The three companies that make the memory in nearly every phone on Earth have reallocated production to AI chips. The cheapest devices, and the people who depend on them, are losing the bid.
Samsung, SK Hynix, and Micron didn't run out of memory for cheap phones. They chose AI, explicitly, publicly, and profitably, and the $80 Tecno Spark Go on a Nairobi retailer's shelf is paying for it.
The three companies control the bulk of the world's production of DRAM and NAND, the working memory and storage in every smartphone, laptop, and PC. Over the past year, all three have shifted wafer capacity — the fabrication slots that turn blank silicon into chips — toward high-bandwidth memory (HBM), the stacked DRAM packages that feed Nvidia's AI accelerators. HBM commands a steep premium over commodity DRAM per wafer, and AI demand has no visible ceiling. Commodity DRAM, the kind that goes into budget phones, has become the residual business.
Micron announced on its investor relations site that it is exiting the Crucial consumer memory brand entirely and redirecting capacity to AI and data-center customers. As Forbes contributor Tom Coughlin, a longtime storage analyst, wrote, the move is a deliberate margin play. Sumit Sadana, Micron's chief business officer, said the "surging demand for memory and storage solutions in the AI-driven data center market" is driving the decision, with a target exit from the consumer market by February 2026. Samsung and SK Hynix have separately warned, according to MSN, that the memory shortage will persist beyond 2027, a long enough horizon to plan around rather than wait out.
For the top of the phone market, the news is a footnote. Apple, Samsung's mobile division, and the premium Android makers absorb component cost increases the way airlines absorb fuel surcharges: quietly, and on the buyer's dime. For the bottom of the market, the $30 to $120 band where Transsion's Tecno and Itel devices dominate across Africa, South Asia, and Southeast Asia, the math does not work. Cheap phones run on thin gross margins, and according to IDC, memory represents 15–20% of the total bill of materials for a mid-range device — and a higher share still for the cheapest devices, where margins are tightest. When memory prices surge, the device either gets more expensive, ships with less storage, or does not ship at all.
IDC's 2026 memory-shortage analysis projects a potential contraction in the global smartphone market alongside rising average selling prices, with the low end bearing the brunt. In IDC's pessimistic downside scenario, the market could contract by as much as 5.2%, with smartphone ASPs rising 6–8% industrywide — and significantly more at the low end, where OEMs have no room to absorb costs. The hit is concentrated at the cheap end, where the consumer is the most price-sensitive and the manufacturer has the least room to absorb a cost shock. David Oks reported in May 2026, citing IDC data, that the IDC projection showed a potential decline of approximately 13% in global smartphone shipments in 2026, with drops above 20% in Africa and the Middle East — what Oks described as the largest single-year decline ever for the industry. (Note: the IDC document accessed shows 2.9% moderate and 5.2% pessimistic downside scenarios; the 13% and >20% figures appear in the Oks blog citing IDC.)
The mechanism is unglamorous and worth naming. Memory fabs run on multi-year capacity plans. As IDC describes the dynamic: "This is a zero-sum game: every wafer allocated to an HBM stack for an Nvidia GPU is a wafer denied to the LPDDR5X module of a mid-range smartphone or the SSD of a consumer laptop." When Samsung, SK Hynix, and Micron collectively decide, as they effectively have, that AI customers are higher priority, the cheap-phone supply pool shrinks by exactly the wafers the AI side consumed. The result is a quiet, mechanical pass-through: the $80 device that sold well last year now costs more to build, ships with half the storage, or disappears from the catalog. The buyer at the bottom of the market does not see a memo about HBM margins. She sees that the phone she was saving for is no longer available.
The framing matters because it determines who is responsible. The cheap phone is not dying because of an impersonal supply chain or a mysterious market force. It is dying because three named firms made a documented, public capital allocation decision, and the people absorbing the cost are the ones with the least ability to switch. As David Oks argued in a recent post, the human consequence lands in markets that technology journalists in San Francisco and London rarely cover: Lagos, Nairobi, Karachi, Dhaka, Phnom Penh. The Next Web has covered the same dynamic from the consumer-impact angle, and the pattern is the same in every regional outlet that bothers to look.
There is a counterargument worth taking seriously. The AI buildout is generating real economic value, and the firms reallocating capacity are responding to price signals, not conspiring. If HBM pays several times more per wafer, the rational move for a publicly traded memory company is to make HBM. The market is doing what markets do. But "the market is doing what markets do" is a description, not a justification, and it leaves unanswered who bears the cost of the reallocation. So far, the answer is: the buyer at the bottom of the global phone market, and the local manufacturer who tried to serve her.
What to watch next: whether any of the three memory makers publicly carves out a guaranteed wafer slice for consumer-grade DRAM, the way some chipmakers have set aside capacity for the auto industry during prior shortages. So far, the public posture is that consumer demand will simply have to clear at the new price. IDC's next quarterly smartphone tracker will be the cleanest read on whether the Africa and Middle East decline materializes as forecast, and whether Chinese fabs — CXMT and YMTC — are quietly backfilling the bottom of the market with domestically produced DRAM and NAND. If they are, the AI reallocation will still be real, and the cheap phone will still be scarcer, but the geography of who loses may shift.