Nothing’s decision to scrap its next budget smartphone didn’t make big headlines. Yet it marked a turning point. For months executives at the London-based company had watched component costs climb. Memory prices in particular refused to stabilize. In the end the math no longer worked. A co-founder explained the team could not deliver a device that represented a genuine step forward at an accessible price. The CMF Phone 3 Pro simply ceased to exist. The Next Web reported the cancellation this week alongside a blunt assessment from industry watchers: AI’s demand for memory has moved beyond raising prices. It has begun to erase products.
Even Apple feels the strain. Tim Cook told The Wall Street Journal in an exclusive interview that price increases on Apple products are “unavoidable.” He added, “We’re doing our best to mitigate the huge increases that are being passed to us, and we’ve been trying to shield our customers from the increases, but the situation has become unsustainable.” Cook described the surge as a “hundred-year flood.” He has spent more than four decades in technology. Nothing like this has crossed his desk before.
The numbers tell a stark story. LPDDR5X memory used in premium iPhones rose 230 percent during 2025. JPMorgan analysts project memory could represent 45 percent of an iPhone’s component cost by 2027, up from roughly 10 percent today. Entry-level Android phones have seen memory’s share of bill-of-materials jump from 15 percent to 50 percent. Some budget handsets now cost 50 percent more than they did a year ago. Global smartphone shipments could fall 15 percent in 2026 according to CCS Insight. That would mark the first sustained contraction in years.
Supply reallocations explain much of the pressure. Samsung, SK Hynix and Micron have shifted production lines toward high-bandwidth memory. The specialized chips deliver several times the revenue per wafer compared with standard DRAM destined for phones. HBM capacity is sold out well into 2027. Data centers building out AI infrastructure snapped up the available supply. Nvidia locked in multi-year deals. Consumer electronics suddenly found themselves at the back of the queue.
By 2027 Nvidia’s upcoming Vera Rubin platform is projected to consume more LPDDR than Apple and Samsung combined. That single forecast, cited across multiple analyses, captures the scale of displacement. AI training and inference do not merely use memory. They hoard it. High-bandwidth variants now claim 20 percent of total wafer production by the end of 2026, up from 2 percent recently. Another 3 percent of high-density DDR heads to AI servers. The remaining pool shrinks fast.
Apple once dictated terms to memory suppliers. Long-term contracts smoothed costs for years at a time. Those agreements began expiring in January 2026. Suppliers now offer only quarterly deals. To secure chips at all, Apple agreed to pay Samsung a 100 percent premium on LPDDR5X destined for iPhones. The company has already raised the price of its cheapest Mac mini and delayed the standard iPhone 18 model until spring 2027. Further hikes look imminent. Cook signaled the firm would dip into its balance sheet to help expand capacity but drew a firm line at building its own memory factories. “We can’t do everything,” he said. “We know what we’re good at.”
The pain spreads far beyond Cupertino. Transsion, the dominant brand across Africa, saw profits drop 54 percent last year and cut shipments 40 percent. Oppo, Vivo and Xiaomi have lowered 2026 targets by double-digit percentages. In India the sub-$100 phone segment collapsed 59 percent in the first quarter. Across much of the global south, where the majority of handsets sell for under $200, the category is vanishing. Memory inflation has done what trade barriers and currency swings could not. It has priced out hundreds of millions of potential buyers.
But the crunch runs deeper than phones. Personal computers, cars, medical devices and networking gear all compete for the same underlying components. Morgan Stanley estimates production capacity will grow only 30 percent by 2027 while demand from AI leaves consumer markets 15 percent short. DRAM prices jumped 90 percent in a single recent quarter. DDR5 tags in some European markets have quadrupled over twelve months. The retail SSD market has essentially disappeared.
Industry leaders outside Apple have begun to speak openly. Micron’s executives called the shortage “unprecedented” and redirected output away from consumer DRAM entirely. SK Hynix watched HBM revenue quadruple in 2024; the category now exceeds 40 percent of its total DRAM business. Memory makers as a group are on track to generate $70 billion in profit this year and more than double that figure in 2026. For decades the sector operated on thin margins. AI changed the economics overnight.
Efforts to ease the bottleneck have so far produced modest results. Google researchers detailed TurboQuant, a compression technique that shrinks the key-value cache used during inference by at least six times without sacrificing accuracy. The approach targets runtime memory rather than training requirements. It will not free up enough chips to refill smartphone assembly lines. Other labs explore compute-in-memory architectures and new interconnects such as CXL. These remain years from volume production. In the meantime data-center operators continue to outbid everyone else.
Analysts warn the reallocation could last through 2027 and beyond. AI build-outs show no sign of slowing. Hyperscalers from Google to Meta to Amazon have poured hundreds of billions into GPU clusters that each require vast amounts of accompanying memory. The same chips that once powered affordable gadgets now keep large language models responsive. Consumers ultimately foot the bill through higher device prices, fewer choices or both.
Nothing’s canceled phone stands as an early casualty. Apple’s coming price adjustments may prove more visible. Yet the pattern points to something larger. An entire tier of consumer electronics is being repriced or retired to feed the AI boom. The industry once debated whether artificial intelligence would eliminate jobs. It is already deciding which gadgets never reach store shelves. Short, sharp shocks to supply chains have happened before. This one feels structural. And the data-center hunger shows little sign of abating anytime soon.
Recent coverage reinforces the trend. A New York Times report from earlier this year detailed how AI companies buying up memory chips caused prices for components in laptops and smartphones to soar, with custom PC builders passing on tripled costs to customers. Another piece from The Wall Street Journal warned the global memory-chip shortage will cost everyone as data centers consume more than 70 percent of high-end supply in 2026. On X, real-time discussion echoes the same alarm. Posts from market analysts and tech accounts this week highlighted Apple’s weakened negotiating power and Micron’s strong positioning amid the crunch.
WebProNews is an iEntry Publication