Key takeaways
- Then: a consumer-driven demand spike, with a clear trigger for the end.
- Now: an industrial demand shift, where consumers compete with data centres for the same components.
- Then: prices fell sharply once demand disappeared.
- Now: prices respond to component costs and allocation decisions, which move slowly.
Probably not before 2027, and for the RTX 5090 possibly not at all in its current form. As of October 2026, NVIDIA says the GeForce RTX 5090 has not been discontinued but describes supply as constrained, retailers have delisted many listings, and the memory shortage behind the squeeze is forecast by suppliers to last well beyond this year. The RTX 5080 is tighter than at launch but far less scarce than the 5090. If you need a high-end card now, the practical advice is to avoid paying large premiums, look at the full RTX 50 and Radeon stack, and decide what you can wait for.
Our top picks
As an Amazon Associate we earn from qualifying purchases at no extra cost to you.
Where Supply Stands in October 2026
The picture differs by card, and mixing them together is the main reason shortage coverage is confusing.
| Card | Chip | VRAM | Supply situation (October 2026) |
|---|---|---|---|
| GeForce RTX 5090 | GB202 | 32GB GDDR7 | Severe shortage; listings removed or marked not carried at several retailers; street prices far above list |
| GeForce RTX 5080 | GB203 | 16GB GDDR7 | Tight but generally findable; prices elevated above launch levels |
| GeForce RTX 5070 Ti | GB203 | 16GB GDDR7 | Reported as heavily constrained and sidelined from shelves in some regions |
| GeForce RTX 5070 and below | GB205, GB206, GB207 | 8GB to 12GB GDDR7 | Available, with prices drifting up |
| Radeon RX 9070 XT | Navi 48 | 16GB GDDR6 | Available; uses GDDR6, so less exposed to the GDDR7 squeeze |
This table reflects reporting from Tom’s Hardware, Digital Trends, TweakTown, VideoCardz and Tech Times, which agree on the direction even when they disagree on severity. Some of the more detailed claims come from industry leakers and smaller outlets, so treat precise percentages as indicative rather than official.
What NVIDIA Has Actually Said
In mid-October 2026, NVIDIA’s senior vice president of personal computing, Jeff Fisher, told Taiwanese media following an event in Taipei that the RTX 5090 has not been discontinued and that NVIDIA will continue to ship it. According to the translated reports, he described the supply chain as somewhat constrained and said the company is doing everything it can to get product to market. He also noted that the 5090 is not going end-of-life even though other products built on the same GB202 chip, such as the professional RTX Pro 6000, are in production.
The denial matters, but it has limits. It addresses end-of-life status, not allocation. Rumours from leakers say NVIDIA has been directing GB202 dies to higher-margin professional products, which would leave add-in-board partners with little to build consumer cards from. NVIDIA has not confirmed or refuted that specific claim, and reports describe partners as reaching the end of their last allocations. The company has also not published a plan or a date for improved availability.
Why the Shortage Exists
1. AI demand for memory and chips
The dominant cause named in industry coverage is the demand from AI data centres. IDC has estimated that data centres will absorb roughly 70 percent of memory output in 2026, and SK hynix has said its DRAM, NAND and high-bandwidth memory are sold out for the remainder of the year. Memory makers prioritise the highest-margin products, which are datacentre parts, and consumer GDDR7 competes for the same wafer capacity and engineering effort.
2. GDDR7 supply constraints
The RTX 50 series was the first consumer generation to use GDDR7. Reports in 2026 say NVIDIA cut RTX 50 supply to partners by roughly 15 to 20 percent because of GDDR7 constraints, with some outlets claiming larger production cuts of 30 to 40 percent in the first half of the year. A September 2026 report states that Micron marked its 2GB GDDR7 parts as end of life; most RTX 50 boards use 2GB modules, so the remaining supply comes from Samsung and SK hynix. Micron’s 3GB parts remain active, but boards would need redesign to use them in the same quantities.
3. Die allocation
One analysis argues that the RTX 5090 problem is more about GB202 allocation than memory. The same silicon sells in professional cards at many times the margin of a GeForce, which gives NVIDIA a strong financial reason to favour those products when wafers are limited. Whether the cause is memory, silicon or both, the result for buyers is the same: fewer cards.
4. Pricing and retail behaviour
When a product becomes scarce, retailers and scalpers raise prices, and manufacturers pass on higher component costs. Gartner has estimated that the memory shortage will push up PC prices across the board in 2026, affecting complete systems as well as individual components.
Will Supply Catch Up? Scenarios and Evidence
| Scenario | What would have to happen | Evidence in 2026 | Likelihood |
|---|---|---|---|
| Recovery by end of 2026 | GDDR7 and GB202 supply normalise within months | Earlier TrendForce baseline expected normalisation by Q4 2026, but later statements from memory makers contradict it | Low |
| Gradual improvement through 2027 | 3GB GDDR7 ramps, boards redesigned, datacentre demand stabilises | Reports link the RTX 50 Super timing to 3GB GDDR7 supply | Moderate |
| Tight supply into 2028 | AI memory demand keeps outpacing new fabs | Micron’s chief executive has said memory will stay tight through 2028; new fabs take four to five years | Moderate to high for memory |
| 5090 replaced rather than restocked | NVIDIA shifts the GeForce flagship role to another card | Rumours of a 24GB RTX 5080 refresh, unconfirmed | Possible |
The most credible signal is the memory industry itself. Memory executives have stated that capacity is sold out and that new fabrication capacity takes years to build. A comment attributed to SK Group’s chairman in June 2026 said the shortfall would stay well above demand for years. If those forecasts hold, consumer GPU supply will improve only as memory allocation to graphics cards improves, and that is a business decision as much as a manufacturing one.
What about the RTX 50 Super cards?
Early 2026 rumours suggested a Super refresh with 3GB memory modules, which would have raised capacities, for example a 24GB RTX 5080. Later reports say the refresh has slipped toward early 2027, with some reporting a 24GB RTX 5080 planned for the first quarter of 2027. TechPowerUp reported that board partners had not yet received development materials, which makes that window look optimistic. None of this has been confirmed by NVIDIA, so do not plan a purchase around unannounced hardware.
What This Means for Buyers
If you want an RTX 5090
Expect a long search and a large premium over list. Avoid listings from sellers without return policies, and be wary of cards sold as new but missing original packaging. Prebuilt desktops that include a 5090 can be a legitimate route, though system prices have also risen, and you should verify the exact GPU model in the specification before ordering. If you can live without 32GB of memory, an RTX 5080 or a used RTX 4090 may be better value.
If you want an RTX 5080
The card is available but not discounted. Waiting rarely helps in a shortage, but paying well above the standard range does not either. Set a price ceiling based on the performance gap to the next card down, and walk away when listings exceed it.
If you are flexible
The most valuable step is to match the card to your actual needs. At 1440p, an RTX 5070 Ti, RTX 5070 or Radeon RX 9070 XT covers nearly all games. At 4K, a Radeon RX 9070 XT with upscaling is a capable alternative, and the RTX 5080 remains the realistic NVIDIA option.
Alternatives to the Scarce Cards
| Option | Key specs | Best for | Trade-off |
|---|---|---|---|
| GeForce RTX 5080 | 10,752 CUDA cores, 16GB GDDR7, 360W | High-refresh 4K and 1440p ultra | 16GB limits the largest workloads |
| GeForce RTX 5070 Ti | 8,960 CUDA cores, 16GB GDDR7, 300W | 1440p and entry 4K | Reported as constrained in some regions |
| Radeon RX 9070 XT | 16GB GDDR6, around 304W | Rasterised performance per watt, 4K with FSR | Ray tracing and software features trail NVIDIA |
| Used GeForce RTX 4090 | 16,384 CUDA cores, 24GB GDDR6X, 450W | 4K, creation and AI with a large memory pool | No warranty in many cases, high power draw |
| GeForce RTX 5070 | 6,144 CUDA cores, 12GB GDDR7, 250W | 1440p at high settings | 12GB can be limiting at 4K |
| Prebuilt system with a high-end GPU | Varies by model | Buyers who need a complete PC | Check component quality, power supply and upgradeability |
Worked Example: Is Waiting Worth It?
Suppose you gain value from a faster GPU every month you play or work with it. If you estimate that a card would save you ten hours of render time per month or give you meaningfully better gameplay, waiting six months to save a premium means giving up six months of use. A useful rule is to divide the premium over the typical list price by the number of months you expect to wait. If the premium, spread over six months, is more than the value you place on a month of use, waiting wins. If it is less, buy now.
For a creator who earns income from the GPU, the premium is usually justified quickly. For a gamer with a working older card, waiting costs little, and prices may fall if demand weakens or a refresh arrives. The key point is that rising prices are not a signal that prices will stop rising; the supply forecasts suggest prolonged tightness, not a short spike.
Lessons From Earlier GPU Shortages
GPU supply crunches are not new, and comparing them clarifies what is different this time. The 2020 to 2022 shortage was driven by pandemic-era demand, cryptocurrency mining and constrained manufacturing and packaging capacity. It ended when mining collapsed after Ethereum moved to proof of stake, demand fell and inventories built up, leaving retailers with surplus cards that sold at discounts. The lesson from that period was that shortages can end abruptly when the demand source disappears.
The 2026 situation has a different structure. The demand source is AI infrastructure spending, which is funded by some of the largest companies in the world and shows no sign of collapsing in the near term. The scarce input is memory rather than the graphics processor alone, and memory capacity cannot be expanded quickly. Building and qualifying a new fab takes years, and suppliers have publicly said they are sold out. A sudden glut of consumer cards is therefore less likely than it was in 2022.
- Then: a consumer-driven demand spike, with a clear trigger for the end.
- Now: an industrial demand shift, where consumers compete with data centres for the same components.
- Then: prices fell sharply once demand disappeared.
- Now: prices respond to component costs and allocation decisions, which move slowly.
This does not mean prices can never fall. Competition from AMD and Intel, weaker consumer demand at very high prices, and new memory capacity coming online could all ease the pinch. It does mean that a buyer who expects a repeat of the 2022 collapse may wait a long time.
What to Watch for Signs of Recovery
- Retailer stock changes. A move from not carried back to in stock for the RTX 5090 would signal improved allocation.
- Official announcements. Details of an RTX 50 Super refresh or a 24GB RTX 5080 would show how NVIDIA plans to use scarce memory.
- Memory-maker earnings commentary. Statements from Micron, SK hynix and Samsung about GDDR7 and 3GB module output are the best leading indicator.
- Street price trends. Falling gaps between list and street prices suggest easing; widening gaps suggest the opposite.
- Board partner behaviour. New models, refreshed coolers or limited-edition releases usually appear only when partners are confident of supply.
Safe Buying Practices During a Shortage
- Buy from authorised retailers or the manufacturer. Marketplace sellers and auction listings carry higher scam and counterfeit risk.
- Check the exact model. Fake listings may show a 5090 in photos with a lower-tier card in the box, or an empty box.
- Use payment methods with buyer protection. Avoid bank transfers or gift cards to unknown sellers.
- Inspect on arrival. Record unboxing, check the serial number against the manufacturer’s registration tool and confirm the card identifies correctly in GPU-Z or the driver panel.
- Avoid cards with damaged power connectors. High-power 12V-2×6 connectors are sensitive to poor seating and bent pins.
- Check return windows. A short return window on a very expensive card is a risk.
- Use stock alerts and waitlists. Retailers and manufacturers release small batches without notice.
How the Shortage Affects Other Parts
The same memory forces influence system memory and storage. DDR5 modules and SSD prices have risen in 2026 as manufacturers divert capacity to AI products, so a budget set aside for a GPU may also need to cover costlier RAM. Builders upgrading from DDR4 to DDR5 should compare total platform cost rather than the GPU in isolation. Power supplies and cases are unaffected, which makes them the safest places to spend if you are buying parts in stages: a quality 1000W unit with a native 12V-2×6 cable will serve whichever flagship you eventually purchase.
Frequently Asked Questions
Has NVIDIA discontinued the RTX 5090?
NVIDIA says no. A senior executive stated in October 2026 that the card has not been discontinued and will continue to ship, while acknowledging that supply is constrained. Retail availability remains very limited, which is why the rumours persist.
Why is the RTX 5090 so hard to find?
Reports point to a combination of limited GB202 chip allocation, GDDR7 memory constraints and very high demand. The same chip is used in professional cards, and memory makers are prioritising AI datacentre products, leaving less for consumer graphics.
When will RTX 5080 and RTX 5090 supply improve?
No official date exists. Memory suppliers have said capacity is sold out through 2026 and tight into 2027 or later, and a RTX 50 Super refresh is rumoured for early 2027. A meaningful recovery in 2026 looks unlikely.
Is the RTX 5080 also in shortage?
The RTX 5080 is tighter than at launch and sells above its original price in many places, but it has not been delisted widely like the 5090. Availability varies by region and retailer.
Should I buy a used RTX 4090 instead?
It can be a good option for 4K gaming, creative work or AI, because it offers 24GB of memory and strong performance. Buy only with return protection, inspect the power connector and cooler, and remember that it draws 450W and may have no remaining warranty.
Will the RTX 50 Super cards fix the shortage?
Not by themselves. Rumoured Super models depend on the same constrained memory supply, particularly 3GB GDDR7 modules, and they remain unannounced. A refresh could add capacity, but it will not necessarily raise the number of cards available.
Bottom Line
Supply is unlikely to catch up with demand for the RTX 5090 or RTX 5080 in the near term, because the cause is structural: memory and chip capacity are being pulled toward AI products, and suppliers expect tightness for years. Buy a flagship only if its performance has clear value for your work, avoid unsafe sellers and large premiums, and consider the RTX 5070 Ti, Radeon RX 9070 XT or a carefully inspected used RTX 4090 as realistic alternatives.
Write Your Review
No reviews yet. Be the first to share your experience!