“NVIDIA Has Informed Its Partners”: a new wave of price hikes hits graphics cards and manufacturers have no choice

A low-key internal memo, rather than a public announcement, has become the real warning sign.

Away from the spotlight, NVIDIA has informed its AIC partners (add-in card manufacturers such as ASUS, MSI and Gigabyte) that a crucial part of every graphics card is about to cost more. That “quiet” change is now showing up as genuine price increases on retail listings across Europe and Asia.

The email that shifted sentiment around NVIDIA GPU memory (GDDR6/GDDR7)

For months, analysts have been flagging that memory pricing was on a steep upward path. Laptops were the first to reflect it, followed by desktop RAM kits. For a time, graphics cards oddly avoided the worst of the turbulence.

That period of relative calm is now ending. Reporting from Taiwanese outlet Benchlife says NVIDIA has formally told partners that GPU memory costs - GDDR6 and newer GDDR7 - will rise as existing supply agreements roll off in January 2026.

NVIDIA is reportedly holding the suggested retail price of its GPU silicon steady, while increasing the cost of the memory that accompanies it.

What looks like a small line-item change matters in reality because VRAM is among the priciest parts of a modern graphics card - particularly on 16 GB and 24 GB models aimed at high-end gaming and AI workloads.

NVIDIA is also said to be increasingly willing to ship “bare” GPUs to certain partners, leaving them to source VRAM modules themselves on the open market. With DRAM pricing trending upwards, that approach makes board makers even more exposed to swings in memory costs.

Why memory pricing is climbing so quickly

This crunch has been building for some time. Over the last year, the biggest DRAM and NAND producers have reduced output to firm up pricing after an oversupply that followed the pandemic period. At the same time, demand for high-bandwidth memory used in AI accelerators has surged, and that pull-through effect is lifting prices for mainstream products such as GDDR6 too.

Because a relatively small number of suppliers account for most global DRAM production, even modest changes in capacity or demand can push contract pricing up at speed. GPU partners have been able to lean on older, cheaper inventory for months - but that cushion is now being used up.

When legacy stock is exhausted, board partners have to buy VRAM at the new, higher rates - and those costs have to be absorbed somewhere.

Other headlines that appeared alongside the report

Board partners begin passing costs on to gamers

As cheaper stock thins out, manufacturers are starting to reset pricing. Taiwan’s Commercial Times reports that MSI acted first by increasing prices on some fresh RTX 50 series models in December. Gigabyte and ASUS are expected to make similar moves, with knock-on effects across both NVIDIA and AMD board designs.

European pricing trends already hint at the direction of travel:

• Cards with 16 GB of VRAM or more are rising by roughly 15–20%.
• AMD Radeon RX 9000 models are commonly up by around 10–18% compared with recent months.
• Entry-level and mid-range 8 GB cards are increasing more slowly, although upward pressure is building there as well.

On a premium card that previously sold for about £800, a 15–20% increase can lift pricing well past the psychological £900 threshold. That affects not only spur-of-the-moment purchases but also how worthwhile an upgrade feels for owners of older GPUs, including the RTX 20 and RTX 30 generations.

NVIDIA offsets part of the rise, but cannot cancel it

According to sources, NVIDIA is trying to keep its platform compelling by absorbing some of the memory inflation instead of pushing all of it on to partners. In practice, reported increases on NVIDIA-based cards appear a touch less severe than rises seen on certain AMD models.

Even so, partial cushioning can only do so much when VRAM contract pricing is climbing in double digits. The negotiation between NVIDIA and its AIC partners determines how the pain is shared - not whether it exists.

Retail pricing is the last domino to fall; once upstream contracts reset, there is limited room for the channel to hide.

NVIDIA vs AMD: two different approaches to performance and price

Both major GPU designers are now trying to balance competitiveness with affordability as memory becomes more expensive.

NVIDIA leans towards lower-cost 8 GB designs

On the NVIDIA side, partners are reportedly favouring more 8 GB products, including the rumoured RTX 5060 and RTX 5060 Ti 8 GB. The logic is straightforward: fewer memory chips reduce the bill of materials, making it easier to keep cards near familiar price points.

The trade-off is longevity. In heavier 1440p and 4K titles - and in creative workloads such as video editing or AI tasks - 8 GB is increasingly restrictive. Shoppers who remember how quickly 4 GB cards felt outdated may hesitate before paying several hundred pounds for a model that could run out of headroom sooner than expected.

AMD bets on 16 GB “XT” positioning

AMD, per the same reporting, is encouraging partners to lean harder into higher-capacity options, with a particular focus on 16 GB XT variants even as memory costs rise. That creates an easy headline in benchmarks and helps in scenarios where VRAM capacity is the limiting factor.

However, that strategy pushes cards into a higher price band. With household budgets under strain, AMD’s wager is that a meaningful slice of gamers will still pay more for extra breathing room and stronger 1440p results rather than accept a shorter upgrade cycle.

Brand strategy	Typical VRAM focus	Main risk	Main advantage
NVIDIA partners	More 8 GB models	Reduced useful lifespan at higher resolutions	Lower upfront pricing and easier positioning
AMD partners	More 16 GB XT models	Higher final retail prices and potentially slower demand	Stronger performance narrative and better future-proofing

What it means for your upgrade plans

If you are still using a GTX 10, RTX 20, or an older Radeon, timing has become far more important. Cards manufactured under earlier supply contracts are still filtering through distribution, and retailers may run short-term offers to clear remaining stock. Once that inventory is gone, new production built with higher-cost VRAM will set the new normal.

A realistic example: a 16 GB mid-range card listed at £550 today could sit nearer £630–£650 in three months’ time - not because anything about the GPU has improved, but because the cost of the underlying memory modules has increased and a newer shipment has arrived.

Value-seekers may also start weighing up prebuilt gaming PCs more seriously. System integrators can sometimes soften component spikes by securing bulk terms, which can make a whole tower appear better value than a standalone GPU upgrade.

One more practical consideration: as pricing becomes volatile, it is worth paying closer attention to the exact specification of the card you are buying (VRAM capacity, memory speed, and memory bus width) rather than relying purely on the GPU name. In periods like this, similarly named variants can end up offering noticeably different real-world performance depending on the memory configuration vendors choose to keep costs down.

Key terms shaping the price moves

Two pieces of jargon sit at the heart of the story:

• VRAM (video RAM): the dedicated memory on a graphics card. Games store textures, geometry and frame buffers there. More capacity can reduce stutter and texture pop-in at higher resolutions.
• MSRP (manufacturer’s suggested retail price): the target price set by NVIDIA or AMD for a GPU tier. Board partners can diverge from it, particularly when their own component costs change - as is happening now with VRAM.

So when NVIDIA says GPU MSRPs are unchanged, that is a statement about the chip itself, not the full graphics card (cooling, power delivery, PCB and memory). AIC partners build around the GPU and then set street prices based on their costs and margin requirements.

Looking ahead: possible outcomes for 2026 and beyond

If DRAM manufacturers expand capacity again in the second half of 2026, VRAM pricing could level off or even ease. That would relieve pressure on graphics card pricing, particularly for models with 16 GB and above. However, continuing growth in AI demand makes that outcome uncertain.

There is also a risk that vendors respond with quieter design compromises. To keep products under specific price points, partners may release more variants with narrower memory buses or fewer chips, which can reduce performance. The name on the box may look familiar, but the in-game experience can differ meaningfully between versions.

For enthusiasts trying to protect long-term value, one defensive approach is to choose a card that balances solid VRAM capacity with a wider memory bus, even if that means buying one tier below the absolute flagship. Over a four- to five-year ownership window, that blend often works out better than chasing the top model during a pricing spike driven by memory costs.