GPU prices rose sharply in 2026, and the memory shortage behind them is still getting worse. At the end of 2025, many graphics cards were selling at or even below MSRP. By September 2026, the videocardprices.com GPU Price Index showed the RTX 5090 selling for roughly three times its launch price, with midrange cards like the RTX 5070 and Radeon RX 9070 XT running 25 to 30 percent above MSRP.
The cause is the same one we warned about last year. Memory manufacturers are redirecting production toward higher-margin AI and data center clients, leaving fewer GDDR6 and GDDR7 chips for consumer GPUs. That squeeze has delayed new graphics card launches, pushed Nvidia to bring back an older GPU, and raised prices on consoles and handhelds too. Here's what happened, and what it means if you're planning an upgrade.
What's happening to the GPU market right now For most of 2025, the GPU market appeared to normalize. Prices stabilized, cards were widely available, and some models dipped below MSRP for the first time since before the pandemic. Underneath, the supply chain was already tightening, and in 2026 that pressure reached store shelves.
Here's how current-generation cards compare to their launch prices as of late September 2026:
GPU
Launch MSRP (USD)
Street price (USD)
Difference
RTX 5090
$1,999
$6,099
+205%
RTX 5080
$999
$1,399.99
+40%
RTX 5070
$549
$714.99
+30%
Radeon RX 9070 XT
$599
$749.99
+25%
Arc B580
$249
$299.99
+21%
RTX 5060 Ti
$379
$425.97
+12%
These figures come from the videocardprices.com GPU Price Index as of September 29, 2026, and prices change daily.
The flagship RTX 5090 took the biggest hit. On September 1, 2026, the cheapest new RTX 5090 in the US was $4,999, compared to $2,250 one year earlier , according to Tom's Hardware tracking. Older generations are the exception: cards like the RTX 3070 and RX 6700 XT now sell for roughly half their original launch prices, which makes last-gen hardware worth a look for budget builds.
Note for Jake: verify final numbers against Tom's Hardware's GPU price index before publishing.
RAM prices exploded, and GPUs depend on the same supply The first warning sign appeared in the RAM market. Throughout 2025, DDR5 prices surged at a pace that caught nearly everyone off guard. Kits that cost around $90 in late summer suddenly climbed to $240 or more , and some configurations saw year-over-year increases of more than 170 percent. Those increases kept climbing into 2026.
This spike isn't limited to desktop memory. GPUs rely on specialized versions like GDDR6 and GDDR7 , which come from the same handful of manufacturers. When DRAM prices spike, VRAM costs rise alongside them. Memory fabs are prioritizing their most profitable customers, AI and data center clients, which leaves far fewer chips for consumer graphics cards. GPU makers must either absorb the cost or pass it on, and in 2026 most passed it on.
Higher-density memory has been hit especially hard. Nvidia reportedly held back its RTX 50 Super cards because 3GB GDDR7 modules cost around three times more than the 2GB chips used in current cards.
AI data centers are consuming global memory production The single biggest force behind the memory crunch is AI. Hyperscale companies such as Nvidia, Microsoft, Amazon, Google, and Meta have dramatically increased their demand for high-performance memory. They are buying enormous quantities of:
HBM (High Bandwidth Memory) Server-grade DRAM GDDR6 and GDDR7 Complete production runs from major fabs AI accelerators require far more memory per device than consumer GPUs, and they generate much higher profit margins. So memory manufacturers keep redirecting capacity away from gaming hardware and toward data centers.
That means consumer GPUs are no longer the priority. When memory is limited, enterprise buyers get first access, and gaming cards get whatever remains. It resembles the crypto-mining era, except this demand is larger, more sustained, and backed by corporate budgets that can outbid the consumer market every time.
Nvidia cut gaming GPU supply In late 2025, reports suggested Nvidia would reduce discrete gaming GPU production by as much as 30 to 40 percent in the first half of 2026 because of VRAM constraints. Later reporting put the reduction lower, with leaker MegasizeGPU claiming Nvidia cut chip supply to board partners by 15 to 20 percent and planned no new Blackwell gaming GPUs for the year.
Even a smaller cut matters. When fewer graphics cards are produced while demand stays steady, price competition weakens, discounts disappear, and average selling prices rise. The cuts also aren't spread evenly. Higher-VRAM configurations consume more scarce memory per unit, which made popular 16GB cards especially exposed.
What about VRAM bundling? Another rumor last year claimed that Nvidia had stopped supplying VRAM alongside its GPUs , forcing board partners to source memory on the open market. That worst case doesn't appear to have happened, since the same 2026 leak reported that Nvidia is still supplying the necessary VRAM with its chips. That's good news for smaller board partners, who would have struggled to outbid AI buyers for memory on their own.
GPU launches were delayed by memory shortages Nvidia was originally expected to refresh its lineup with RTX 50 Super models in early 2026. By late 2025, reports said those launches might slip toward late 2026 . The delay has since stretched further. In early 2026, The Information reported that Nvidia had no plans to launch new GPUs that year and had delayed its next-generation RTX 60 "Rubin" series .
By summer, the RTX 50 Super cards had reportedly reached board partners, but Nvidia put the launch on hold over the cost of 3GB GDDR7 memory . The rumored lineup includes the RTX 5080 Super and RTX 5070 Ti Super with 24GB of VRAM and the RTX 5070 Super with 18GB. Launch timing has reportedly moved from late 2026 to potentially early 2027, and Nvidia has not officially announced any Super models.
The RTX 3060 comes back One of the strangest effects of the shortage is the return of the RTX 3060 12GB, a card first launched in 2021 and phased out in 2024. While waiting for memory costs to improve, Nvidia has reportedly resumed RTX 3060 12GB shipments. Because the card uses older GDDR6 memory and Samsung's 8nm process, Nvidia can increase RTX 3060 supply without taking capacity from its newer cards. Its 12GB of VRAM is more than the 8GB found on entry-level RTX 4060 and RTX 5060 cards, though its overall performance is lower.
AMD faces the same squeeze. Its RDNA 4 lineup relies on GDDR6, and those inventories are tight as well.
Consoles and handhelds got more expensive too The memory shortage hit every device that relies on DRAM or VRAM, including consoles and handheld PCs. Xbox had already seen price increases and unstable pricing by late 2025, and 2026 brought more.
Memory makes up a large share of console pricing. TrendForce estimates that memory accounts for about 35 percent of the price of PS5 and Xbox systems, so console prices move with the memory market.
Will GPU prices go down? Not soon, based on current forecasts. Jefferies Equity Research expects memory prices to rise another 40 to 50 percent in Q3 2026 and 30 to 40 percent in Q4 2026, followed by a further 40 to 45 percent increase in 2027. As long as AI demand keeps absorbing memory supply, GPU makers will keep paying more to build each card.
A few factors could still bring relief. New memory fab capacity, a slowdown in AI spending, or a shift back toward consumer memory production would all ease the pressure, but none of those is expected to happen quickly. For now, the most practical way to save is to look beyond standalone flagship cards: last-generation GPUs, entry-level models, and complete gaming systems can all offer better value than chasing an RTX 5090 at today's prices.
Final thoughts At the end of 2025, it looked like the GPU market had finally returned to normal. That window closed. The memory shortage driven by AI expansion raised graphics card prices across the board in 2026, delayed new launches, and pushed console and handheld prices up with them. With memory prices forecast to keep rising, relief isn't likely in the next few months.
If you need an upgrade now, a complete system can be a smarter buy than a standalone card at today's inflated prices. Acer's Predator and Nitro gaming laptops and desktops come with powerful GPUs built in, and many models are discounted during seasonal sales. Students can also take advantage of Acer's education program, which offers up to 15 percent off select devices from the Acer Store . For help choosing the right specs, see our guide to the best laptop specs for college in 2026 .
GPU prices FAQ Why are GPU prices so high in 2026? GPU prices are high mainly because of a global memory shortage. AI data centers are buying up DRAM and VRAM, so graphics card makers pay more for GDDR6 and GDDR7 memory, and Nvidia has reduced gaming GPU supply.
How much has the RTX 5090 gone up in price? The RTX 5090 launched at $1,999. By late September 2026, its street price was around $6,000, roughly three times MSRP.
Will GPU prices go down in 2027? Probably not in the near term. Analyst forecasts call for memory prices to keep rising through the end of 2026 and into 2027, which would keep GPU prices elevated.
When will the RTX 50 Super launch? Nvidia has not announced RTX 50 Super cards. Reports say the cards are ready but on hold because of high GDDR7 memory costs, with launch timing possibly moving into early 2027.
Are older GPUs a good buy right now? They can be. Many last-generation cards sell well below their original launch prices, and Nvidia has reportedly brought back the RTX 3060 12GB to help fill supply gaps.
Recommended Products
Predator Helios 18 AI (RTX 5090) Buy Now
Acer Nitro 60 (RTX 5070Ti) Buy Now
Predator BiFrost AMD Radeon RX7600 OC Graphics Card Buy Now
