The cheap smartphone is dying, and AI is holding the knife. That's the argument in a piece climbing Hacker News this week: the memory shortage pushing up DRAM and NAND prices isn't a blip but a repricing — and the budget end of consumer electronics is what gets cut first.
Sit with that. The AI buildout's first tax on ordinary people isn't job loss or deepfakes. It's the memory in a $200 phone. The same chips that serve frontier models sit in your laptop, your console, your car — and a data center outbids you for every one.
The same chips that serve frontier models sit in your laptop, your console, your car — and a data center outbids you for every one.
Memory is fungible in a way compute isn't. A hyperscaler ordering DRAM by the trainload sets the price for the kid buying a single handset. When demand steps up by an order of magnitude and new fab capacity takes years to arrive, the price curve doesn't wiggle — it shifts. A flagship absorbs a forty-dollar bump inside a thousand-dollar bill of materials. The budget phone can't. It gets worse, or it disappears.
The rebuttal writes itself: memory is cyclical, gluts always follow shortages, this corrects by 2028. True — for a normal cycle, where demand wobbles around a flat line. This isn't that. Inference at planetary scale isn't a spike that clears; it's a new floor under memory demand. Fabs will catch up to where demand was, not to where it's racing.
So watch the bottom of the market, not the top. AI's real cost won't show up in a model card. It'll show up in the phones that quietly stop existing.