DRAM prices forecast to rise 400% as AI redirects memory supply

By Rebecca Stone

JP Morgan projects that DRAM prices will have risen 400% since 2024 by the end of 2026, as AI data centers absorb memory faster than suppliers can manufacture it. The forecast turns a trade in high-bandwidth memory into a broader consumer technology cost problem.

The same JP Morgan report says the Consumer Price Index for software and accessories has increased by nearly a quarter over the past two years. That spread matters because factory decisions driven by AI memory demand now set costs for DRAM, GPUs, consoles and laptops.

Tom’s Hardware’s chips tracker lists Nvidia’s most powerful consumer GPU, the GeForce RTX 5090, at $4,920 above the $1,999 suggested price set when it launched in January 2025. The report describes GPU price increases that have outpaced economy-wide inflation as “chipflation.”

In Japan, some gaming GPUs have risen roughly 41% above their American counterparts, according to the report. The Economist’s 2026 report found similar price increases for video game consoles and laptops, which it linked to an uncharacteristic fall in PC shipments.

Resellers have used the price gap. GameStop markets the Sony PlayStation 5 Pro for roughly $500 above MSRP, extending the same markup pattern from graphics cards into game hardware.

The report says component prices are rising as supplies dwindle because chipmakers continue to shift resources toward AI. Consumers now absorb those decisions in products they did not buy for AI workloads, from standalone GPUs to complete PCs.

Why is a GPU shortage also a memory shortage?

Peaks and valleys in chip supply are routine, but the industry’s shift from DDR4 to DDR5 had already delayed output while manufacturers changed production lines, according to the report. The rapid data center buildout then worsened the shortage. At the core, the report describes a memory problem, not merely a shortage of finished graphics cards.

AI developers need HBM, a specialized form of DRAM with faster connections and more bandwidth, to train and run their models. The report says demand for those chips far outpaces current market supply.

HBM competes for the same wafer capacity as traditional DRAM and other memory products. Suppliers have shifted production toward HBM because of both demand and profit margins, selling it for roughly 30% more than traditional DRAM. Every reallocated line therefore narrows the supply available for conventional memory.

The transition adds another constraint. Suppliers are moving to HBM4 and HBM4E, so factory changes for the newer generations further limit immediate supply.

A Samsung executive told Reuters in September that HBM could account for nearly 30% of total DRAM production in 2027, up from 20% in 2026. Some forecasts cited in the report put that share at almost 40% by 2029.

The report projects that AI data centers will consume roughly 70% of global DRAM supplies this year. With AI taking the largest share of output, consumer markets depend on what suppliers can spare after meeting data center demand.

Who controls the memory market?

Three companies — Micron Technologies, Samsung Electronics and SK Hynix — account for roughly 90% of global DRAM production. The report says all three expect long-term shortages despite major capacity expansions.

Suppliers have continued to tilt capacity toward data centers at the expense of consumer GPUs. Micron, for example, ceased direct-to-consumer sales last year to dedicate resources to AI production, according to the report. The move contributed to continued RAM price increases.

That concentration leaves a handful of producers deciding how much memory reaches gaming, PC and other consumer markets. Their shift toward higher-margin AI orders transmits the data center buildout directly into consumer product pricing.

How long will the shortage last?

In a recent earnings call, Micron CEO Sanjay Mehrotra said supply would be “much tighter in calendar 2027 and 2028” than in 2026. His forecast extends the current squeeze across the next two production cycles.

SK Hynix CEO Kwak Noh-jung told Reuters in July: “we forecast that next year will be the worst year in the industry's history from the supply perspective.”

SK Hynix chairman Chey Tae-won said at a Nvidia event that it will take “at least four to five years” to build adequate wafer supply. Reuters reported that production would remain 20% behind demand until at least 2030. That timeline extends the shortage into the 2030s.

The report says many chip producers have already sold much of their 2027 supply and now focus sales on 2028 and beyond. Several major suppliers have sold most of next year’s production capacity, making an immediate release to consumer buyers unlikely.

• Micron has reportedly sold 75% of its 2027 HBM output. Mehrotra said via Yahoo Finance that those sales carried prices “much higher than 2026 prices.”
• Samsung has sold almost 80% of its 2027 memory output, according to KB Securities. Roughly 70% of that output is tied up in five-year contracts.

KB’s figures and Micron’s sales figure point in the same direction: the market for 2027 memory is being allocated before consumer buyers reach the market. The shortage therefore has a procurement dimension as well as a production one.

Could new supply break the squeeze?

Acer CEO Jason Chen told DigiTimes that prices could plateau next year, citing a potential influx of cheaper Chinese alternatives. That outcome would require new supply to arrive quickly enough to offset HBM allocations and factory transitions.

The report also notes skepticism about whether AI’s growth will generate enough demand to justify suppliers’ massive buildout. A surplus would pressure prices, but current contracts and production decisions leave the market committed to the shortage through at least 2027.

In October 2026, one South Korean publication sent memory stocks tumbling after quoting experts who predicted a “flood of new supply” in 2029 would transform the market “into a red ocean,” according to CNBC’s translation.

Nvidia head Jensen Huang dismissed the possibility of a supply surplus when he told CNBC: “One of the things that's really great about memory is that you can always use more of it.”

The competing paths leave two possible inflection points: cheaper Chinese memory could slow price increases next year, while the 2029 “flood of new supply” cited by those experts could turn shortage into surplus. Until then, AI memory demand will remain the dominant force setting consumer technology prices.