Are graphics cards getting more expensive?

After running around offline PC-building markets and working with channel distributors recently, you can clearly feel that a new round of graphics card price hikes has already rolled out across the board. This isn’t retailers fabricating hype to raise prices—upstream original manufacturers, memory chip suppliers, and foundry capacity are all increasing costs in sync, and for every single price increase, there’s solid evidence to point to.

On July 23, NVIDIA officially issued notices to AIC manufacturers such as ASUS, MSI, and Colorful, raising the supply prices for bundled GeForce graphics card GPU + VRAM kits across the entire lineup. Whether they’re new RTX 50 series models using GDDR7 or older mid-range cards using GDDR6 memory, all of them are included in the repricing scope. Earlier than that, AMD also implemented price adjustments: the supply price for the entire line of Radeon graphics cards was increased uniformly by 10%. With both chipmakers tightening supply and raising the cost of purchasing chips, terminal price increases are only a matter of time.

The main driver behind the hikes is runaway VRAM chip costs. Samsung, SK hynix, and Micron—three of the largest memory manufacturers—now have more than 70% of their high-end DRAM capacity fully diverted to AI server HBM high-bandwidth memory. HBM profits are far higher than consumer-grade GDDR VRAM, so factories prioritize AI orders and continue to shrink the production lines allocated to gaming graphics cards. By comparing spot prices, you can see the increase directly: the per-GB price of GDDR6 VRAM rose from $2.5 at the end of last year to $7.5 today—tripling directly. The bulk purchase price of 16GB GDDR7 VRAM modules jumped from $80–$90 to $200–$300, and the per-card VRAM material cost increased by nearly 1,000 yuan immediately.

Beyond VRAM, the entire set of add-on supplies for graphics cards—PCB substrates, power-supply components, cooling fans, and metal backplates—are also seeing synchronized price hikes. On top of that, NVIDIA is proactively reducing consumer graphics card supply by 15%–20%, causing the channel side to move into a “stop shipments and ration sales” situation. Many distributors have paused external quotes, waiting to reset pricing once the next batch of higher-priced inventory arrives. Looking at the terminal price increases: overall costs for mid-range cards with 8GB VRAM have risen by around 600 yuan, 12GB models are up 900 yuan, and high-end 16GB cards have reached an increase of 1,200 yuan. For the previously 2,599 yuan tier RTX 5060, the channel’s current purchase price has already touched 3,199 yuan. The secondary-market price of the flagship RTX 5090 has broken through the 30,000-yuan mark.

According to industry analysts, this round of price hikes is due to a structural imbalance in capacity, and it will be difficult to relieve it in the short term. Building new storage clean rooms and wafer production lines has a long construction cycle. HBM high-profit orders have already been locked with factory capacity for the long term, and capacity allocated to consumer graphics cards cannot be expanded in the short term. Industry consensus is that the high-price trend will remain at least through 2027, and may even extend into 2028.

For players preparing to build a PC or upgrade their graphics cards, there is currently no room for a “wait a bit and the price will drop” scenario. Once the channel’s existing low-price inventory is consumed, all graphics cards currently on sale will be updated with higher prices. For people with urgent needs—building PCs, doing local AI deployments, or playing new AAA titles—there are still spot low-priced cards available, so they can buy early. For non-urgent users, they can only keep watching for the long run, and there are no signs of a cost decline in the short term. #存储芯片 #ai #显卡