognizing that a graphics card is a tool, and the right tool at a fair price beats the perfect tool at a 5x markup every single time.
What to Buy Instead of the RTX 5090
Here is the section that will save you the most money in this entire article. Do not pay $7,488 to $10,000 for an RTX 5090. Full stop. No gaming workload, no creative project, no fear of missing out justifies a 3.75x to 5x markup on a still-in-production card. The rational move is to buy a cheaper card now, enjoy it, and revisit the 5090 when — not if — the market normalizes. Here are the serious alternatives, ranked by sense.
| Alternative | Class | Who It Is For | Why It Makes Sense Now |
|---|
| RTX 5080 | High-end (one tier down) | 4K gamers who want flagship-adjacent performance | Far closer to sane pricing; delivers the large majority of the 5090's gaming experience at a fraction of the street price |
| RTX 5070 Ti | Upper mid-range / high-end | 1440p and entry-4K gamers | Excellent price-to-performance when found near list; handles modern titles with high settings comfortably |
| Previous-generation flagship (used) | High-end (last gen) | Budget-conscious power users | Used-market prices reflect real depreciation; a careful purchase can deliver strong 4K performance for a fraction of 5090 street prices |
| Wait with your current card | — | Anyone whose current GPU still runs their games | Costs nothing; every month you wait is a month closer to normalized pricing |
The RTX 5080 deserves the top slot because it is the closest thing to a 5090 that exists in a functioning market. It is the same generation, the same architecture family, the same feature set — modern ray tracing, current-generation upscaling, full driver support — sitting one tier down the stack. In real gaming, the gap between a 5080 and a 5090 is real but narrow compared to the gap between their street prices: you are giving up a meaningful but not transformative slice of performance to save thousands of dollars. That is the best trade in PC hardware right now.
The RTX 5070 Ti is the value play. It targets the gamer who plays at 1440p or who is willing to lean on upscaling technologies at 4K — which, in 2026, is a completely respectable way to play. Modern upscaling has matured to the point where the visual difference at 4K is subtle in most titles, and the 5070 Ti class has the headroom to drive those modes well. If your monitor is 1440p high-refresh, this card is arguably the smarter buy than a 5090 even at MSRP.
The used previous-generation flagship is the contrarian pick, and it requires care: buy from reputable sellers, check warranty transferability, stress-test on arrival. But the logic is unassailable. Last generation's halo card was itself a monster, and its used price reflects a market where sellers compete — the opposite of the 5090's anchor-less resale market. For creators doing GPU rendering or AI inference on a budget, this route can deliver the most raw compute per dollar of any option on this list.
And then there is the option nobody wants to hear but everyone should consider: wait. If your current card runs the games you play at settings you enjoy, the highest-return move is to change nothing. GPU generations are long, game requirements move slowly, and the money you do not spend on a $8,500 card today is money that buys a fairly-priced flagship — or two generations of excellent mid-range cards — tomorrow. Patience is the only strategy with a guaranteed positive return in this market.
One more consideration for creators and AI hobbyists, who are often the most tempted by the 5090's massive memory and compute: evaluate whether your workload actually needs a single 5090, or whether it needs memory capacity and compute that could be served differently — a workstation card, cloud GPU rental for burst workloads, or simply a smaller card while you wait. Cloud GPU rental, in particular, has matured enormously: for intermittent heavy workloads, renting top-tier compute by the hour is dramatically cheaper than buying a $10,000 card to use at 20 percent utilization.
The Waiting Playbook: How to Buy a 5090 at a Fair Price
If your heart is set on the RTX 5090 — and for some buyers, with 4K high-refresh monitors or heavy creative workloads, it genuinely is the right card — then do not buy one today. Buy one correctly, later. Here is the playbook.
Step one: set your target price and write it down. A fair target is the $1,999 MSRP plus a modest premium for the specific board partner model you want — partner cards have always cost somewhat more than the baseline, and that is legitimate. What is not legitimate is any price starting with a 7, 8, 9, or five digits. Your target is MSRP-plus-a-little, and you do not move until the market comes to you. Writing it down matters because it converts a vague intention into a rule, and rules are what protect you from 2 a.m. impulse purchases.
Step two: build your monitoring stack. Bookmark Tom's Hardware's GPU price tracker and check the RTX 5090 floor regularly — remember, $7,488 on October 16 is your baseline; you are watching for sustained downward movement. Set stock alerts through reputable tracker services and retailer notification systems. Follow VideoCardz and the hardware press for supply reporting. The goal is to be informed the week the market turns, not the month after.
Step three: prepare your purchase path for the restock moment. When legitimate supply returns, it will sell out in minutes — the pent-up demand is enormous. Have accounts created and payment methods saved at the retailers most likely to get first-party stock. Know exactly which board partner models you will accept and which you will not. Decide in advance whether you will buy a Founders-style baseline card at MSRP or wait for a premium partner model. The buyers who get cards at fair prices during restocks are the ones who did their deciding beforehand.
Step four: beware the false dawn. The first sign of improving supply will be falling resale prices — $6,000, then $5,000 — and it will be tempting to declare victory and buy. Do not. A falling scalper price is not a fair price. Wait for first-party retail at or near MSRP, or at minimum for the tracker floor to approach the low-$2,000s. The difference between buying at $5,000 in excitement and $2,200 in patience is a whole second graphics card.
Step five: consider the generational clock. Every month you wait is a month closer not just to normalized 5090 pricing but to the eventual next generation. There is a point — unknowable in advance, recognizable in hindsight — where waiting for the 5090 stops making sense because its successor is on the horizon. Keep one eye on the usual rumor cycle, and be willing to update the plan. Flexibility is part of the playbook.
The through-line of all five steps is discipline. The market is currently designed — not by conspiracy but by the emergent logic of scarcity — to separate impatient buyers from their money. Your edge is simply refusing to play that game.
Frequently Asked Questions
Why is the RTX 5090 so expensive right now?
Because supply has collapsed while demand has not. Street prices run $7,488 to $10,000 — 3.75x to 5x the $1,999 MSRP — with Amazon listings near $8,500 and Newegg listings up to roughly $9,600–$10,000. The leading reported explanation is rising costs: ZDNet Korea reports Nvidia is raising GDDR memory kit prices to board partners, which increases the cost of every card built. Allocation pressure toward higher-margin customers and resale-market dynamics amplify the effect.
Is the RTX 5090 discontinued?
No. Nvidia says the RTX 5090 is not discontinued — it remains an active product. This is a supply problem, not an end-of-life. That distinction matters because it means supply can recover and prices can fall, which is the core reason not to pay scalper prices today.
What does Micro Center's "Not Carried" tag mean?
It means the retailer is not currently stocking the RTX 5090 at all — distinct from "Out of Stock" (we sell it, check back) or "Discontinued" (it is dead). For a chain famous for aggressive GPU stocking, stepping out of the 5090 market entirely is a strong signal that wholesale supply has dried up or that current wholesale economics make the product unworkable at retail.
Where is the cheapest place to buy an RTX 5090?
Per Tom's Hardware's US price tracker on October 16, 2026, the best available price anywhere in the US was $7,488 — and that is still 3.75 times MSRP. There is currently no retailer selling the card at or near its $1,999 list price with reliable stock. If you see a listing dramatically below the tracker floor, treat it with extreme skepticism.
Should I buy an RTX 5090 at $8,500?
No. At $8,500 you are paying more than four times MSRP for a card that is still in production. Buy an RTX 5080 or 5070 Ti-class card at a sane price instead, or wait. The money you save — several thousand dollars — will buy your next two GPU upgrades.
Will RTX 5090 prices ever return to $1,999?
Possibly, but not on any predictable timeline. Watch four signals: board partner memory-cost trends (per reporting like ZDNet Korea's), retailer listing statuses (especially Micro Center's "Not Carried" tag flipping back), the Tom's Hardware tracker floor moving down from $7,488, and Nvidia's official communications. Sustained improvement across those signals means relief is coming.
What is the best alternative to the RTX 5090 right now?
The RTX 5080: same generation, same feature set, one tier down, at a far more rational price. For 1440p gamers, the RTX 5070 Ti class is the value sweet spot. And for anyone whose current card still does the job, waiting costs nothing and has a guaranteed positive return.
Could a new GPU generation make the 5090 irrelevant?
Eventually, yes — every flagship is eventually superseded. That is part of the argument against paying 5x MSRP today: you would be paying a luxury premium for a card whose successor is always, inevitably, coming. Factor the generational clock into your waiting strategy.
How We Verified This Article
- Street prices of $7,488–$10,000 (3.75x–5x MSRP): compiled from Tom's Hardware's US price tracker (Oct 16, 2026 — best available price $7,488), Amazon listings near $8,500, and Newegg listings up to roughly $9,600–$10,000, as reported by tech-insider.org (Oct 18, 2026).
- $1,999 MSRP: Nvidia's official launch price for the GeForce RTX 5090, widely reported at launch and referenced across all cited outlets.
- Micro Center "Not Carried" listing status: observed on Micro Center's RTX 5090 product listings, as reported by tech-insider.org (Oct 18, 2026).
- Nvidia says the card is not discontinued: Nvidia's stated position, as reported by tech-insider.org (Oct 18, 2026).
- Nvidia raising GDDR memory kit prices to board partners: reported by ZDNet Korea, via tech-insider.org (Oct 18, 2026).
- Shortage theories (allocation pressure, resale dynamics): presented as analysis and inference grounded in the verified facts above — not as reported facts. Where the article reasons from market structure, it says so explicitly.
- Alternative-card recommendations: editorial judgment based on product positioning within Nvidia's current stack; no performance or pricing claims beyond the verified 5090 figures are asserted as fact.
Sources
The Bottom Line
The RTX 5090 did not get more expensive because it got better. It got more expensive because it got rarer — a $1,999 flagship now changing hands at $7,488 to $10,000, with America's most gamer-friendly retailer declining to carry it at all, while Nvidia insists the card is not discontinued and board partners quietly pay more for the memory inside every unit. That is the mystery, and the reported facts point to an answer with no villain and no quick fix: rising component costs, capacity allocated to higher bidders, and a resale market spinning freely without the anchor of legitimate retail.
But a mystery is only interesting until it costs you money. The practical conclusion of this entire investigation fits in one sentence: do not buy an RTX 5090 in October 2026. Buy the RTX 5080 or a 5070 Ti-class card at a fair price, or simply wait with the card you have, monitor the four signals we laid out, and let the market come back to you. The fastest graphics card ever made is not worth five times its price. Nothing is — except, perhaps, the patience to wait until it is.
Comments
Post a Comment