European GPU buyers just got hard numbers confirming what their wallets already knew: graphics cards cost dramatically more than they did over the summer. A European GPU-market tracker measured the average Nvidia graphics card price climbing from roughly €783 on June 28 to about €1,107 by September 28, a jump of 41.3% in three months. AMD’s Radeon lineup rose more slowly but still painfully, from around €644 to €826, a 28.3% increase over the same window. The pricing data was reported on September 29, 2026 and corroborated independently by PCGuide, which attributed the surge to the ongoing memory shortage rather than any change in chip design or manufacturing cost for the GPUs themselves.
The timing matters. These are not holiday-season markups or a one-off import tariff. They track almost exactly with a broader DRAM, NAND, and HBM supply crunch that chipmakers have been warning about since late 2025, one that has already reshaped PC and smartphone pricing (see our coverage of DRAM now costing more than TSMC’s 2nm silicon). What’s new this week is a clean, dated before-and-after snapshot showing exactly how much of that cost has already landed on gamers and PC builders in Europe, and why Nvidia buyers got hit harder than AMD’s.
The numbers: Nvidia up 41%, AMD up 28% in three months
The tracker in question measures average street prices across retailers in the eurozone, covering current-generation Nvidia GeForce RTX 50-series and AMD Radeon RX 9000-series cards. Between late June and late September 2026, the average Nvidia card rose from about €783 to about €1,107. AMD’s average moved from roughly €644 to €826, with most of that increase concentrated in September itself, going from around €786 to €820 in just that one month according to the republished figures.
Why the gap between the two vendors? The most plausible explanation, repeated across multiple outlets covering the memory shortage, is memory capacity per card. Nvidia’s flagship GeForce RTX 5090 carries 32GB of GDDR7, a memory standard that is newer, lower-volume, and more exposed to the same fabs now prioritizing high-bandwidth memory (HBM) for AI accelerators. AMD’s Radeon RX 9070 XT, by contrast, uses 16GB of the more mature GDDR6 standard. Less exotic memory means less exposure to the acute supply squeeze. That difference in bill-of-materials sensitivity lines up with the gap in price inflation between the two brands.
It is worth being precise about what is and is not confirmed here. The aggregate averages, 41.3% and 28.3%, are well-documented across two independent republications. Specific per-model price points in euros, card by card, are thinner in the public record. One tracker listing showed a minimum price of €5,299 against a €2,099 MSRP for an unspecified RTX 50/RX 9000 card, which looks like an outlier or a scarcity-driven scalper listing rather than a representative market price, and should be read with caution rather than treated as typical.
Why GPU prices are rising: the memory shortage, not the chips
Nothing about the silicon inside an RTX 5090 or a Radeon RX 9070 XT changed this year. What changed is the cost of the memory chips soldered around that silicon. GDDR6, GDDR7, and HBM all draw on the same upstream DRAM manufacturing capacity at Samsung, SK hynix, and Micron. When AI data center operators started placing enormous advance orders for HBM to feed Nvidia and AMD’s AI accelerators, those three manufacturers began shifting wafer, packaging, and testing capacity away from commodity DRAM and graphics memory and toward HBM, where margins and contract value run much higher.
Micron’s own earnings commentary shows how severe the shortfall has become. Micron CEO Sanjay Mehrotra told investors the company can currently meet only about 50% to two-thirds of what its largest customers are asking for, and said he expects the aggregate memory shortage to persist well beyond 2026, with meaningful new industry-wide supply unlikely to come online before 2028, according to Tom’s Hardware’s report on Micron’s earnings call. Mehrotra has pointed to the multi-year construction timeline for new fabs as the structural bottleneck: pouring the shell of a new fab, equipping it, and qualifying production lines takes years, not quarters, so pricing pressure alone cannot pull forward supply that physically does not exist yet.
CNBC separately reported Micron’s view that aggressive customer price negotiations earlier in the cycle contributed to the shortage, by discouraging the kind of capacity investment that would have been needed to meet today’s AI-driven demand, a dynamic laid out in CNBC’s coverage of Micron’s CEO remarks. PC Gamer’s reporting on the same earnings cycle noted Micron’s position that shortages will persist beyond 2026 even as the company frames more memory capacity as essential to deliver the AI experience customers are demanding, a framing detailed in PC Gamer’s writeup. None of that is good news for anyone trying to buy a graphics card in the next 12 to 18 months.
AMD’s price hikes: a slower, staged climb
AMD’s board partners have been more transparent about the mechanics of their price increases than Nvidia has. Tom’s Hardware reported that AMD notified partners of roughly a 10% increase on Radeon GPU-and-memory kits earlier in 2026, followed by confirmation that prices would rise by at least another 10% starting in August, a pattern of staged hikes rather than a single shock, as detailed in Tom’s Hardware’s reporting on AMD’s pricing notices to partners. A separate Tom’s Hardware report on rumored board-partner pricing floated specific dollar figures: roughly $20 increases on 8GB Radeon cards and $40 on 16GB models, tied directly to rising GDDR6 spot prices, covered in that outlet’s report on the GDDR6 rumor.
That staged, model-by-model approach is consistent with the European tracker’s finding that AMD’s average rose more gradually than Nvidia’s across the June-to-September window, with the steepest jump concentrated in September. AMD is still raising prices. It’s just raising them in smaller, more frequent increments tied to specific memory-capacity tiers, rather than the broader, steeper repricing Nvidia’s lineup appears to have absorbed. For buyers, the practical effect is similar either way: a card that cost roughly the same in June now costs meaningfully more by the time you check out in October.
How this compares to the broader PC and smartphone market
GPUs are not an isolated case. Gartner’s analysis of the memory market forecasts combined DRAM and SSD prices rising roughly 130% by the end of 2026 compared with 2025 levels, a surge the firm expects to push PC prices up about 17% and smartphone prices up about 13%, according to figures reported by Tech Times’ coverage of the Gartner forecast. That framing helps explain why the GPU-specific numbers from the European tracker, severe as they look in isolation, actually sit inside a much larger repricing of nearly every memory-dependent consumer electronics category.
Our earlier coverage of RAM now making up 60% of some phones’ bill of materials and the parallel story on the RTX 5090 vanishing from US retail shelves entirely both point to the same underlying mechanism driving the European GPU numbers: when memory becomes the scarce input, every device that depends on it gets repriced according to how much memory it carries, not according to how much the silicon itself costs to manufacture.
Table: European average GPU prices, June vs. September 2026
| Metric | June 28, 2026 | September 28, 2026 | Change |
|---|---|---|---|
| Nvidia GeForce average (EU) | €783 | €1,107 | +41.3% |
| AMD Radeon average (EU) | €644 | €826 | +28.3% |
| AMD average, September only | €786 (early Sept.) | €820 (late Sept.) | +4.3% |
| AMD board-partner kit price hike (reported) | Baseline | 10%+ since August | Staged increase |
| Rumored 8GB Radeon price hike | Baseline | +$20 (reported rumor) | Tied to GDDR6 spot cost |
| Rumored 16GB Radeon price hike | Baseline | +$40 (reported rumor) | Tied to GDDR6 spot cost |
Table: Memory shortage’s reach across device categories
| Category | Reported impact | Source |
|---|---|---|
| GPUs (Europe, Nvidia) | +41.3% in three months | European GPU tracker, republished Sept. 29, 2026 |
| GPUs (Europe, AMD) | +28.3% in three months | European GPU tracker, republished Sept. 29, 2026 |
| Combined DRAM + SSD pricing (global, 2026 vs. 2025) | ~130% forecast rise | Gartner, via Tech Times |
| PC prices (global, forecast) | ~17% rise | Gartner, via Tech Times |
| Smartphone prices (global, forecast) | ~13% rise | Gartner, via Tech Times |
| Micron customer order fulfillment rate | 50%-66% of demand met | Tom’s Hardware, Micron earnings call |
Historical context: how we got here
GPU pricing has gone through sharp shocks before, most memorably the cryptocurrency mining boom of 2017-2018 and again in 2021, when scarcity and scalping pushed GeForce and Radeon cards to double or triple their list prices. Both of those episodes were demand shocks hitting a relatively elastic supply of finished cards. What is happening in 2026 is structurally different. The bottleneck is not finished-card assembly or even GPU die supply from TSMC. It is the memory chips that go onto the card, and those chips are produced by a tiny number of manufacturers (Samsung, SK hynix, Micron) who are simultaneously fulfilling multi-year, enormous HBM contracts for AI accelerator makers like Nvidia and AMD’s own data center divisions.
That overlap is the twist that makes this shortage harder to fix than past ones. Nvidia and AMD are, in effect, competing against their own AI data-center product lines for the same upstream memory capacity that goes into consumer gaming GPUs. Our earlier reporting on Nvidia CEO Jensen Huang’s forecast locking up over a third of global HBM supply shows just how far in advance that capacity is already committed, leaving little room for memory suppliers to redirect output back toward consumer GDDR even if they wanted to.
Competitive landscape: Nvidia vs. AMD pricing strategy
Nvidia and AMD are navigating the same supply crunch with visibly different public postures. AMD has been relatively forthcoming with board partners about staged, percentage-based increases tied to specific memory tiers, which is consistent with the gentler slope seen in the European tracker’s AMD average. Nvidia’s pricing moves have been less transparently communicated to the public, and the tracker data shows a steeper, more front-loaded jump across the same three months.
Part of that gap likely comes down to product mix. Nvidia’s current GeForce RTX 50 lineup leans more heavily on GDDR7, a newer and lower-volume standard than GDDR6, which makes Nvidia’s bill of materials more sensitive to the same supply squeeze hitting HBM. AMD’s Radeon RX 9000 series still uses GDDR6 almost across the board, a memory type that, while also rising in price, has a larger installed manufacturing base and more alternative supply sources than GDDR7. That single technical choice may explain a meaningful share of the 13-point gap between Nvidia’s 41.3% increase and AMD’s 28.3%.
Desktop GPU shipment data adds another layer. Nvidia’s dominant share of the discrete GPU market, detailed in our coverage of desktop GPU shipments hitting 12.5 million units with Nvidia holding roughly 90% share, means Nvidia has far less competitive pressure to absorb memory-cost increases out of margin. AMD, fighting for the remaining share of the market, has more incentive to shield gamers from the full cost pass-through, even if it cannot avoid passing along some of it.
Supply chain mechanics: why board partners can’t just absorb the cost
It helps to understand who actually eats a memory-cost increase before it reaches a storefront. Nvidia and AMD design the GPU and sell the core chip or the full board to add-in-board (AIB) partners like ASUS, MSI, Gigabyte, Sapphire, and PowerColor. Those partners buy memory chips separately, often under shorter-term contracts than the GPU vendors’ own HBM deals, which means AIB partners feel spot-price spikes in GDDR6 and GDDR7 almost immediately, sometimes before Nvidia or AMD formally revise their own suggested retail prices.
That is part of why the rumored $20 and $40 price increases on Radeon cards reported by Tom’s Hardware were framed as board-partner decisions tied to GDDR6 spot pricing rather than a single AMD-wide price list update. Margins at the AIB level are thin to begin with, often in the single digits, so a sudden jump in memory input costs leaves partners with little room to absorb the hit internally. Retail price increases become close to automatic once a memory contract renews at a higher rate, and the European tracker data captures the end result of that pass-through process rather than any single company’s one-time decision.
What this means for PC builders right now
If you are planning a new build or an upgrade in the next few months, the practical takeaway from this data is straightforward: prices are not likely to retreat soon, and every month of delay has, on average, meant a higher price over the June-to-September window this tracker covers. Buying now locks in today’s price rather than a further markup. Waiting makes sense only if your current system is functional and you are not under time pressure, since Micron’s own guidance points to continued tightness well into 2027 and possibly beyond.
For anyone shopping specifically for memory capacity rather than raw GPU performance, the data suggests AMD’s GDDR6-based cards currently offer a somewhat more stable price trajectory than Nvidia’s GDDR7-based lineup, simply because GDDR6 is less exposed to the acute end of the shortage. That is not an argument about raw performance, only about near-term price stability.
Predictions: where GPU pricing goes from here
- Expect further incremental AMD price increases through early 2027, following the same staged, percentage-based pattern seen since mid-2026, rather than one large single hike.
- Nvidia’s GDDR7-heavy lineup will likely keep seeing steeper average price inflation than AMD’s GDDR6-based cards for as long as GDDR7 production stays concentrated among fewer suppliers.
- Entry-level and budget GPU tiers will likely see the sharpest percentage increases in 2027, mirroring Gartner’s prediction that low-end PCs become economically unviable as memory eats a growing share of bill-of-materials costs.
- Relief is unlikely before new DRAM and HBM fab capacity comes online, which Micron’s own leadership has pointed to as a 2028 event at the earliest, meaning 2026-2027 should be treated as a sustained high-price period rather than a temporary spike.
- Expect more granular, dated price-tracking reports like this one to keep surfacing as shoppers and retailers both seek hard numbers to plan around, rather than relying on anecdotal reports of individual price hikes.
Market impact beyond gaming PCs
The ripple effects reach well past gamers. Cloud GPU rental pricing has already moved in response to the same memory constraints, as covered in our report on Nebius raising AI cloud prices 21% while memory costs jumped 41%, showing that enterprise buyers of GPU compute face the identical upstream pressure as consumer shoppers, just denominated in cloud-hour pricing rather than retail sticker prices. Component resellers, system integrators, and prebuilt PC makers are all absorbing the same cost increases, and several have already begun passing them through in list prices for pre-assembled gaming desktops.
Retailers across Europe have reportedly started adjusting restock pricing more frequently than in previous years, a shift away from the relatively stable MSRP-anchored pricing that characterized most of the 2022-2024 period after the prior mining-driven shortage eased. That dynamic pricing behavior, where shelf prices change week to week rather than quarter to quarter, is itself a symptom of how volatile upstream memory costs have become.
System integrators building prebuilt gaming desktops face a harder problem than individual component buyers: they typically quote prices weeks ahead of actual assembly and shipping, which means a memory-driven price swing mid-quarter can turn a profitable order into a loss if the integrator doesn’t hedge with shorter quote windows or price-adjustment clauses. Several smaller system builders have reportedly shortened the validity window on their published prebuilt configurations as a direct response to this volatility, a practical, if unglamorous, sign of how deep the pricing instability has cut into normal retail operations.
Frequently asked questions
Why did Nvidia GPU prices rise more than AMD’s in Europe?
The most likely explanation is memory type. Nvidia’s current RTX 50-series lineup relies heavily on GDDR7, a newer and lower-volume memory standard than the GDDR6 used in most of AMD’s Radeon RX 9000 series. GDDR7 is more exposed to the same fab capacity being prioritized for HBM in AI accelerators, which has pushed its cost up faster.
Is the GPU price increase only happening in Europe?
No. The European tracker data gives the clearest dated before-and-after snapshot, but reporting from Tom’s Hardware and others shows AMD and Nvidia both raising board-partner pricing in the US and globally through 2026, driven by the same global DRAM and GDDR shortage.
When will GPU prices come back down?
Based on Micron’s own earnings commentary, meaningful new memory supply is unlikely before 2028, since new fab capacity takes years to build, equip, and qualify. Analysts broadly expect the shortage to persist through at least 2027.
Should I buy a GPU now or wait for prices to drop?
Nothing in current reporting points to a near-term price drop. If you need a GPU for work or a build that’s otherwise ready, buying now avoids locking in a further markup. If your current setup still works, waiting carries real risk of paying even more later rather than less.
What is driving the memory shortage behind these price increases?
AI data center buildouts have created enormous demand for HBM, which is produced using the same fabs, wafer capacity, and packaging lines as conventional DRAM and graphics memory like GDDR6 and GDDR7. Manufacturers are prioritizing HBM contracts, which squeezes supply for everything else, including the memory that goes into consumer GPUs.
Are AMD and Nvidia raising prices because manufacturing the GPU chips got more expensive?
No. The chips themselves, the GPU dies, have not seen a comparable cost increase. The price pressure comes almost entirely from the memory chips surrounding the GPU die, not the processor itself.
Does this affect laptop GPUs too?
Yes, indirectly. Laptop pricing overall is expected to rise as memory costs climb, per Gartner’s forecast of roughly 17% higher PC prices in 2026, and laptop GPUs draw on the same constrained memory supply chain as desktop cards.
Related
- DRAM Now Costs 54% More Than TSMC 2nm Silicon
- RTX 5090 Vanishes From US Stores, Hits $9,500
- RAM Now 60% of Phone Cost, Sparks Fake-Chip Checks
- Huang’s 2x Chip Vow Locks Up 37% of HBM Supply
- Desktop GPU Shipments Hit 12.5M, Nvidia Takes 90%
