As artificial technologies intelligence develop, they also rewrite the standards in the hardware world. Autonomous artificial intelligence (Agentic AI) models, which have been on the rise recently, cause processors (CPUs) in data centers to require much higher memory capacities.

Latest reports from the industry show that processors will soon be equipped with huge memories such as 400 GB. However, this situation makes the ongoing global DRAM shortage much more intractable towards 2027.

Processor Need Is Increasing in Data Centers

Until now, most of the artificial intelligence workload in data centers was undertaken by graphics cards (GPU). While the GPU and CPU ratio in a data center was around 8:1 in the past, as autonomous artificial intelligence models require higher overall processing power, this ratio first drops to 4:1, and is expected to approach 1:1 in the near future.

According to a report by Korea-based SE Daily, based on industry sources, processor manufacturers now plan to equip their artificial intelligence-focused CPUs with memory between 300 and 400 GB. Considering the 96-256 GB DRAM values ​​offered per chip in current systems, this increase means a huge leap for the hardware world. It is not yet clear whether this high capacity will be provided by standard DIMM modules on the motherboard or by new generation memory standards (HBM, etc.) integrated directly into the processor package. However, AMD is known to have produced EPYC processors with HBM memory in the past, and the industry is expected to focus more on these integrated solutions.

Memory Capacity Race Among Competitors

Competition in memory capacity is not limited to standard processors; A fierce race is also going on in the field of graphics cards and special artificial intelligence chips. Nvidia’s new generation artificial intelligence chip “Vera Rubin” offers 288 GB memory capacity through eight HBM chips.

Its biggest rival, AMD, increases this value to 432 GB with its new generation MI400 accelerator. On the other hand, Google’s recently announced 8th generation special artificial intelligence chip TPU 8i model is expected to have 288 GB HBM capacity. The fact that Intel’s

What Awaits the End User?

So how is this huge demand for AI hardware affecting standard consumer electronics? Although memory manufacturers earn huge revenues, they have great difficulty keeping up with this ever-increasing demand. Samsung’s previous pessimistic warnings for 2027 confirm the general concerns in the industry. Due to increasing market needs, memory manufacturers are dedicating their production lines to high-end artificial intelligence memories, which have much higher returns.

Just like Samsung stopped producing LPDDR4 memory and turned entirely to the more profitable LPDDR5, companies are slowly withdrawing from lower segment products. This situation leads to a shortage of RAM in standard consumer electronics products such as smartphones, computers and tablets that we use in daily life, and therefore to much more serious price increases in the coming years.

What do you think about the development of artificial intelligence hardware and possible memory price increases? Don’t forget to express your thoughts in the comments section below.