Instead, a discreet internal memo has set alarm bells ringing.
Behind the scenes, NVIDIA has told its board partners that a key part of every graphics card is about to get more expensive, and that quiet message is now turning into very real price rises on shop shelves across Europe and Asia.
The email that changed the mood
For months, industry analysts have warned that the cost of memory chips was heading sharply upward. Laptop prices jumped first, then RAM kits for desktops followed. Graphics cards, for a while, managed to sit in a strange bubble of stability.
That calm is now over. According to reports from Taiwanese outlet Benchlife, NVIDIA has officially notified its AIC partners (the board makers such as ASUS, MSI and Gigabyte) that the price of GPU memory – GDDR6 and the newer GDDR7 – is going up as existing supply contracts expire in January 2026.
NVIDIA is keeping the suggested retail price of its actual GPUs, the silicon chips, stable – but charging more for the memory that goes with them.
On paper, that sounds like a minor accounting detail. In practice, VRAM is one of the most expensive components on any modern graphics card, especially for 16 GB and 24 GB models targeting high-end gaming and AI workloads.
NVIDIA is also said to be increasingly interested in shipping “bare” GPUs to some partners, who then have to secure their own VRAM modules on the open market. In a context of rising DRAM prices, this exposes board makers even more to memory cost swings.
Why memory costs are spiralling
The memory squeeze does not come out of nowhere. Over the past year, major DRAM and NAND manufacturers have cut production to stabilise prices after a pandemic-era glut. At the same time, demand for high-bandwidth memory for AI accelerators has exploded, pulling up prices for more conventional chips like GDDR6.
When a handful of suppliers control most of global DRAM output, even relatively small shifts in capacity or demand can push contract prices up quickly. GPU partners have been drawing on older, cheaper inventory for months, but that buffer is now gone.
Once legacy stock runs out, board partners are forced to buy VRAM at new, higher prices – and that cost has to land somewhere.
➡️ I made this hearty recipe and felt instantly relaxed after eating it
➡️ After 60, “poor winterization can crack the shell” during the first freeze
➡️ This Boeing 737 looks like no other you’ve ever seen because it’s been adapted for Canada’s Arctic
➡️ Goodbye, Christmas tree : meet the plant hitting florists that’s set to trend in
➡️ At 50 and beyond, “underestimating total ownership costs is the most common mistake”
Board partners start passing the bill to gamers
With older stock drying up, manufacturers have begun to adjust retail prices. The Commercial Times reports that MSI moved first, raising prices on some new RTX 50 series models in December. Gigabyte and ASUS are expected to follow, affecting both NVIDIA and AMD cards.
The pattern is already visible in European pricing data:
- Cards with 16 GB of VRAM or more are seeing price rises in the 15–20% range.
- AMD Radeon RX 9000 models are typically up 10–18% over recent months.
- Entry-level and mid-range 8 GB cards are moving more slowly, but upward pressure is growing there too.
For a high-end card that used to sit around £800, a 15–20% increase can push it well beyond the psychological £900 barrier. That changes not just impulse buys but also the perceived value of upgrading from older GPUs like the RTX 20 or 30 series.
NVIDIA absorbs some, but not all, of the hit
Sources suggest NVIDIA is trying to keep its ecosystem attractive by swallowing part of the memory inflation rather than passing everything to partners. The reported hikes on NVIDIA-based cards look slightly lower than on some rival AMD models.
Still, even a partial offset cannot fully counter double-digit increases in VRAM contract prices. The game of margin-sharing between NVIDIA and its AICs only decides how pain is split, not whether it exists.
Retail prices move last, but when upstream contracts reset, the retail side has almost nowhere to hide.
Different playbooks at NVIDIA and AMD
The two big GPU designers now face a tricky balancing act between performance and affordability.
NVIDIA leans on cheaper 8 GB designs
On the NVIDIA side, partners are said to be prioritising cards with 8 GB of VRAM, such as the rumoured RTX 5060 and 5060 Ti 8 GB. These models use fewer memory chips, which keeps bill-of-materials costs lower and allows manufacturers to stay closer to existing price brackets.
That strategy, though, has its own risk. In demanding 1440p or 4K games, or in video editing and AI workloads, 8 GB is starting to feel tight. Buyers who remember how quickly 4 GB cards became obsolete may think twice before spending several hundred pounds on a card that could age badly.
AMD pushes 16 GB performance play
AMD, according to the same reports, is leaning into higher-capacity models. Its partners are encouraged to push 16 GB XT variants, even as memory costs climb. That gives AMD bragging rights in raw VRAM capacity, which helps in benchmarks and in memory-heavy scenarios.
But those cards now sit in a pricier tier. In a context where disposable income is under pressure, AMD’s bet is that some gamers will still pay extra for more headroom and better 1440p performance, rather than compromise on longevity.
| Brand strategy | Typical VRAM focus | Main risk | Main advantage |
|---|---|---|---|
| NVIDIA partners | More 8 GB models | Shorter useful lifespan at higher resolutions | Lower up-front prices, easier to market |
| AMD partners | More 16 GB XT models | Higher final retail prices, slower sales | Stronger performance story and future-proofing |
What this means if you plan to upgrade
For anyone sitting on a GTX 10, RTX 20, or older Radeon card, timing now matters a lot. Stock that was produced under previous contracts is still dribbling through channels, and retailers may run promos to clear shelves. Once that stock is gone, new batches built with expensive VRAM will set the baseline.
A practical scenario looks like this: you see a 16 GB mid-range card at £550 today. Three months from now, that same model might hover closer to £630–£650, not because performance changed, but because the underlying memory modules jumped in cost and a new shipment arrived.
Budget-conscious buyers may start comparing prebuilt gaming PCs more seriously. Sometimes, system integrators can smooth component price spikes by negotiating bulk deals, making a full system oddly better value than just the GPU upgrade on its own.
Key terms that shape these price shifts
Two bits of jargon sit at the centre of this story:
- VRAM (video RAM) is the dedicated memory on your graphics card. Games load textures, geometry and frame buffers into it. More capacity reduces stutter and texture pop-in at high resolutions.
- MSRP (manufacturer’s suggested retail price) is the target price set by NVIDIA or AMD for a given GPU tier. Board partners can deviate from it, especially when their own component costs change, as is happening now with VRAM.
When NVIDIA says GPU MSRPs are unchanged, that refers to the chip alone, not the full card with cooling, power delivery and memory. Board makers build around that and set actual street prices according to their own costs and margins.
Looking ahead: scenarios for 2026 and beyond
If DRAM manufacturers ramp up capacity again in the second half of 2026, VRAM prices could stabilise or even ease. That would reduce pressure on GPU pricing, especially for models with 16 GB and above. But the surge in AI demand makes that far from guaranteed.
There is also a risk of design compromises. To keep cards under certain price points, partners might ship more models with narrower memory buses or fewer chips, affecting performance. On paper, you still get the same GPU name; in practice, variants can behave quite differently in real games.
For enthusiasts, one defensive tactic is to target cards with a mix of decent VRAM capacity and wide memory buses, even if it means staying one tier below the absolute top end. That balance often delivers better value over a four- to five-year span than chasing the flashiest flagship during a pricing spike.
Originally posted 2026-03-13 00:12:27.