What is HBM memory and why is it making RAM and GPUs more expensive in 2025?

Last update: 09/12/2025

HBM Memory

Have you recently tried to buy a high-end graphics card or upgrade your computer's RAM? You were probably surprised by the prices, which have more than tripled for some kits. What's behind this increase? Among other reasons, the demand for HBM memoryBut what is HBM memory and why is it driving up the price of RAM and GPUs in 2025?

What exactly is HBM memory?High Bandwidth Memory)?

HBM Memory

It all started as a moderate trend in 2024 that quickly became an unavoidable reality in 2025. Suddenly, RAM and GPU prices began to climb rapidly. We've already alluded to this market movement in previous posts, along with the reasons behind it and its global impact(See the topics) The price increase of AMD GPUs due to the memory shortage y DDR5 RAM prices skyrocket: what's happening with prices and stock).

But today we're here to talk about the main protagonist of this shake-up: HBM memory. This acronym stands for High Bandwidth Memory or High Bandwidth Memory, and refers to a hardware technology which is generating a lot of buzz. And as you probably already suspect, it addresses a specific need related to Artificial Intelligence.

Unlike traditional GDDR memory, which is arranged horizontally on the motherboard, HBM chips are stacked verticallySo they introduce a radical architectural change: they look like small silicon skyscrapers. This 3D arrangement achieves extraordinary density in a minimal space: more in less.

Exclusive content - Click Here  AMD FSR Redstone debuts in Black Ops 7 with Ray Regeneration

Higher speed and lower consumption

And how do they connect to each other and to the processor? Through Through-Silicon Vias (TSVs), thousands of microscopic connections that run vertically through the chipsThese connections create ultra-fast data highways between the memory layers and the processor. For everything to work so well, it's essential that the HBM memory is as close to the processor as possible.

So, instead of being on separate chips soldered to the motherboard, HBM memory is stacked directly on or next to the processor (GPU or CPU). This is achieved using a silicon interposer, a special substrate that acts as a high-density connection platform. Thanks to this design, the electrical paths are shorter, resulting in a lower energy consumption, lower latency, and enormous bandwidth.

To give you an idea: GDDR6X memory, a latest-generation model, reaches approximately 1 TB/s of bandwidth. In contrast, Current versions of HBM3E exceed 1.2 TB/sAnd with HBM4 on the way, performance is expected to be much higher.

HBM Memory: Why RAM and GPUs are getting more expensive in 2025

Logically, given the widespread adoption of artificial intelligence, there is a growing demand for HBM memory In the technology market, all generative AI models have one thing in common: their voracious consumption of memory bandwidth. Traditional hardware technology can't keep up, but the HBM configuration solves this problem elegantly and efficiently.

Exclusive content - Click Here  Why your CPU never goes above 50% in games and how to fix it

But AI isn't the only driver. Other sectors, such as quantum computing, molecular simulation, or high-fidelity virtual realityThey also benefit from HBM's capabilities. It's clear: as these applications become more complex and demanding, the transition to architectures with high-bandwidth memory is inevitable.

So much so, that companies like NVIDIA, Google, and Amazon Web Services They have signed multi-year contracts to secure the supply of HBM memory.Who are the manufacturers? The epicenter is in Asia and the US: Samsung, SK Hynix Micron and Microsoft are the companies responsible for meeting this demand. They also manufacture traditional RAM... and that's the root of its higher price.

Production falls… prices rise

Memory slots

Naturally, all the time and resources of semiconductor manufacturing companies have been diverted to the production of HBM memory. And logically, this This reduces the available capacity for manufacturing GDDR and DDR memory. conventional. Production falls… shortages arise… prices rise… it's that simple.

It's important to note that manufacturing these memories is a complex process, substantially different from the production of GDDR and DDR. Therefore, it's not as simple as halting production of one model to resume production of the other. The same applies to the raw materials: specialized materials are required to produce HBM memory. In summary: They are separate design lines.

Exclusive content - Click Here  Musk's xAI is preparing a massive data center in Saudi Arabia with support from Humain and Nvidia chips.

And to this must be added the geographical concentration of producers in South Korea and the United States. This reality limits the global response capacity, which This further increases prices in Europe.What has been the impact on users? The following figures illustrate how much this has influenced the price increase of RAM and GPUs in 2025:

  • A 20% to 40% increase in wholesale prices by 2025 for the DDR5 RAM memories.
  • Quarter-on-quarter increases of 8% to 13% in DRAM for servers, with extreme cases of up to 40% – 50%.
  • Supply shortages graphics memory (GDDR6/GDDR7), affecting consumer GPUs.
  • Quarter-on-quarter increase of over 10% in LPDDR5X memories for mobiles.

HBM Memoirs: What to Expect in the Future

Free up RAM in Windows 11 without restarting your computer-0

In conclusion, we can say that Producing conventional memories is no longer a priorityAll eyes are on HBM memory. For example, Micron, one of the world's three largest memory manufacturers, recently announced it is leaving the retail market. For more details, read the article. Micron shuts down Crucial: the historic consumer memory company says goodbye to the AI ​​wave.

While other solutions emerge, consumers and businesses will have to contend with high prices and limited availability. HBM memory is directly responsible for the increased cost of RAM and GPUs in 2025. such a strategic resource for the advancement of AI and other technologiesIt is not surprising that it will continue to garner resources and attention in the months and years to come.