Consumer RAM prices have surged in 2026, with a 32GB DDR5 kit cited at $374.97 on Tom’s Hardware’s early June tracker after selling near $80 to $120 a year earlier, according to the source material. The increase matters because memory is becoming a much larger share of PC costs as DRAM makers steer more capacity toward AI-focused high-bandwidth memory.

The price shock is broad across common desktop configurations. A 64GB DDR5 kit that sold near $150 to $200 through much of 2025 now often lists at $600 or more, according to the report. The source material says consumer DRAM is running three to six times its 2024 and 2025 lows, depending on the kit and timing.

The increase is also showing up inside finished PCs. HP told investors that memory had risen to about 35% of build materials, up from 15% to 18% a quarter earlier. That shift means RAM is no longer a secondary line item for many buyers; it can be one of the largest costs in a build.

The core confirmed mechanism is capacity allocation. Samsung, SK Hynix and Micron make nearly all global DRAM, and the same advanced manufacturing base that supplies DDR5 for PCs can be used for high-bandwidth memory, or HBM, used with AI accelerators. The report says HBM modules can bring far more revenue than standard DDR5, giving suppliers a strong incentive to favor AI-related output.

AI Demand Reprices PC Builds

The price increase matters because PC builders, gamers, workstation buyers and small system vendors are all exposed to the same memory market. Higher DRAM costs can raise the price of new desktops, make upgrades harder to justify and force buyers to choose between capacity, storage, graphics and CPU performance.

The report frames the increase as a cost passed through from the AI boom. HBM uses more wafer area per bit than standard DDR5, so a shift toward AI memory can reduce consumer supply by more than a simple one-for-one capacity move. According to the source material, one bit of HBM consumes about three to four times the wafer area of one bit of DDR5.

That does not mean every price increase is permanent or that all suppliers are acting in lockstep. It does mean buyers should treat late-June 2026 memory pricing as a changed market rather than a routine retail spike. The report says relief is not forecast before 2028, and even then prices may settle 30% to 50% above pre-crisis levels.

HBM Is Eating Wafer Output

The source material says HBM now uses about 23% of total DRAM wafer output, up from 19% a year earlier. It also says AI-related demand is on track to absorb about one-fifth of all DRAM capacity in 2026. Those figures help explain why the consumer market is seeing pressure even without a single factory outage or short-term logistics shock.

Past memory cycles often ended when high prices encouraged suppliers to add output, eventually creating oversupply and lower prices. The current report says that pattern is slower this time because new fab capacity takes years, with major expansions not expected to reach meaningful volume until 2027 or 2028.

Several market signals already point to strain. The report cites Micron retiring the Crucial consumer brand, Apple raising prices, Framework DDR5 rising about 50% and DDR4 reaching or exceeding DDR5 pricing per gigabyte. Those details suggest the squeeze is affecting both current-generation and older memory channels.

““Doubled” is the polite version.”

— Thorsten Meyer AI report

Price Relief Remains Unclear

It is not yet clear how high consumer DDR5 prices will go or whether late-2026 retail pricing will stabilize near current levels. The figures in the report are described as point-in-time late-June 2026 data, and component prices can move quickly based on availability, retailer inventory and contract supply.

It is also unclear how much of the increase comes from physical wafer constraints, how much from supplier product mix and how much from channel pricing. The source material says suppliers appear to be managing scarcity and favoring higher-margin products, but that is presented as a market interpretation rather than a confirmed admission by the companies.

Buyers Face Longer Wait

The next milestones are second-half 2026 DRAM pricing, AI accelerator demand forecasts and updates from Samsung, SK Hynix and Micron on capacity plans. Buyers should also watch whether PC makers keep raising prices or absorb part of the added memory cost.

For consumers, the practical takeaway is limited but clear: buy the RAM capacity you genuinely need, avoid panic-buying unused capacity and expect budget PC builds to remain harder to price. The report’s next installment is expected to focus on how HBM demand has changed fab allocation.

Key Questions

Why did RAM prices rise so sharply in 2026?

The source material points to AI-driven demand for HBM, the specialized memory used with AI accelerators. Because HBM and DDR5 rely on overlapping DRAM manufacturing capacity, more HBM output can mean less room for consumer RAM supply.

How much have DDR5 prices increased?

According to the report, a 32GB DDR5 kit that cost about $80 to $120 a year earlier was listed at $374.97 in early June 2026 on Tom’s Hardware’s tracker. A 64GB kit that sat near $150 to $200 in 2025 now often lists at $600 or more.

Will RAM prices fall soon?

The report says broad relief is not expected before 2028, because new capacity takes years and suppliers are prioritizing higher-margin AI memory. That forecast is not guaranteed, and prices may still move with inventory and demand.

Who benefits from the memory squeeze?

DRAM suppliers can earn more by selling HBM into AI systems than by selling standard DDR5 into consumer channels, according to the report. The main customers with stronger allocation are likely to be hyperscalers and AI hardware buyers.

Should PC builders buy RAM now?

The report’s guidance is to buy the capacity you actually need rather than waiting for a quick return to old prices or buying excessive memory out of fear. For many builds, that means balancing RAM, GPU, CPU and storage budgets more carefully than in 2024 or 2025.

Source: Thorsten Meyer AI