GPU Down, RAM Up: PC Chaos

In what can only be described as cosmic irony for PC enthusiasts, the market for computer components has entered a bizarre phase where two critical pieces of hardware are moving in opposite directions. After years of sky-high graphics card prices driven by cryptocurrency mining and supply chain disruptions, GPU prices are finally stabilizing and returning to their manufacturer-suggested retail prices (MSRP). However, just as gamers and PC builders might dare to dream of building their ideal systems, RAM prices have skyrocketed—some kits now cost three times what they did just three months ago.

The GPU Market: Finally Coming Back to Earth

For PC builders and upgraders who have waited on the sidelines since the 2020s began, there’s finally some good news. The graphics card market, which has been characterized by scarcity and inflated prices, is showing signs of normalcy. Popular models like AMD’s Radeon RX 9060 XT and 9070 series are now available at or slightly under their MSRP. Similarly, Nvidia’s RTX 5060, 5070, and 5080 series can be found without the excessive markups that plagued consumers throughout the pandemic and AI boom years.

This stabilization comes as a relief to many who had postponed their upgrades, waiting for prices to drop to reasonable levels. The normalization is partly attributed to improved supply chains and a gradual decrease in demand from cryptocurrency miners and data centers focused on AI workloads—at least for the GPU units themselves.

Meanwhile, RAM Prices Are Going Through the Roof

However, joy over affordable GPUs is short-lived for anyone looking to build a new system or upgrade their existing one. Memory prices have entered what can only be described as a crisis phase, with costs for RAM kits (especially DDR5) surging dramatically in just a few months. According to price comparisons from August to November 2025, the increases are staggering:

Patriot Viper Venom 16GB (2 x 8GB) DDR5-6000 memory kit: $49 → $110 (124% increase)
Team T-Force Vulcan 32GB DDR5-6000 memory kit: $82 → $310 (278% increase)
Team Delta RGB 64GB (2 x 32GB) DDR5-6400 memory kit: $190 → $700 (268% increase)

Unfortunately, DDR4 hasn’t escaped this price surge either, and the crisis extends to storage solutions like SSDs, though to a lesser extent. DDR5 RAM from Micron

The AI Factor: Memory’s New Master

The culprit behind this unexpected memory crisis is the booming AI industry. Data centers worldwide are scrambling to secure massive amounts of memory to support their AI workloads, which require enormous quantities of DRAM for training large language models and running neural networks. Unlike traditional computing applications, AI training requires keeping vast datasets in memory simultaneously, pushing demand for high-capacity, high-speed DDR5 and even more specialized memory solutions like HBM (High Bandwidth Memory).

As reported by XDA Developers, the AI industry’s demand for memory is particularly intense because “generative AI requires a huge amount of both memory and storage, as training and running models require data centers filled with GPUs; GPUs that can have hundreds of gigabytes of DRAM paired with multiple terabytes of flash storage.”

Why AI Consumes So Much Memory

AI applications are fundamentally memory-intensive because they must hold enormous datasets and model parameters in memory simultaneously during processing. These models, often containing billions or even trillions of parameters, need to keep their weights and intermediate calculations in fast memory (like DRAM) to process information quickly. Large language models, in particular, require what’s called a “context window”—the amount of text that the model can process at once—which necessitates keeping all that text in memory simultaneously.

The exponential growth in AI model complexity has meant that data centers now require exponentially more memory than traditional server applications. According to research, the memory requirements for AI training have been doubling approximately every 3-4 months, far outpacing other types of computing workloads.

The Dilemma for PC Builders

This creates a unique and frustrating predicament for PC builders. For the first time in years, one can reasonably expect to find a graphics card at or near its intended price point. But if they want to build a new system or upgrade their memory to keep pace with modern software requirements, they’re faced with exorbitant costs that make even modest upgrades prohibitively expensive.

“It’s not a bad time to upgrade your gaming PC,” notes Ars Technica’s Andrew Cunningham. “But it’s not a good time to build a new PC or swap your older motherboard out for a new one that needs DDR5 RAM.”

Industry Response and Predictions

Major tech companies are scrambling to secure their memory supplies. According to Bloomberg, Lenovo is “stockpiling memory and other critical components to get it through 2026 without issues.” Apple, with its vertical integration approach, may be in a relatively better position due to its ability to secure memory allocations in advance and its substantial profit margins.

However, smaller companies like Framework are facing significant challenges. The laptop maker has already announced it will need to increase memory pricing and has stopped selling standalone RAM kits in its online store to combat scalpers capitalizing on the shortages.

The situation is expected to persist into 2026, with analysts predicting that DRAM prices could rise another 30% in Q4 2025, followed by up to 20% more in early 2026. This comes atop roughly 50% price increases already seen this year.

Historical Context: Memory Price Cycles

This isn’t the first time the PC industry has faced memory price volatility. The memory market is notoriously cyclical, with periods of shortage followed by oversupply. During the pandemic-related chip shortages of 2021 and 2022, demand spikes and supply chain disruptions created shortages across all semiconductor components. However, unlike GPU price fluctuations driven by cryptocurrency mining, the current DRAM crisis is being driven by long-term structural shifts in where the memory is being used.

Previous shortages in 2016-2017 led to oversupply and dramatically low prices in 2018-2019. Whether that pattern will repeat this time around is uncertain, given that the AI demand for memory appears to be structurally higher and more sustained than previous demand spikes.

Looking Forward: When Will Prices Normalize?

Unfortunately, there’s little indication that memory prices will drop significantly in the near future. The fundamental issue is that memory manufacturing requires years of lead time for new capacity, and manufacturers must make decisions about future supply based on today’s uncertain demand patterns. The ongoing expansion of AI infrastructure means that demand for memory is likely to remain high for the foreseeable future.

Additionally, as reported by various tech publications, memory companies might actually prefer the current situation for their own profitability. High-bandwidth memory (HBM) used in data center applications typically has higher profit margins than consumer DRAM, creating an incentive for manufacturers to prioritize HBM production over standard consumer modules.

Conclusion

The current PC hardware market presents a curious paradox—consumers can finally purchase graphics cards at reasonable prices after years of frustration, only to find that memory has become prohibitively expensive. This new landscape reflects broader shifts in technology where AI workloads are redefining demand patterns for basic components.

For PC builders and enthusiasts, the best strategy might be patience and selectivity. While GPU upgrades are now more affordable, memory upgrades should probably wait for supply chains to adjust. In the meantime, those looking to build systems will face difficult trade-offs between processing power and memory capacity.

This situation also highlights the interconnected nature of the global technology supply chain, where developments in one industry (AI) can dramatically affect consumers in seemingly unrelated markets (PC building). It’s a reminder that in our increasingly connected technological world, the impact of new technologies ripples far beyond where we might initially expect.

Sources: