AMD has reportedly raised the price of the GDDR6 memory it supplies to graphics card makers, effective October 1, 2026. According to information reported by the Taiwanese hardware site BenchLife on October 10, the Chinese industry site "Bobantang" confirmed the price revision. It applies to memory that AMD supplies to partner companies alongside its GPUs, and the size of the increase has not been disclosed. AMD has not made an official announcement, and none has been confirmed. Still, in the memory market, the traditional assumption that choosing the older GDDR6 standard keeps costs down is becoming harder to rely on. GDDR6 production is being scaled back, and the newer GDDR7 is also limited by production capacity. What will shape future Radeon pricing is not just the memory standard, but whether each company can secure the volume it needs, and when added capacity actually reaches the market.
AMD Raises GDDR6 Supply Prices, Affecting Graphics Card Manufacturing Costs
The report concerns a price revision for components that AMD supplies to its partners. It does not mean AMD itself manufactures GDDR6 memory. Graphics card makers obtain GPUs and compatible memory from AMD, then combine them with circuit boards, power circuits, cooling hardware and other parts to build and sell finished products. The reported increase affects the procurement cost of a component needed in that manufacturing process.
GDDR6 is also used in current Radeon products. For example, according to AMD's official product specifications, the 16GB model of the Radeon RX 9060 XT is equipped with GDDR6 memory. A shrinking GDDR6 supply is therefore a concern not only for repairing or selling inventory of older-generation cards, but also for manufacturing current products.
That said, a higher memory procurement price does not necessarily mean finished-product prices will rise by the same proportion. The actual increase depends on how much of the production cost memory accounts for and how far board makers are willing to adjust their margins. And if a manufacturer still holds components purchased before the increase, it may take time for the new procurement price to show up in product prices. The supply price from AMD to partners, the manufacturer's suggested retail price, and the actual store price should each be considered separately.
Memory is not the only thing pushing up manufacturing costs. BenchLife's report says electronic components such as capacitors, resistors and power control ICs have also become more expensive, and that printed circuit boards (PCBs) are seeing both higher prices and longer lead times. However, the specific price increases and each part's share of production cost are not given, so it is not possible to attribute any rise in graphics card prices entirely to GDDR6.
Even if memory can be secured, finished products cannot ship if other parts are short. Judging the graphics card supply situation requires looking not just at the prices of individual components, but at the procurement of all the parts needed for manufacturing.
GDDR6 Production Is Shrinking, and Next-Generation GDDR7 Faces Supply Limits
Behind this price increase is a shift in the supply structure of the graphics memory market.
According to an analysis published by the market research firm TrendForce on September 30, memory makers are reducing GDDR6 production, and some are gradually exiting the market. Meanwhile, some PC makers are reconsidering GDDR6 in order to hold down component costs. On the demand side, there is a push for cheaper memory, yet on the supply side, production of it is being reduced.
Would moving to the newer GDDR7 solve the problem? According to TrendForce's analysis, GDDR7 is not securing sufficient supply either, because advanced manufacturing capacity is being allocated preferentially to products such as server DRAM. Demand for GDDR7 from AI chips is also strong, so supply and demand are no longer determined by gaming graphics cards alone.
| Memory standard | Supply constraint noted by TrendForce on September 30 |
|---|---|
| GDDR6 | Production cutbacks by manufacturers and gradual market exits |
| GDDR7 | Priority allocation of advanced manufacturing capacity to server DRAM and similar products |
Even within graphics memory, the problems differ: for GDDR6, the cutback of a legacy product; for GDDR7, competition for production capacity with other, more profitable products.
This is TrendForce's analysis of the memory market as a whole, and it does not verify AMD's individual procurement prices or contract terms. Even so, it explains why choosing older memory does not guarantee cheap, stable procurement.
In fact, memory prices rising despite weakening demand had already been observed. In a survey TrendForce published on July 3, it explained that demand for GDDR6 and GDDR7 fell because of factors such as sluggish notebook PC shipments, while manufacturers redirected production capacity to other mainstay products, tightening supply and pushing up prices.
In other words, memory does not necessarily become more expensive because gaming GPU sales are strong. Even when demand is falling, prices can rise if supply falls faster. And even for a mature standard, it is difficult to maintain the traditionally low prices once manufacturers scale back production.
Even as DRAM Fabs Expand, GDDR6 Prices May Not Fall Soon
Memory makers are expanding production capacity in response to the shortage. But the announcement of a new fab does not mean enough product will immediately reach the market.
In its fiscal fourth-quarter 2026 prepared remarks, published on September 30, Micron laid out plans to begin initial production at its first Idaho fab in mid-2027, and at its second fab and its DRAM expansion fab in Japan in late 2028.
| Micron DRAM fab plans | Planned start of initial production (calendar year) |
|---|---|
| Idaho Fab 1 (ID1) | Mid-2027 |
| Idaho Fab 2 (ID2) | Late 2028 |
| Japan DRAM expansion fab | Late 2028 |
These are planned start dates for initial production at the fabs. They do not indicate when GDDR6 output will increase or when prices will fall.
In the same document, Micron explains that it will take several quarters after a new fab begins initial production before it contributes meaningfully to supply growth. The time a fab is approved, the time its first wafers are produced, and the time enough product reaches the market are all different.
Which types of DRAM the new fabs will produce also matters. Even if overall DRAM capacity grows, that increase will not necessarily go toward GDDR6. As TrendForce notes, if manufacturers continue to prioritize server memory, supply of memory for gaming graphics cards may not improve right away even as fab expansion proceeds.
When supply cannot grow quickly, securing the needed volume in advance becomes as important as negotiating prices.
In the same materials, Micron disclosed that it has signed 26 multiyear agreements with strategically important customers. These contracts use a structure in which customers bear payment obligations even if they do not actually take the contracted volumes, allowing them to secure long-term supply.
It has not been confirmed that AMD is a party to any of these agreements. Still, they show that in the memory market, supply is secured not only by negotiating unit prices but also by committing to purchase volumes over a set period.
Because the output gains from new fabs will not appear immediately, GPU makers face the added challenge of deciding on what terms to secure the memory they need in the meantime. The report of AMD's price increase alone does not reveal such long-term procurement terms.
NVIDIA Has Not Issued a Price Increase Notice, and May Adjust Supply Volumes Instead
In its October 10 article, BenchLife reports that NVIDIA has not sent partner companies a notice of higher shipment prices for the fourth quarter of 2026.
The outlet also suggested that, rather than raising GDDR6 or GDDR7 prices, NVIDIA may manage GeForce RTX margins by adjusting the supply volume of GPU-and-memory bundles. However, this is BenchLife's speculation, not an official NVIDIA policy.
A price freeze is also a separate matter from manufacturers being able to buy the volumes they want.
Even if the unit procurement price is unchanged, a reduction in the supplied quantity of GPUs and memory limits the number of graphics cards board makers can produce. Conversely, even if procurement prices rise, manufacturers may be able to keep shipping finished products if they can secure the volumes they need reliably.
For that reason, whether a price increase notice has been issued cannot by itself tell us whether AMD or NVIDIA products will be cheaper to buy going forward.
The information that NVIDIA has not notified partners of a price revision at this point is also no guarantee that prices will hold through the year. Because GDDR7 also faces production capacity constraints, the absence of an NVIDIA price increase notice does not mean the overall shortage of graphics memory has been resolved.
To judge the impact on Japanese consumers, it will be necessary to see whether prices change when the same graphics cards come back in stock, and whether relatively affordable models continue to be replenished.
If only higher-priced products remain on store shelves, the budget needed to buy a card rises even if suggested retail prices do not change.
Whether GDDR6 and GDDR7 supply has improved can be judged not only by memory prices, but also by whether graphics cards with the required memory capacity can be bought steadily at multiple retailers. The key question going forward is whether an improved component procurement environment shows up in actual product prices and supply volumes.