Memory Shortage 2026: Causes and When It Ends
The 2026 memory shortage explained: why AI and HBM demand starved consumer DRAM, how high prices have climbed, and when the RAM shortage is likely to end.
By Capital & Compute
The 2026 memory shortage is the tightest the DRAM market has been in over a decade, and unlike past cycles it was not triggered by a fire, a flood, or a demand collapse. It was a choice. Memory makers redirected wafer capacity toward high-bandwidth memory for AI accelerators, where the margins are several times higher, and left the DDR4 and DDR5 that goes into laptops, desktops, and phones structurally undersupplied. The result: the mainstream DDR4 chip spot price has roughly tripled since late 2025, and the relief date keeps sliding toward 2028.
What is the 2026 memory shortage?
The 2026 memory shortage is a global shortfall of DRAM and NAND flash memory, driven by AI data-center demand pulling manufacturing capacity away from the standard memory used in consumer devices. It began in late 2025, pushed spot prices for mainstream memory chips up roughly threefold within a year, and is expected to persist into 2027 or later. It is severe enough that Wikipedia now tracks it as the 2025 to present global memory supply shortage.
What makes it different from the memory gluts and squeezes of the past decade is the cause. DRAM is famously cyclical: makers overbuild, prices crash, they cut output, prices recover. This time the shortage is not a hangover from overbuilding. It is the direct result of where the three big memory makers, Samsung Electronics, SK Hynix, and Micron Technology, chose to point their wafers.
What caused the memory shortage?
AI accelerators need high-bandwidth memory (HBM), and HBM eats the same silicon wafers that would otherwise become ordinary DDR5. That is the whole story in one sentence, but the mechanics are worth spelling out, because they explain why this shortage is so hard to fix.
Start with the wafer. A memory maker has a fixed amount of cleanroom capacity and a fixed wafer budget each quarter. HBM stacks many DRAM dies vertically and, per gigabyte delivered, consumes substantially more wafer area than a plain DDR5 chip: analysts and reporting such as IEEE Spectrum’s coverage of the DRAM shortage put the penalty at roughly three to four times the wafer capacity per gigabyte. So every wafer a maker moves to HBM removes far more than one wafer’s worth of consumer DRAM from the market.
Now add the margin. HBM sells for far more per wafer than commodity DRAM, and demand from NVIDIA-class AI accelerators has been effectively unlimited. When a supplier can earn several times the revenue per wafer by making HBM, and that HBM is sold out well in advance, the rational move is to starve the low-margin product. Micron said as much when it reported a record-profitable quarter and noted it could fill only 55 to 60% of its core customers’ demand, steering the rest toward higher-value parts (see its fiscal Q3 2026 results).
The demand side is enormous. Gartner estimates AI workloads absorb around 20% of global DRAM wafer capacity in 2026, and that share climbs as each new server generation ships with more memory. This is why I keep calling the shortage a choice rather than an accident: the capacity to make consumer DRAM exists, but it is more profitable to make something else. I go deeper on that argument in why RAM is so expensive in 2026.
How bad is the shortage?
Bad enough that memory went back on allocation, a word the industry had not used seriously since the last major crunch. Here is the scale in numbers.
The raw chip is where the squeeze shows first. The mainstream DDR4 1Gx8 3200 spot price, the figure TrendForce quotes in dollars, went from about $12 per chip in November 2025 to about $37 by July 2026. That is the number our DRAM price tracker follows week to week, rebased into a RAM price index that now sits near 305 against a November 2025 base of 100.
DDR5 rose just as hard. TrendForce reported DDR5 spot prices up 307% from September to November 2025 alone. And the pain is not confined to chips: Gartner expects combined DRAM and SSD prices to rise about 130% across 2026, lifting average PC prices by roughly 17% and cutting global PC and smartphone shipments.
Supply is the reason none of this snaps back fast. IDC projects 2026 DRAM bit-supply growth of only about 16% year on year, with NAND flash at about 17%, both below historical norms, because so much of the new output is HBM rather than commodity DRAM. Counterpoint Research reckons the industry needs sustained 20%-plus supply increases to start bringing prices down, a bar 2026 does not clear.
| Item | Value |
|---|---|
| DRAM supply growth, 2026 (IDC) | 16% |
| NAND supply growth, 2026 (IDC) | 17% |
| Growth needed for relief (Counterpoint) | 20% |
How the shortage built: a timeline
The shortage did not arrive overnight. It built over about a year, each step making the next harder to reverse.
Late 2025
Spot DRAM starts climbing
Makers divert wafer capacity to AI memory. The loose-chip spot market is the first place the squeeze appears, with DDR5 spot up 307% from September to November 2025 (TrendForce).
Early 2026
Memory goes back on allocation
DDR4 and DDR5 contract prices jump quarter over quarter. PC makers including Dell, HP, and Lenovo warn of double-digit price increases; smartphone brands begin downgrading base memory specs.
Q1 to Q2 2026
AI absorbs about a fifth of DRAM wafers
Gartner estimates AI workloads take roughly 20% of global DRAM wafer capacity in 2026. Samsung and SK Hynix tell investors they will not expand commodity DRAM aggressively.
Mid 2026
The chip price roughly triples
The mainstream DDR4 1Gx8 spot chip reaches about $37, up from about $12 the previous November (TrendForce), and Micron reports it can meet only 55 to 60% of core-customer demand.
2027 to 2028
New supply slowly arrives
A new memory fab takes 18 to 24 months to build. Micron does not expect its Idaho fab at volume until 2027, and Intel CEO Lip-Bu Tan says the industry told him there is no relief until 2028.
Who is hit hardest?
Not everyone feels this equally, and the order matters.
Consumers and PC builders are near the back of the queue. Older DDR4 is the worst-hit segment, because fabs are winding it down to free capacity for DDR5 and HBM, so legacy-memory spot prices have in some weeks run above comparable DDR5. If you are buying a memory kit today, you are paying the shortage in full. Our guide to how much RAM you actually need to run a local LLM is a useful gut-check before you overspend on capacity you will not use.
PC and phone makers eat the cost or pass it on. Several major OEMs warned of 15 to 20% price increases for 2026, and some are quietly shipping less memory for the same money rather than raise the sticker price, a move I unpack in what Apple’s price hike means for local AI.
Data centers and AI buyers get served first. They are the reason for the shortage, and they hold the contracts that clear before consumer parts. When relief does arrive, it reaches hyperscalers before it reaches a retail memory aisle.
When is the RAM shortage going to end?
The RAM shortage is expected to ease no earlier than late 2027, and parts of the industry point to 2028. There is no firm date, because the end depends on two things that move slowly: new fab capacity coming online, and AI memory demand cooling enough to free existing capacity for consumer DRAM.
The physical clock is unforgiving. A new memory fab takes 18 to 24 months to build and longer to yield well, so capacity decided today lands in 2027 at the earliest. Counterpoint Research has pointed to Q4 2027 as the earliest realistic inflection point. Gartner sees no meaningful relief before late 2027. And Intel CEO Lip-Bu Tan has said flatly that the memory industry told him there is no relief until 2028.
Even that is only the start of relief, not a return to old prices. Prices are sticky on the way down, makers have relearned that scarcity pays, and the baseline is likely to settle above the pre-AI floor. For the full buy-now-or-wait breakdown by buyer type, see our 2026 RAM price forecast.
Bottom line
The 2026 memory shortage is what happens when the most profitable customer in the industry, AI, can absorb everything the fabs can make. Consumer memory did not get scarcer because the world stopped making it. It got scarcer because making it stopped being the best use of a wafer. That framing matters, because it tells you the shortage will not break on a supply surprise or a demand dip. It breaks when the economics flip, and the people who make the chips are being paid very well to keep them flipped the other way for now.
Frequently asked questions
- What caused the 2026 memory shortage?
- AI demand. Samsung, SK Hynix, and Micron redirected wafer capacity to high-bandwidth memory (HBM) for AI accelerators, which earns far more per wafer than consumer DDR4 and DDR5. HBM consumes roughly three to four times the wafer area per gigabyte, so shifting capacity to it starved the mainstream memory market and drove prices up sharply from late 2025.
- When is the RAM shortage going to end?
- No earlier than late 2027, and possibly 2028. New memory fabs take 18 to 24 months to build, Counterpoint Research points to Q4 2027 as the earliest inflection, Gartner sees no meaningful relief before late 2027, and Intel CEO Lip-Bu Tan says the industry told him 2028. Consumer parts ease after data-center and HBM demand is met.
- How much have memory prices risen in the shortage?
- The mainstream DDR4 1Gx8 spot chip roughly tripled, from about $12 in November 2025 to about $37 by July 2026 (TrendForce), and DDR5 spot rose 307% from September to November 2025. Gartner expects combined DRAM and SSD prices to climb about 130% across 2026, raising average PC prices by roughly 17%.
- Is the memory shortage caused by AI?
- Yes, primarily. AI data-center demand for high-bandwidth memory pulled manufacturing capacity away from standard DRAM. Gartner estimates AI workloads absorb about 20% of global DRAM wafer capacity in 2026, and because HBM is more profitable per wafer, makers prioritized it over consumer memory.
- Will RAM prices go back to 2024 levels?
- Unlikely. Even optimistic forecasts describe a gentle easing toward 2028 rather than a return to the pre-AI floor. Prices are sticky on the way down, makers are rationing supply to protect record margins, and structural AI demand keeps the baseline elevated.
Sources
- U.S. Bureau of Labor Statistics and TrendForce, via the Capital & Compute DRAM price tracker. Primary/independent; DDR4 1Gx8 spot chip price and RAM price index. Verified 2026-07-12.
- TrendForce (2026). DRAM Spot Price (mainstream DDR4 1Gx8 3200MT/s chip). Primary; chip spot price. Verified 2026-07-12. trendforce.com
- Gartner (2026). Gartner Says Surging Memory Costs Will Reduce Global PC and Smartphone Shipments in 2026 (press release, February 26 2026). Independent forecast; ~130% price rise, ~17% PC price impact, ~20% AI wafer share. Verified 2026-07-12. gartner.com
- IDC (2025). Global Memory Shortage Crisis: Market Analysis and the Potential Impact on the Smartphone and PC Markets in 2026. Independent; ~16% DRAM and ~17% NAND bit-supply growth, below historical norms. Verified 2026-07-12. idc.com
- Counterpoint Research (2026). Advanced Memory Prices Likely to Double as DRAM Crunch Spreads on NVIDIA Pivot. Independent; Q4 2027 earliest inflection, 20%-plus supply growth needed for relief. Verified 2026-07-12. counterpointresearch.com
- Micron Technology (2026). Reports Results for the Third Quarter of Fiscal 2026 (SEC Form 8-K exhibit, June 24 2026). Primary; record quarter, 55-60% demand fulfilment. Verified 2026-07-12. sec.gov
- Bloomberg (2026). Intel CEO Says There’s ‘No Relief’ on Memory Shortage Until 2028. Secondary; Intel CEO statement. Verified 2026-07-12. bloomberg.com
- IEEE Spectrum (2026). AI Boom Fuels DRAM Shortage and Price Surge. Independent; HBM wafer-penalty mechanism. Verified 2026-07-12. spectrum.ieee.org
- Wikipedia. 2025 to present global memory supply shortage. Secondary; used for framing only. Verified 2026-07-12. en.wikipedia.org