AMD, one of the largest providers of desktop CPUs and GPUs, is reportedly planning sweeping price increases for all of its existing products. According to Overclocked 3D, based on unnamed industry sources, AMD reportedly told its partners that it is setting new suggested retail prices for all of its CPUs, including the latest Ryzen 7 9800X3D as well as previous AM5 processors in the 7000 series. Overclock3D claims the new retail prices were scheduled to arrive Monday night, though at least for now, there are several pending Black Friday deals to keep the ship afloat.
GPU cost is also likely to be higher
All CPUs and GPUs rely on some amount of memory, even if it’s not as obvious as RAM (random access memory) and SSDs (solid state drives). When it comes to graphics cards, PC builders will have even more to worry about. Well-known GPU leaker harukaze5719, who has a relatively strong track record of sharing information on GPUs, posted on X that AMD’s GPUs will see universal price increases. Any GPU with 8GB of VRAM (aka lower-end GPU) will demand $20 more and a model with 16GB of VRAM will force PC builders to cough up $40 more. This comes after weeks of rumors about rising GPU costs. PC industry analyst Dan Nystedt claimed last month that AMD had already informed supply chain partners that it was raising GPU prices.
AMD 8GB VRAM: +20$
AMD 16GB VRAM: +40$– Channelgate
– 포시포시 (@harukaze5719) 2 December 2025
Gizmodo contacted AMD for comment, and we’ll update this post if we hear back. AMD’s 9000 series GPUs, such as the Radeon RX 9070 and RX 9070 XT, were nearly impossible to get at the suggested retail price at launch in March. The XT variant—which is built for 4K gaming—is still hard to buy at its estimated price of $600. Some GPUs with 8GB of VRAM include a version of the nearly $300 RX 9060 XT (there’s another with 16GB) and the older RX 7600. While this may not seem like much, considering you’re already spending over $500 on a graphics card, if GPU manufacturers pass the additional VRAM cost on to customers, it will become a much bigger problem.
Unlike Nvidia, AMD does not manufacture any GPUs of its own. Board partners, the companies that actually make these graphics cards, rely on companies to offer them the GPUs and VRAM they need to make these AIC, or add-in cards, that PC builders stick inside their desktop towers. Nvidia typically sells VRAM to GPU manufacturers to ensure that finished cards are in compliance with specifications. Last week, more unverified rumors stemming from China-focused social media site Weibo suggested that Nvidia had stopped supplying partners with the needed VRAM. This means each company must negotiate for its own memory. That’s a lot of small fish battling each other for scrap from major semiconductor makers.
GPU makers are already telling customers to expect higher prices. A customer service representative from AIC manufacturer PowerColor told users on Reddit to “buy before the last week of the year before prices go up.”
We need more options than a brand new PC
DRAM failures are hurting every PC component, and will no doubt affect self-contained mobile devices as well. Valve’s highly anticipated Steam Machine – a small-sized PC/console hybrid – will use a “semi-custom” AMD-designed SoC or system on chip, with reported graphics capabilities equivalent to the mobile version of the RX 7600. Some users have expressed concerns about the limited 8 GB VRAM. Valve software developer Pierre-Loup Griffais told Ars Technica that Valve has “done some work to improve video performance management” so that games are not hindered by low memory.
In the end, having lower memory specs may be the only way to keep costs down. The Steam Machine is likely to be more expensive than a PlayStation 5 today. We can only hope that it won’t be too expensive. Consumers need real options because buying a PC will become very expensive in 2026.
At the core of the issue is the increasing demand for high-speed memory from AI datacenters. They’re growing like weeds in communities across the US and beyond, and those making them are commanding such a premium on memory that semiconductor companies that typically make DRAM and NAND storage for consumer devices have focused on pumping more oxygen into the ever-expanding AI bubble.
If you want an example of just how limited DRAM is, Samsung – the Korean tech monolith that has a semiconductor division named DS – is reportedly limiting the number of DRAM chips demanded by Samsung’s own smartphone division. This comes as Samsung is on the verge of releasing its Galaxy S26 in early 2026. A report by Seoul Economic Daily (SeDaily), based on industry sources, reveals that DS is reportedly asking the Mobile Experience Division to renegotiate the terms every three months instead of maintaining supplies for every 12 months. The semiconductor arm reportedly does not want to miss out on the huge amount of money it is getting from the demand for AI data centers.
As companies find themselves eaten from the inside out to supply AI datacenters, consumers are left footing the bill. Just don’t expect the pain to go away anytime soon.
<a href