Nvidia Raises RTX GPU Prices Up to 30% in Third Hike of 2026

Photo by Mariia Shalabaieva on Unsplash

Nvidia has raised the price of GeForce RTX GPU bundles sold to board partners by 20% to 30%, the third Nvidia GPU price increase 2026 has seen and the first to hit the entire product stack rather than just the flagship. Board partners ASUS, Gigabyte, and MSI are now paying close to a third more for the GPU-and-memory kits they turn into finished cards. Retail hasn't caught up yet on every model, but the ceiling is already visible in the UK and China.

The January round lifted MSRPs by 10% to 15%. The May round hit only the RTX 5090. This July round covers Blackwell GDDR7 cards and the older GDDR6 stack together, meaning budget buyers who sat out the first two increases are now inside the blast radius.

The pattern matters. Two isolated adjustments in a single year is unusual. Three, with the third covering the whole stack, signals the pricing pressure is structural rather than a one-off supplier squabble.

What the Third Nvidia Graphics Card Price Increase Actually Covers

The increase applies to what Nvidia calls the GPU package, the bundle that ships to add-in board partners containing both the graphics silicon and the GDDR memory soldered around it. Basic Tutorials reports the pass-through on those kits is close to 30% at the top end, with ASUS, Gigabyte, and MSI all absorbing the initial hit before deciding what to move to retail.

The scope is what separates this round from the previous two. Eteknix confirmed that January's increase raised prices roughly 10% to 15% across the lineup, and that May's follow-up only touched the GeForce RTX 5090. This July round pulls in the whole GeForce RTX Blackwell price sheet and the remaining GDDR6 SKUs at the same time.

Nvidia has issued no official statement. Gagadget notes the company notified partners privately, which is why every figure in circulation traces back to Taiwan's supply chain reporting rather than a press release. That distinction matters for anyone trying to time a purchase against a confirmed MSRP change.

It also matters for accountability. When a manufacturer raises prices via press release, retailers and consumers get a fixed reference point. When the same manufacturer raises them via partner notification, retail pricing drifts unevenly across regions, and the only signal buyers get is a shelf tag that quietly climbs week by week.

The Samsung DRAM Price Increase Behind the GPU Hike

Samsung raised DRAM contract prices by roughly 20% for Q3 2026, and Basic Tutorials reports the increase covers both GDDR6 used on older cards and GDDR7 used across the Blackwell generation. That single supplier decision is the mechanism behind the Samsung DRAM price increase GPU buyers are now paying for at the till.

The non-linear pricing of GDDR7 is where this gets ugly. A standard 2 GB GDDR7 module now trades around $20, while a 3 GB module costs between $60 and $70 despite offering only 50% more capacity. High-VRAM cards get punished disproportionately, and the gap between an 8 GB and a 16 GB SKU on the same die is widening faster than raw memory scaling would suggest.

That curve has a second-order effect nobody advertises. A budget SKU is being cannibalized by its own bill of materials before it ships in volume, because the 3 GB parts needed for higher-VRAM configurations on entry-level cards eat most of the margin that made the SKU viable in the first place. Any card that needs multiple 3 GB modules to hit a competitive VRAM figure has a memory bill approaching what mid-range cards cost outright a year ago.

TrendForce reported on July 27, 2026 that graphics DRAM pricing has climbed alongside the broader memory market as suppliers redirect wafer capacity toward higher-margin products. On July 3, the firm projected conventional DRAM contract prices would rise a further 13% to 18% quarter-over-quarter in Q3 alone, tied to AI server demand pulling capacity away from consumer parts.

Those projections are the leading indicator to watch. If they hold, the July hike is not the last one this year.

Why AI Data Centers Are Draining Consumer GPU Memory

AI data centers now consume roughly 70% of global high-end DRAM production, according to analyst estimates from IDC and TrendForce cited by Gagadget. The HBM modules that pair with Nvidia's own H100 and B200 server chips require four times the wafer area of standard DRAM, which directly crowds out the GDDR supply gaming cards depend on.

Supply growth is not catching up. IDC projects DRAM supply growth of just 16% year-on-year, well below the 20% to 30% historical norm for the category. Market observers quoted by Basic Tutorials expect the shortage to persist until at least 2028, with no meaningful relief until additional fabrication capacity comes online.

There's an awkward loop hiding in that math. Nvidia sells the H100 and B200 systems whose HBM demand is pulling wafer capacity away from GDDR7, and Nvidia is now raising GeForce prices because the resulting GDDR7 shortage is inflating its own kit costs. The company benefits from the enterprise side of the same shortage that is squeezing its consumer division. Gamers are absorbing the collateral damage of a supply chain the company itself is bidding against.

This is the piece of the story that rarely makes it into the mainstream write-ups. It isn't a case of Nvidia being unable to secure memory. It's a case of Nvidia's most profitable customers, the hyperscalers running its data center chips, outbidding the consumer arm for the shared wafer pool. The consumer division loses that internal auction every quarter that AI demand holds.

And it will hold. Every publicly disclosed capex plan from Microsoft, Google, Meta, and Amazon points to further data center build-out through 2027, with memory content per accelerator rising rather than falling. There is no scenario in the current pipeline where hyperscaler HBM demand cools before the fabrication side catches up.

RTX 5090 Price Hike and What Retail Already Shows

Retail has already moved on the RTX 5090 price hike in markets where inventory is thin. Gagadget tracks the RTX 5090 from a UK launch MSRP of around £1,919 to Amazon listings between £3,599 and £3,649, roughly a 90% premium over launch price and well beyond what the announced kit increase alone would justify.

The RTX 5080 tells a cleaner story because it's a more common SKU. God is a Geek points to an Asus TUF Gaming RTX 5080 that sold for $1,199 last winter and now lists on Amazon at $1,595, a roughly 33% increase that sits inside the reported kit range. That's the retail behaviour to expect from cards where partner inventory has already rotated once.

China is where the compression is showing up fastest. The 16 GB GDDR7 RTX 5070 Ti sold for 7,200 yuan, about $1,064, in November 2025, rose to 9,299 yuan and then to 10,700 yuan, roughly $1,581, over the past several weeks. Some Chinese platforms have started removing Nvidia SKUs from sale entirely, expecting still higher pricing in the coming months rather than list at current wholesale rates.

The retailer behaviour is the tell. Pulling stock rather than repricing it says the retailer expects the next reprice to be worse than the current one. That is not a signal a mature market gives when a hike is one-off.

IGN's estimate, cited by Gagadget, puts a full pass-through scenario for the RTX 5090 near $6,500 in the US against the roughly $5,000 it was already fetching earlier in the year. That's the number to watch if you want to gauge how much of the wholesale increase board partners are actually absorbing versus passing on.

AMD and Intel Are Not the Escape Route

Switching brands doesn't get you out. Basic Tutorials reports AMD and Intel have both raised GPU and memory prices in 2026, and the timing lines up with the same Samsung contract adjustments that pushed Nvidia's kit costs. AMD's Radeon lineup relies on GDDR6, which sits inside the same wafer reallocation squeeze, so the escape route to team red closes as fast as it opens.

Console pricing offers no refuge either. The PlayStation 5, Steam Deck, and Nintendo Switch have all seen price increases in 2026 per Basic Tutorials, which is the clearest evidence that the problem is industry-wide memory economics rather than a Nvidia-specific pricing decision. The usual advice to wait out a GPU hike, or to sidestep it by buying a console, is unlikely to work this cycle.

Second-hand pricing follows new pricing with a lag, historically six to twelve weeks. Buyers who normally bridge a hike by picking up last-generation cards on resale markets should expect that window to compress this time, because the same memory economics apply to any card still being manufactured or refurbished with new modules.

What to Watch Before Holiday 2026

The near-term marker is inventory. Board partners are still selling through GPU kits purchased at pre-hike prices, and Gagadget reports they're absorbing part of the cost while that stock clears. Once it does, retail should align with the new wholesale rate, and the gap between US and UK pricing on the same SKU is the cleanest signal of how far that alignment has progressed.

The date to circle is Samsung's Q4 2026 DRAM contract negotiations. If contract prices settle at another double-digit increase in line with the TrendForce Q3 projection, a fourth Nvidia kit adjustment before the end of the year is more likely than not. If contract prices flatten, board partners have room to hold the line into the holiday buying window without a further round of MSRP moves.

One last variable. Nvidia has not publicly confirmed the increases, and gHacks noted via God is a Geek that the exact impact on retail prices remains unclear because the reported figures reflect supplier-side pricing rather than confirmed shelf pricing. That gap between wholesale and retail is where the next few weeks of price discovery will happen, and it's the reason no one should treat any single retail listing as the new baseline yet.