DDR5 memory kits are up nearly 500 percent in a year as AI buyers absorb DRAM supply

The AI build-out has reached the shelf where ordinary people buy computer parts, and it did not arrive gently. Consumer DDR5 memory now costs roughly five to ten times what it did a year ago.

The numbers

Tom’s Hardware price tracking puts a 128GB DDR5-6400 kit at $3,399, about ten times its lowest tracked price.

The mainstream kits tell the same story. A DDR5-6000 2x32GB kit that averaged around $222 in August 2025 now averages about $1,272, a 473 percent rise. DDR5-5600 2x32GB went from $191 to $1,118, up 485 percent. Even the entry-level 2x16GB kit went from $108 to $572.

The increases are consistent across speeds and capacities, which points to the whole category repricing rather than a shortage of one specific part.

Why: a supply shock, not a factory problem

The cause is a supply shock rather than a manufacturing failure. Hyperscale AI customers are reserving much of the industry’s future DRAM capacity, and memory that would otherwise flow to the consumer channel is being committed years ahead to data center buyers. The modules that used to be the predictable part of a build are now among the most expensive components in it.

That distinction matters for anyone hoping the situation resolves on its own. A yield problem or a plant outage clears with time. Long-dated capacity commitments from the largest buyers in the industry do not, and the consumer market is at the end of the queue.

What it means for builders

This is the first stage of the data center boom that consumers can feel directly, and memory is only where it shows up first. Anyone planning a build, an upgrade, a home lab or a local model rig is now competing for supply with companies buying capacity years in advance, and competing on price rather than on timing.

The practical implication is uncomfortable for the usual strategy of waiting. Component pricing is usually discussed as cyclical, with the implied advice to wait out a spike. The current DDR5 market behaves less like a cycle and more like an auction in which the largest bidders have already placed their orders, and waiting for prices to settle assumes a settling point that nothing in the current data indicates.