I watched a Surface laptop get pushed back from a cart when the sticker jumped. You feel that pinch — the sudden math of memory costs. I’ve been following the numbers, and the squeeze isn’t letting go.
On the showroom floor, price tags have teeth.
Microsoft blamed “recent increases in memory and component costs” when it raised some Surface models by as much as $500 (€460). Raspberry Pi bumped prices by up to $150 (€138) this month. Meta pointed at memory shortages when it hiked the Quest 3 by $100 (€92). Apple’s low-priced MacBook Neo remains an outlier, and yes, it’s selling fast.
Why are RAM prices rising?
You’re seeing symptom and cause at once: demand for memory has exploded because AI and cloud workloads gobble more RAM per rack than consumer laptops ever did. At the same time, production ramps lag. Nikkei Asia warns production needs about 12% growth per year through 2027 to meet demand — we’re only seeing roughly 7.5% growth.
Inside fabs, the new lines hum but not fast enough.
At Samsung’s Pyeongtaek complex, crews are building capacity that matters — but “full-scale mass production” for mainstream DRAM isn’t expected until next year. HBM, the high-bandwidth memory that AI accelerators crave, is now penciled in for wider rollout only around 2028.
The memory market is a pressure cooker: demand keeps rising, yet adding capacity takes years and billions in capital.
When will RAM prices fall?
Realistically, don’t expect broad price relief before late 2027, and even that is optimistic. Counterpoint Research’s MS Hwang tells Nikkei that supply and demand probably won’t normalize until 2028. Until HBM and DRAM capacity actually hit the floor, AI clusters and hyperscalers will bid up what’s available.
On the factory gate, labor tensions are loud.
Samsung recently asked a court to block labor actions at Pyeongtaek, according to Reuters — the union called that move a “declaration of war.” Those disputes can delay startups, create quality hiccups, and tighten output when every wafer matters.
At the AI compute table, appetite outstrips the buffet.
There’s roughly a 40 percent gap between current supply and demand, Nikkei reports. AI-related purchases are the main driver; geopolitical shocks that push up energy and material costs are making expansion more expensive.
SK Hynix’s Cheongju facility will be the primary HBM source through the rest of 2026 while Samsung focuses more on traditional DRAM. Samsung has even allowed competitors a toehold in HBM while it prioritized consumer-facing chips, and that has consequences for server builders and cloud providers.
The supply chain is a leaking dam: new capacity plugs holes slowly while demand keeps pouring in.
How does the RAM shortage affect device prices and availability?
Higher module prices ripple through PCs, gaming consoles, VR headsets, and single-board computers. OEMs either absorb costs and eat margins, or pass them to you — as we’ve just seen. For enterprise buyers, the shortage raises operating costs and pushes purchase timing decisions into strategic territory.
I’ve tracked semiconductor cycles long enough to know one thing: markets correct, but they don’t owe you a quick fix. If Samsung, SK Hynix, Microsoft, Meta, and the cloud giants are all angling for the same scarce pieces of silicon, who pays when supply finally loosens—your wallet or their margins?