For years, the price of memory chips fell so reliably that each new device could deliver more power without costing more. That era is over.
When Apple takes the stage this week, analysts expect it to debut the iPhone 18 Pro and Pro Max alongside its first foldable — while holding the standard iPhone 18 and a refreshed Air until spring 2027. The Wall Street Journal, citing TechInsights estimates, calculates the iPhone 18 Pro could start at $1,299 — $200 above the iPhone 17 Pro. Apple has already raised prices across its Mac and iPad lines; the company's then-CEO Tim Cook described memory pricing as a '100-year flood', citing 'exponential increases' in component costs.
Behind the price tags sits a shortage years in the making. AI data centers have an insatiable appetite for high-bandwidth memory (HBM), the stacked DRAM that feeds accelerators from Nvidia, AMD and others. HBM is far more lucrative than conventional DRAM, but it consumes roughly three times the wafer capacity for the same memory volume. With Samsung, SK Hynix and Micron controlling about 90 percent of the market, the incentives have shifted decisively: instead of betting on how many phones consumers will buy, the big three are signing multi-year deals with AI giants willing to commit capacity years in advance.
'The three big memory guys are just basically allocating the capacity that they have to these companies,' says David Naranjo of Counterpoint. The figures are stark: Counterpoint estimates smartphone DRAM prices rose 56 percent quarter over quarter in Q1 2026 and another 83 percent in Q2. A 16GB DRAM package that cost roughly $42 in mid-2025 went for about $181 a year later — an increase of more than 300 percent. The financial results show who is winning: SK Hynix posted a record 76 percent operating margin, Micron's adjusted gross margin hit 85 percent, and Samsung's semiconductor profit jumped roughly 250-fold year over year.
Consumers feel it across the board. Microsoft raised Xbox prices by $100 to $150 (some models now cost up to $300 more than at launch) and added $500 to some Surface Pro models; Meta put $100 on the Quest 3. Budget-constrained makers are shipping fewer devices, trimming memory or pivoting to premium models — meaning revenue can hold steady even as unit sales fall.
Supply is coming, but slowly. Micron poured the first concrete for its New York mega-fab in July after more than two years of permitting; at 2.4 million square feet of cleanroom space it will be the largest semiconductor manufacturing site in US history, yet the company does not expect meaningful output until 2030. Its Idaho fab should begin wafer output in mid-2027. SK Hynix is investing 600 trillion won in its Yongin cluster and, with Samsung, plans four new memory plants in southwestern South Korea — a combined commitment of roughly $588 billion. A potential fourth player, China's CXMT, is expanding but trails on technology, and Apple's testing of its chips would require US government approval.
Analysts at Counterpoint and IDC do not expect supply to catch up with demand before late 2027 or early 2028 — if then. '2027 is pretty much locked and loaded,' says Naranjo. Price increases may slow, but they will stabilize at a new, higher plateau. The new normal is 'at least triple what they used to be', says IDC's Nabila Popal.
The only scenario that changes the math quickly is an AI spending bust. For now, manufacturers are betting on the boom — and consumers are paying for it.
