Benchmarks & Evaluations
Memory Prices Double as AI Eats the World's RAM
Data centers will consume 70% of global memory production this year, leaving PC builders and enterprise buyers scrambling for increasingly expensive scraps.
Data centers will consume 70% of global memory production this year, leaving PC builders and enterprise buyers scrambling for increasingly expensive scraps.
If you tried to build a PC this week, you probably noticed something alarming: a 32GB DDR5 kit that cost $150 in September now runs $240. By March, it could hit $500.
This is not a temporary blip or seasonal fluctuation. According to new data from TrendForce, we're witnessing what analysts call a "permanent reallocation" of global memory capacity—away from consumer electronics and toward AI infrastructure. The numbers are stark: conventional DRAM prices jumped 63% between September and December 2025, with another 55-60% increase forecast for Q1 2026.
The culprit appears to be High Bandwidth Memory, or HBM—the specialized RAM that powers AI training clusters. Manufacturing HBM requires three times the wafer capacity of standard DRAM, according to AIwire's analysis of production bottlenecks. Every wafer allocated to HBM production means three wafers' worth of regular memory that won't reach the market.
"Data centers will absorb 70% of global memory production in 2026," Tom's Hardware reports, citing industry forecasts. The remaining 30% must somehow serve PCs, smartphones, cars, appliances, and industrial equipment.
The shortage hits legacy memory particularly hard. DDR4, which still powers most enterprise servers and networking equipment, saw prices spike 50% in Q1 2026 alone, according to TrendForce. Suppliers are abandoning DDR4 production entirely to chase higher margins in AI-centric memory. Enterprise buyers are now "aggressively sourcing remaining DDR4 stocks," creating panic buying that further inflates prices.
Get the latest model rankings, product launches, and evaluation insights delivered to your inbox.
NAND flash storage faces similar pressure, with prices up 33-38% this quarter. The surge affects everything from SSDs to the memory chips in washing machines.
For video creators and AI professionals, the implications are immediate. Building a workstation with 64GB or 128GB of RAM—standard for 4K video editing or local AI model testing—could soon cost more than the GPU itself. Server costs are climbing too: The Register notes that OEMs like Dell and HPE are expected to pass component price increases directly to customers as inventory buffers run dry.
The timing couldn't be worse for the creator economy. Just as AI tools democratize video production, the hardware to run them efficiently becomes prohibitively expensive. A creator who might have upgraded their system for better Stable Diffusion performance now faces a memory bill that exceeds their entire previous build budget.
This echoes the GPU shortage of 2020-2021, but with a crucial difference: crypto miners eventually sold their cards back into the market when profits declined. AI data centers, by contrast, represent permanent demand. Those memory chips aren't coming back.
PC builders should buy RAM now, as prices will likely peak in March 2026. Enterprise IT departments face steep server refresh costs through the year. Video creators may need to lean on cloud rendering as local workstation costs soar, while legacy DDR4 systems become unexpectedly valuable as replacement parts vanish. Consumer electronics prices will rise as manufacturers pass through memory costs.
Memory manufacturers show no signs of expanding capacity for consumer markets. They're chasing the margins where a single AI cluster order can exceed an entire quarter's worth of gaming PC sales. The question is not whether this shortage will ease, but how much of the computing world will be priced out before equilibrium returns.