The price of remembering just went up. Not the human kind — the silicon kind. DRAM, the volatile memory chips that let your phone juggle apps and your laptop run spreadsheets, is getting significantly more expensive. And the cost increases are already landing on consumer price tags.
Samsung’s newest Galaxy S25 lineup and Microsoft’s Surface laptops are among the first high-profile casualties. According to TechRadar, the memory components inside these devices have surged in cost, and manufacturers are passing that burden downstream — to you. The pattern isn’t new. But the scale of this latest cycle, and its timing alongside rising demand for AI-capable hardware, makes it particularly painful.
DRAM pricing is cyclical by nature. The market swings between glut and shortage with a regularity that would be almost comical if billions of dollars weren’t at stake. The current upswing began in earnest in late 2024, after a prolonged downturn that saw memory makers slash production and delay capital expenditure to stanch the bleeding from oversupply. Samsung, SK Hynix, and Micron — the three companies that collectively control roughly 95% of the global DRAM market — all cut output. The strategy worked. Supply tightened. Prices recovered. And then they kept climbing.
Contract prices for DDR5 modules — the current standard for flagship phones and modern PCs — have risen more than 20% over the past two quarters, according to market research firm TrendForce. DDR4, still widely used in budget and mid-range devices, has seen similar increases. Server-grade memory, driven by insatiable demand from hyperscale data centers building out AI infrastructure, has climbed even faster.
For Samsung, the timing is awkward. The Galaxy S25 Ultra ships with 12GB of RAM, the same as its predecessor, but the cost of that memory to Samsung’s mobile division has risen materially. Samsung’s semiconductor arm profits from higher memory prices, of course. But its consumer electronics division absorbs them. It’s a peculiar form of corporate self-taxation. The company hasn’t publicly broken out how much of the Galaxy S25’s price reflects higher component costs, but teardown analyses from firms like Counterpoint Research suggest memory now represents a larger share of the bill of materials than it did a year ago.
Microsoft faces a different version of the same problem. Its Surface Pro and Surface Laptop lines compete in a segment where consumers are especially price-sensitive. A $100 or $200 increase on a $1,300 device doesn’t just dent demand — it can push buyers toward competitors or, worse, toward simply keeping their existing machines another year. The new Surface devices equipped with Qualcomm’s Snapdragon X Elite processors require DDR5, and there’s no downgrading to cheaper memory without sacrificing the performance and efficiency gains those chips promise.
So who’s actually benefiting?
The memory makers themselves, obviously. SK Hynix reported record quarterly profits in early 2025, driven largely by sales of High Bandwidth Memory (HBM) chips to Nvidia and other AI accelerator companies. Samsung’s semiconductor division has returned to profitability after a brutal 2023. Micron, the lone American player in the DRAM oligopoly, has seen its stock price recover sharply from its 2023 lows.
But the gains are unevenly distributed even among the winners. HBM — the specialized, vertically stacked memory used in AI training chips — commands margins that commodity DRAM can only dream of. SK Hynix has been the primary beneficiary, having locked in early supply agreements with Nvidia for its HBM3E chips. Samsung, which stumbled on HBM quality issues in 2024, has been playing catch-up. The result is a two-tier memory market: sky-high margins on AI memory, and merely elevated margins on the conventional DRAM that goes into phones and PCs.
The ripple effects extend well beyond Samsung and Microsoft. Every major PC OEM — Dell, HP, Lenovo, Asus, Acer — sources DRAM from the same three suppliers. Every smartphone maker, from Apple to Xiaomi, faces the same pricing pressure. Apple, with its enormous purchasing scale and long-term supply agreements, is better positioned to absorb or negotiate around price increases. Smaller manufacturers have less leverage. Some will eat the margin hit. Others won’t.
There’s an AI dimension to all of this that’s impossible to ignore. The explosion in demand for large language models, image generators, and enterprise AI tools has created a voracious appetite for memory at the data center level. Training a single frontier AI model can require thousands of GPUs, each paired with expensive HBM modules. Inference — running those models once they’re trained — requires less exotic memory but still enormous quantities of it. This demand competes directly with the memory that would otherwise flow into consumer devices. Fabs have finite capacity. When a chip manufacturer can sell HBM at a premium to a hyperscaler, the economic incentive to produce commodity DDR5 for laptops diminishes.
TrendForce has projected that DRAM prices will continue rising through at least the third quarter of 2025, though the rate of increase may slow. Much depends on whether Samsung and Micron can ramp HBM production fast enough to satisfy AI demand without further constraining conventional DRAM supply. Capital expenditure plans announced by all three major memory makers suggest significant investment in HBM capacity, but new fab capacity takes 18 to 24 months to come online. Relief, if it comes, won’t be immediate.
The situation is compounded by geopolitical factors. Export restrictions on advanced semiconductor equipment to China, imposed by the United States, the Netherlands, and Japan, have limited China’s ability to produce cutting-edge memory domestically. Chinese firms like CSET and ChangXin Memory Technologies (CXMT) have made progress on older-generation DRAM, but they remain largely shut out of DDR5 and HBM production. This keeps the supply base concentrated among Samsung, SK Hynix, and Micron — and keeps pricing power firmly in their hands.
For consumers, the practical impact is straightforward. That laptop you’ve been eyeing? It might cost $50 to $150 more than an equivalent model did a year ago, even if the processor, display, and storage haven’t changed. That phone upgrade? Same story. Manufacturers can try to obscure the increases through creative bundling, promotional pricing, or spec adjustments — offering 8GB instead of 12GB at the same price point, for instance — but the underlying economics are clear.
Industry analysts have started drawing comparisons to the DRAM price-fixing era of the early 2000s, when Samsung, Hynix, and other manufacturers were found to have colluded on pricing. There’s no evidence of collusion this time. The current dynamics are driven by legitimate supply-demand imbalances and rational production decisions by the oligopoly. But the structural outcome — a market with three dominant suppliers and limited pricing transparency — naturally invites scrutiny.
And then there’s the question of how long the AI boom sustains memory demand at current levels. If AI spending plateaus or if hyperscalers pull back on data center construction, the memory market could swing back to oversupply remarkably fast. The industry has seen this movie before. The 2022-2023 downturn was triggered in part by pandemic-era overbuilding, and it devastated memory maker balance sheets. The cyclical nature of the business means today’s shortage is tomorrow’s glut. Or not. Nobody knows with certainty.
What is certain: right now, in mid-2025, the cost of memory is rising, and it’s showing up in the devices people buy every day. Samsung’s Galaxy phones. Microsoft’s Surface laptops. And virtually everything else with a power button and a need to think. The memory tax is real. And for the foreseeable future, consumers are the ones paying it.
WebProNews is an iEntry Publication