“AI Is Eating All the Memory.”

“AI Is Eating All the Memory.”

That may be true and it is also the perfect alibi.

A 32 GB DDR5 kit now costs around €500. A 96 GB kit can reach €2,300. An RTX 5090 is listed near €5,000.

The usual explanation is simple: AI data centers consumed supply, and manufacturers shifted capacity to HBM.

But look one layer deeper.

The physical cost of producing ordinary DRAM has not multiplied at the same rate. Micron recently reported an 87% gross margin in its mobile and client memory unit. Samsung, SK hynix and Micron control roughly 90% of global DRAM. On June 25, a US antitrust class action accused them - allegations not yet proven - of coordinating supply reductions while using the HBM transition as cover.

This reveals a wider regulatory risk: AI can become a universal explanation for any engineered shortage.

I call it AI scarcity laundering.

Real AI demand creates the initial constraint. Then manufacturers, distributors or infrastructure owners can restrict supply, maintain “capacity discipline,” raise prices and say:

“AI ate everything.”

The same pattern could spread from DRAM to NAND, advanced packaging, optical networking, power electronics, transformers and cooling.

And the damage does not stop at gaming PCs. It passes through phones, vehicles, industrial automation, medicine, communications, defense ,- and c = a + b: persistent local human - AI systems whose memory, identity continuity and autonomy depend on affordable silicon under local custody.

This is the earthly engineering problem: if silicon memory is redirected to the central AI “brain,” the distributed biological and machine body loses its ability to sense, remember and act. A larger brain does not compensate for starving organs.

Retail price caps alone are not enough. Regulators need upstream capacity and cost transparency, monitoring of parallel production cuts, minimum supply obligations for general-purpose memory, and strict conditions on public fab subsidies.

Excess profits do not automatically build the capacity society needs. They may simply fund more of the premium capacity causing the shortage.

AI development is strategic. So is access to memory.

We should not allow “AI demand” to become a license to privatize the bottlenecks of civilization.