Nvidia is reportedly raising graphics card prices for the third time this year, this time by 20% to 30%, according to Taiwan's Economic Daily News. The detail that makes this one hurt more than the previous two is scope. Where earlier increases largely spared everything below the flagship RTX 5090, this hike reportedly hits the entire lineup, from that $2,000-plus halo card all the way down to budget-tier options, and it covers both the new GDDR7 memory in current cards and the older GDDR6 in previous-generation and entry-level models. Nvidia has notified board partners like ASUS, Gigabyte, and MSI, though it hasn't made any public announcement, so treat the specifics as a supply-chain report rather than confirmed fact for now. The corroboration across outlets, however, is extensive enough that few in the industry doubt it's happening.
The bitter irony sitting underneath this is impossible to ignore, and it's the part worth dwelling on. The reason GPU prices keep climbing is the memory crisis, and the memory crisis is being driven, more than by anything else, by the insatiable demand for DRAM from AI data centers. And who is the single largest beneficiary of that AI build-out? Nvidia, whose H100 and B200 server chips are the engines of the entire boom. So Nvidia is profiting enormously from the AI gold rush that is consuming the world's memory supply, and then turning around and charging its gaming customers more because that same gold rush has made memory scarce and expensive. The company sits on both sides of the equation, minting record profits selling the picks and shovels while passing the resulting cost inflation onto the consumer division that built its brand. When CEO Jensen Huang dismisses AI "doom talk" as "complete nonsense," it lands differently once you notice his own gaming customers are footing part of the bill for the boom he's cheering on.
The Numbers Behind the Squeeze
The upstream cause is straightforward once you follow the chain. Samsung reportedly raised its DRAM prices by around 20% for the third quarter of 2026, squeezing the margins of everyone who buys memory, Nvidia included. Because Nvidia sells its board partners complete "kits" bundling the GPU silicon with its memory, that cost increase flows directly into what ASUS, MSI, and Gigabyte pay, and from there onto retail shelves once existing stock runs dry. The scale of AI's appetite is the root of it all, with analyst estimates from IDC and TrendForce suggesting AI data centers now consume roughly 70% of global high-end DRAM production. The high-bandwidth memory packed into AI server chips requires around four times the wafer area of standard DRAM, physically crowding out the GDDR supply that gaming cards depend on, and IDC projects DRAM supply growing just 16% year-on-year against a historical norm of 20-30%. There simply isn't enough to go around, and gaming is at the back of the queue.
One genuinely revealing distortion illustrates how broken things have become. A standard 2GB GDDR7 module currently costs around $20 per chip. A 3GB module, offering 50% more capacity, costs between $60 and $70, roughly three times the price for half again the memory. That pricing imbalance makes it brutally difficult for manufacturers to build higher-capacity mid-range cards at palatable prices, and it's a big reason Nvidia's planned RTX 50 SUPER refresh, which was set to bump several cards to larger memory configurations, is reportedly on hold at board partners while the pricing gets sorted out. The memory has become so expensive that it's actively stalling the products meant to offer better value.
The Damage Is Already Visible at Retail
You don't have to imagine the impact, because the existing price trajectory already tells the story. The RTX 5090 launched at an MSRP of around $1,999 and now sells for anywhere between $3,599 and $4,500 depending on region, roughly a 90% premium over its launch price before this latest hike even lands. Even the nominally budget-friendly RTX 5060 Ti has crept from a $379 MSRP up past $569 on partner models. Stack another 20-30% on top of figures like those and the notion of an affordable modern graphics card starts to feel genuinely endangered. Over the last 18 months, GPU prices have risen by hundreds and in some cases thousands of dollars, and each hike has confirmed the previous one wasn't the ceiling.
No Bottom in Sight, and Nowhere to Run
This is the third increase of 2026 alone, following a 10-15% hike in January and a narrower May adjustment, and it validates every grim forecast we've tracked. Valve's own engineers recently warned the crisis is "still getting worse," with retail lagging bulk supply by three to six months, meaning worse pricing is already baked in and heading for shelves. Micron and other suppliers have signaled no meaningful relief until at least 2028. The Economic Daily News expects the repeated hikes to have a genuine chilling effect, predicting a 10-20% decline in overall PC hardware sales as buyers are simply priced out. And crucially, there's no escape hatch by brand-hopping, since AMD and Intel have both raised their GPU and memory prices this year too, drawing from the same constrained supplier pool. The pain has spilled well beyond PCs as well, with the PlayStation 5, Steam Deck, and Nintendo Switch 2 all seeing price increases driven by the same shortage.
More:Xbox's Grand Growth Plan Rests on the Staff It Just Cut
Read
Source Editing
Visual Editing
Related news
