The rise of artificial intelligence is having a less obvious consequence than the proliferation of new chatbots or data centers: buying a cheap mobile phone is becoming more difficult and more expensive. The reason lies mainly in two components present in practically any smartphone, DRAM memory, used to run applications and keep processes open, and NAND memory, where photographs, videos, applications, and the operating system itself are stored. Both are experiencing a strong price escalation as major chip manufacturers concentrate a growing part of their investments and productive capacity in the artificial intelligence server business.
The GSMA, the international association representing the mobile industry, warns that memory prices more than doubled between the third quarter of 2025 and the first quarter of 2026 and rose again by 80% to 90% during the second quarter of this year. TrendForce has also recorded extraordinary movements: in the second quarter, it estimated increases of at least 70%-75% for certain LPDDR4X memories commonly used in budget devices and 78%-83% for LPDDR5X.
What ChatGPT has to do with a mobile's memory
Artificial intelligence does not use exactly the same memories as a 150-euro smartphone, but both markets are part of an industry with limited manufacturing capacity. Large AI servers require enormous amounts of HBM, DRAM memory for servers, and high-performance NAND storage, higher value products with greater margins for manufacturers like Samsung, SK Hynix, or Micron. In light of the explosion of orders from data centers, these companies have been allocating more resources and capacity to those products.
The effect ultimately runs through the entire chain. If a factory dedicates more capacity, investment, and production nodes to HBM or server memory, there is less margin for certain components intended for computers and phones. TrendForce points out that suppliers continue to prioritize AI-related applications when distributing available capacity, which also keeps the supply of mobile memory tense. In the case of NAND storage, manufacturers are prioritizing enterprise SSDs used in data centers.
It is not about a data center literally buying the memory intended for a specific model of Xiaomi, Samsung, or Motorola, but rather a competition for industrial capacity, investment, and raw materials within the same semiconductor chain. The more attractive it is to produce memory for AI infrastructure, the greater the pressure on the lines intended for consumer electronics.
Cheap phones are the ones with the least margin to absorb the blow
The increase affects the entire industry, but a phone costing 100 or 150 euros has a problem that a model costing 1,000 euros can face much more easily: its economic margin is extremely reduced. Adding ten, twenty, or thirty euros to the manufacturing cost can completely change the product's profitability or force a considerable increase in its final price.
Counterpoint estimates that the rise in memory prices already caused an increase of more than 20% in the total cost of components for certain entry-level smartphones in the first quarter of 2026. By mid-year, the consultancy also noted that the cost of memory had surpassed that of the processor in all market segments.
Manufacturers have several options, none particularly attractive: raise the price, accept a lower margin, install less memory, use older components, maintain previous models for longer, or simply stop producing the least profitable devices. Counterpoint has precisely detected this strategy and speaks of reducing configurations, simplifying catalogs, and greater use of low-cost 4G platforms in some models.
Less RAM, less storage, or a higher price
The crisis is starting to modify the characteristics of phones. TrendForce estimates that 4 GB of RAM is consolidating as the usual configuration in the entry-level range, while manufacturers of mid-range models are refocusing around 8 GB and the adoption of 16 GB configurations is decreasing even in higher segments.
Counterpoint points to a kind of technological "reduflation": the consumer may end up paying more for a new model without receiving the increase in memory, cameras, or storage that would have been usual in previous generations. Brands are also reducing the number of available variants to increase the volume of each configuration and negotiate better the purchase of components.
In the United States, there is already a clear example. Sales of smartphones costing less than 100 dollars fell by 64% year-on-year during the second quarter of 2026, according to Counterpoint, after some manufacturers abandoned that segment or raised prices to offset the cost of memory and storage. The phenomenon is especially important in emerging markets, where those devices represent the main gateway to the internet for millions of people.
The global market is heading for a historic decline
The pressure on the low-end is contributing to a general deterioration of the market. IDC expects global smartphone shipments to decrease by 16.7% in 2026, to just over 1 billion units, which would represent the largest annual decline recorded by the consultancy. At the same time, the average selling price would increase by 27.6%, to 581 dollars.
Counterpoint has different estimates, but agrees on the direction of the problem. In July, it estimated that shipments of processors for entry-level smartphones could fall by more than 30% during 2026, while the memory shortage could extend into the second half of 2027. During the second quarter, global shipments of phones had already decreased by 11% compared to the previous year and were at their lowest level for that quarter since 2013.
The specific figures vary among consultancies because they use different methodologies and update timings, so they should not be interpreted as a single forecast. However, there is a shared diagnosis: the economic range is suffering much more than the premium.
The paradox: AI makes phones that incorporate less AI more expensive
There is also a paradox. The smartphones most affected are precisely those that have fewer possibilities of running advanced artificial intelligence models directly on the device. Phones with local generative AI need more powerful processors and larger amounts of RAM, which continues to concentrate these functions mainly in mid-high and premium ranges.
Counterpoint estimates that smartphones prepared for advanced generative AI functions could represent around 45% of global shipments in 2026, but notes that the additional memory needed to store and run models continues to hinder these capabilities from reaching the more economical terminals. According to the consultancy, devices with advanced generative AI continue to concentrate above about 400 dollars in wholesale price.
Thus, artificial intelligence exerts pressure on cheap mobiles in two distinct ways: data centers absorb more and more memory and manufacturing capacity, while the phones themselves prepared for AI need more RAM. The premium segment can pass much of that cost onto the consumer; the economical one has much less room to do so.
A threat to close the digital divide
The GSMA believes that the consequence transcends the technology market. More than 3.4 billion people still do not use mobile internet, although more than 90% live in places that already have coverage. For many of them, the problem is not the absence of a network, but being able to afford a device to connect.
The sector had been trying for years to bring basic smartphones to increasingly lower prices. The rising cost of memory now threatens that strategy. According to the GSMA, the deterioration of the segment of phones under 100 dollars could widen the gap between those who can access new digital tools and those who cannot even afford the necessary device to use them.
When could prices start to drop again?
There is still no clear date. Memory manufacturers are expanding capacity, but building new factories and increasing production takes years and billions in investment. Furthermore, as long as the large tech groups continue to spend extraordinary amounts on data centers, HBM, server DRAM, and enterprise NAND will continue to be priority and especially profitable products.
There are some elements that could alleviate the pressure. The Chinese company CXMT announced in September that its new DRAM platform has already entered mass production and presented new LPDDR5X chips aimed, among other devices, at smartphones, an expansion that could introduce more supply and competition.
For now, however, consulting firms continue to warn of a tense market. Artificial intelligence is creating an extraordinary demand for memory precisely when mobile manufacturers need those same industrial resources to keep their costs low. The result may seem contradictory: the more money is invested in building the future of AI, the harder it becomes to maintain one of the most basic technological products of the present, the cheap smartphone.