AI is killing the cheap smartphone

One of the most remarkable things about the last few decades is how cheap computers have gotten.

In 1985, if you were a reasonably affluent American, the best computer that you could afford was the IBM PC AT. The PC AT would cost you about $6,000—$19,400 in 2026 dollars—and thus represented about a quarter of the median American’s annual income; and it ran on an Intel 80286 processor, capable of something like 900,000 instructions per second. Today, if you find yourself in a market stall in Nairobi or Lagos, you’ll be able to find a cheap smartphone—like the Tecno Spark Go, manufactured by China’s Transsion—for somewhere between $30 and $120. That phone will run on a processor capable of billions of calculations per second.

In other words: you can buy a computer thousands of times more powerful than the best consumer device from 40 years ago, for something like 0.3 percent of the price. No other good in history has experienced a decline in cost on that scale: poor people can now carry around in their pockets computers many orders of magnitude more powerful than what the richest slice of the world’s population could afford a few decades ago. And that great cheapening of consumer electronics has enabled a diffusion of computing power to the world’s population that is nothing short of miraculous. Hundreds of millions of the world’s poorest people are able to access the internet because of cheap smartphones like the Tecno Spark Go.

That era is now coming to an end.

In 2026, the International Data Corporation, which tracks the smartphone market, predicted that worldwide smartphone shipments would fall 13 percent, their largest single-year decline ever. The crash would be most intense in Africa and the Middle East, where smartphone shipments would fall by more than 20 percent, and would be concentrated in the cheapest end of the smartphone industry. This shock represented not a temporary blip but indeed “a structural reset of the entire market”: a huge share of the world’s population is getting priced out of smartphone ownership.

So the trend of the last few decades, of consumer electronics getting better and cheaper every year, faces a sharp reversal: the poor world is now entering a smartphone crisis.

This is happening for a simple reason.

Smartphones, like other computers, use memory: and the global supply of memory is remarkably inelastic, because memory is really hard to produce. For a long time, most memory went to smartphones and laptops; but in the last few years, AI has emerged as an enormous and hugely profitable consumer of memory. This has resulted in a huge reallocation of memory away from consumer electronics and toward AI. The inevitable result is that smartphones are much more expensive to make now than they were a few years ago. In the short term, this means that the cheap smartphone, which spread computing and internet access to the poorest parts of the world, is dead.

But at the rate that things are going, it seems like the poor world will only be the first to get hit. If AI consumption continues to grow at current rates—or if it accelerates, as seems manifestly possible—it won’t be long before the smartphone crisis spreads to the rich world. Consumer electronics are about to get much more expensive.

Smartphones are computers. They’re very small computers, and also have things like touchscreens and radios. But in terms of their internal architecture, smartphones are basically the same as what you’d get with a laptop or a server. They have a processor that performs calculations and runs the logic that makes the device do what you tell it to do. They have memory that holds the data that the processor is currently working on. They have storage that retains data when the device is turned off. And they have a circuit board that connects all these different things together.

The big story of computing over the last few decades is the processor. You can think of the processor as a huge array of transistors—tiny switches that flip ON and OFF to perform logical operations. We’ve done a good job—a very good job—of figuring out ways to make transistors smaller and more efficient, which means that processors have improved at an exponential rate over the last few decades. This is Moore’s Law.

But processors can only process the data that they have access to: and the data that they have access to is what they get from memory: specifically, in modern computers, from DRAM, “dynamic random access memory.” Here the story is very different. DRAM has gotten better; but it hasn’t gotten better at anything like the rate that processors have: in the 1980s and ‘90s, processor speeds improved at 60 percent per year, while DRAM speeds improved at just 7 percent per year.

And that means that for the last few decades, the main bottleneck for computer performance has been memory. Computer scientists call this the “memory wall.” A huge amount of the work in computer architecture over the last few decades has been finding various ways around the mismatch between processors and DRAM.

So why hasn’t DRAM improved as fast as processors?

Simply put: it’s just a really hard problem. Just like a processor is a huge array of transistors, a memory chip is basically a huge array of memory cells: and each memory cell has both a transistor and a storage unit called the capacitor, which holds the electrical charge corresponding to an individual bit of data. We know how to shrink the transistor. But shrinking the capacitor is a lot harder. As the capacitor gets smaller, it becomes harder for it to reliably store its electrical charge: the charge might leak out, or disappear, or be altered by interference from its neighbors. So if you want to make DRAM more efficient, you need to resort to all sorts of increasingly exotic architectures.

And that’s exactly what’s happened. DRAM needs to get more efficient, in order to keep up with the improvements in processors. So modern DRAM manufacturing is an extraordinarily complex and expensive process. Building a single state-of-the-art DRAM fabrication facility, a “fab,” will cost you about $15 to $20 billion; acquiring all the necessary equipment, like lithography tools and etching machines, will cost you another few billion; and then it’ll take you a few years of producing substandard and defective memory chips before your yields start to look competitive.

Which leads us to the peculiar economics of the companies that manufacture DRAM: the “memory makers.”

The most important thing to know about memory, beyond the fact that it’s expensive and difficult to make, is that it’s fungible. Processors are bespoke: you can’t swap an Intel chip for an Apple chip. But memory chips are not bespoke. DRAM chips all conform to the same industry-wide standards, so a chip from one memory maker will slot into the same device as a chip from any other. DRAM, in other words, is a commodity.

And that combination—capital-intensive manufacturing plus fungibility—is a punishing combination. Because memory is fungible, the industry is intensely cyclical: the entire history of the DRAM industry is a history of boom-and-bust supercycles. First, strong demand from one sector or another—like Windows PC adoption in the 1990s—drives surging prices and a wave of investment from every player; cumulative overinvestment in an undifferentiated good produces oversupply; and then oversupply leads to collapsing prices.

And because production is so expensive, those down-cycles turn out to be existential: the memory industry is marked by constant wreckage. Intel dominated the memory game in the early 1970s but left in the 1980s, opting to focus on processors. Texas Instruments and IBM, also once major players, left in the 1990s. Germany’s Qimonda collapsed in 2009; Japan’s Elpida, once the world’s third-largest DRAM manufacturer, declared bankruptcy in 2012.

And decades of collapse and consolidation left only a few players standing. In the 1990s, there were perhaps 20 meaningful producers of DRAM around the world; today there are three that account for more than 90 percent of global production. South Korea has two, SK Hynix and Samsung; and the United States has one, Micron.

And these memory makers have learned a very particular lesson from the unforgiving history of their industry: always leave demand unmet. The only way to survive in a capital-intensive and cyclical industry was to demonstrate an almost superhuman degree of capital discipline. Demand might rise now, but it would always fall. So it was better to let prices spike and see the marginal memory consumer priced out than to expand production and risk destruction when demand inevitably softened.

And this, it turns out, is a brutal calculus for smartphone customers.

Earlier, I said that memory is “fungible.” That requires a qualification. Memory is fungible between manufacturers: a chip from Samsung will slot into the same device as a chip from SK Hynix. But that doesn’t mean all computers use memory in the same way. The MacBook Pro on which I’m writing this piece needs memory that can keep up with a powerful processor running many programs at once: so it uses a standard called DDR, “double data rate,” which runs at a reasonably high voltage and offers high bandwidth. The processor on my iPhone is less powerful, so it needs less data at any given moment; but voltage matters enormously, since every milliwatt allocated to memory is drained from the battery. So smartphones use LPDDR, “low-power double data rate,” a variant of DDR engineered to operate at lower voltages. And in the data centers where Claude and ChatGPT are run, an entirely different standard is used: HBM, “high-bandwidth memory,” which I’ll get back to shortly.

All three of these are made the same way, from the same starting material. Memory makers receive thin silicon discs called wafers; over several months, they etch billions of memory cells onto them; and then they cut wafers into individual chips and ship them.

The key question facing a memory maker, then, is how to allocate its wafers between DDR, LPDDR, and HBM. Some percentage of wafer allocation is locked in through long-term agreements with major purchasers, like Apple or Dell; and some is sold on the spot market, to buyers who want flexibility or lack the scale for long-term agreement. So every quarter, the wafer allocation teams at Samsung, SK Hynix, and Micron decide—based on prices, contracts, and their best guesses about the direction of future demand—how to distribute their wafers across the three categories.

For most of the history of the industry, this allocation was straightforward. In the late 2010s, margins were broadly similar for DDR, LPDDR, and HBM; what interested the memory makers most was volume, and wafer allocation basically tracked end-market demand. Phones were the single largest market for memory, so LPDDR got most of the wafers. DDR took most of the rest. And HBM was a niche product for high-performance computing customers, so it got only a small sliver.

That changed dramatically with AI.

Training and running AI models is extraordinarily computationally intensive. Even simple queries require billions of matrix multiplications, done in sequence and in parallel, over and over again. AI workloads need computers that can do enormous numbers of operations in parallel — which is why specialized hardware like Nvidia’s GPUs and Google’s TPUs has become so important. But because GPUs and TPUs perform so many calculations at once, they need to be fed data at a correspondingly enormous rate. Otherwise the expensive hardware sits idle. What was needed, in other words, was memory engineered to deliver vast quantities of data to many processors at once, at extraordinarily high speeds.

That is exactly what HBM was designed to do.

The core idea of HBM is simple. You take lots of DRAM dies, stack them on top of each other, connect them with thousands of tiny vertical channels so that many data paths can operate in parallel, and then place the whole stack right next to the GPU or the TPU. Actually doing this is very hard. But if it works, you can transfer an order of magnitude more data than you could with DDR.

The catch with HBM, beyond the difficulty of producing it, is that it is enormously wafer-intensive. It is not just that you are stacking a lot of dies together. Because of all the peripheral circuits and all the vertical channels, a single gigabyte of HBM consumes more than three times the wafer capacity that a gigabyte of DDR or LPDDR does. Every gigabyte of HBM produced is, in effect, three gigabytes of commodity memory not produced.

For a long time, this didn’t really matter, because HBM demand was small. When ChatGPT was released in November 2022, the memory makers were in the middle of a demand slump, and it took them some time to register that something had shifted. In early 2023, the industry trade press was still hedging, with reporting limited to suggestions that “AI chatbots may help shorten the DRAM market slump.”

But HBM demand increased much faster than the memory makers expected. AI usage continued to explode; and as usage shifted to more intensive models—from chatbots to long-running agents—it became clear that demand for HBM would be much, much bigger than anyone had originally anticipated. The memory makers were caught flat-footed. By the end of 2024, a full HBM shortage had set in; by 2025, HBM margins were running at 70 percent or higher, while margins for DDR and LPDDR sat between 20 and 30 percent.

The rational response, for the memory makers, was obvious: pump out more HBM. And so they reallocated a massive amount of capacity. In 2023, HBM accounted for 2 percent of the memory makers’ wafers; in 2024, 5 percent; in 2025, 10 percent; and by the end of 2026, the share is expected to hit 20 percent, with an additional 3 percent allocated toward high-density DDR for AI servers. And so in the space of three years HBM went from a peripheral product category to the very core of the memory industry. SK Hynix, which had been first to reach volume production of the leading-edge HBM node, saw its HBM revenue increase fourfold in 2024 alone; by the end of that year, HBM accounted for more than 40 percent of the company’s DRAM revenue, up from roughly 5 percent two years earlier.

But even this reallocation hasn’t been enough. Demand continues to outrun supply, and the memory shortage remains one of the defining features of the AI buildout. (It has, in turn, produced all sorts of workarounds, like quantization or DeepSeek’s multi-head latent attention.) So heated has the race for memory become that at the end of 2025, executives from hyperscalers like Microsoft and Google were reportedly “practically taking up permanent residence in Korea” lobbying Samsung and SK Hynix for allocation. More than 30 percent of hyperscaler capital expenditure is now going to DRAM alone.

This has been fantastic news for the memory makers. In 2025, they earned a collective $70 billion in profit; in 2026 they’re expected to earn more than double that amount. Samsung, SK Hynix, and Micron are now among the most profitable companies in the world.

But things are not so happy for the purchasers of commodity DRAM.

Recall what we said earlier about the capital discipline of the memory makers. SK Hynix, Samsung, and Micron survived previous DRAM cycles by refusing, almost as a matter of principle, to supply enough chips to meet all their customers’ demand: the lesson of Elpida and Qimonda was that idle fabs were fatal, while unmet demand was not.

So when the memory makers looked at the rising tide of HBM orders in 2024 and early 2025, they took a deliberately conservative approach and refused to expand production. It was only in 2025, as memory prices began an unprecedented surge, that the memory makers started to build new fabs targeted at HBM, all slated to start producing chips in 2027 or 2028. Even now they’ve been careful not to expand capacity too drastically. As late as December 2025, Samsung stressed that it would “prioritize long-term profitability over rapid capacity expansion.”

And that meant that the only way that the memory makers could meet surging HBM demand was to reallocate wafers away from DDR and LPDDR. As Tom’s Hardware reported at the end of 2025: “with wafer starts flat and packaging lines locked, every wafer pushed into HBM removes capacity from commodity DRAM.” By the end of 2025, SK Hynix was allocating 30 percent of its wafer capacity to HBM, with almost all of that capacity having been taken away from DDR and LPDDR. Micron, meanwhile, opted to simply exit the commodity DRAM market entirely. In December 2025, Micron discontinued its consumer-oriented Crucial brand and announced that it would cease all consumer shipments, redirecting all capacity to AI and enterprise.

And so the supply of memory available for DDR and LPDDR has cratered over the last few years. Accordingly prices have spiked. Between the first quarter of 2025 and the first quarter of 2026, prices for the LPDDR4 standard increased 250 percent; LPDDR5 prices increased by 220 percent. In some corners of the market, the spike was more severe: DDR5 prices in Germany increased 414 percent over the course of a year.

And so memory has rapidly become the most expensive component going into consumer electronics. The memory share of the bill of materials on a budget Android phone has gone from around 15 percent to as much as 50 percent.

This means higher prices for all consumer electronics. But it’s particularly damaging for the marginal consumers, those least able to pay higher prices. In the case of memory, that means the makers and consumers of budget smartphones.

For a long time, the budget smartphone companies—like Transsion, Oppo, Vivo, and Lava—followed a simple model. They would buy last-generation components on the spot market, assemble them cheaply as Android handsets, and then sell the finished product for an extremely low price. The budget phone makers had extremely thin margins, usually somewhere in the low single digits; but they sold phones at huge volumes. Transsion, for example, shipped 105 million phones in 2024, compared to Apple’s 230 million. And in cheaper markets, like Africa or South Asia, these companies were dominant: Transsion alone held 48 percent of the African smartphone market.

But that model breaks when memory prices spike as much as they’re now spiking. The sub-$100 smartphone risks becoming “permanently uneconomical” as a product.

And that means that the budget smartphone makers have been forced to pass memory costs onto consumers: smartphones that sold for $50 are now selling for $120 or more. And price-sensitive consumers have responded by simply not buying phones. In the early months of 2026, Transsion announced that its net profit for 2025 had fallen by 54 percent, and that it would cut its annual shipment target by 40 percent. We’re seeing the same with other low-market and mid-market smartphone companies. Oppo slashed its shipment target by more than 20 percent; Vivo, in the same position, cut by nearly 15 percent. In the first quarter of 2026, Xiaomi’s annual shipments fell 19 percent year over year.

And that repricing has had a stark effect in poor countries. In India, the sub-$100 smartphone market collapsed 59 percent year-on-year in the first quarter of 2026: surging memory prices resulted in a “forced premiumization” of the Indian smartphone market. But in the poorest markets, such premiumization isn’t a possibility. In 2025, 81 percent of smartphone shipments in Africa were in the sub-$200 category: as smartphone prices surge, many African consumers will simply be priced out of phone ownership entirely.

That’s where we are now. HBM demand is already crowding out DDR and LPDDR; and this is already resulting in a large and growing share of consumers being priced out of smartphone ownership.

But there’s no reason to think that this trend will stay confined to the poorest consumers. Companies higher up on the DRAM food chain are starting to feel the pain of higher memory prices; it won’t be long before the consumers of the rich world feel themselves being priced out of the electronics market.

We’re already seeing early signs of this. Samsung’s consumer division, for example, found itself unable to secure a long-term LPDDR agreement with Samsung’s memory division; it thus had to ship its Galaxy S26 phone with less memory than expected and at higher prices. This didn’t do much to help: Samsung executives warned that the company would record its first-ever annual net loss on smartphones. (More than balanced out, of course, by its enormous profits on memory.) We’re seeing the same repricing with Dell, which hiked laptop prices by 15 to 20 percent in December 2025.

Even Apple, the king of the electronics world, is starting to feel the bite of memory costs. Apple traditionally enjoyed significant bargaining power with the Korean memory makers, negotiating long-term agreements to smooth prices out for years at a time; but now the memory makers are the ones with the leverage. When Apple’s latest long-term agreement expired in January 2026, the memory makers refused anything lasting longer than a quarter at a time. In February, in order to secure supply at all, Apple agreed to pay Samsung a 100 percent premium on the LPDDR5X memory destined for the iPhone.

And so the pricing pressure on Apple has grown massively over the last six months. Over the course of 2025, the 12GB LPDDR5X chips that power the iPhone 17 Pro had increased in price by 230 percent; without its long-term agreements to protect it, Apple would face the full brunt of the memory crunch. In order to cope, Apple has announced a wave of delays over the last few months. The iPhone 18 standard model has been delayed to spring 2027; the new Mac Studio was delayed from summer to fall.

There’s no sign that things are going to get better anytime soon. Indeed, even if the memory makers stop reallocating wafer capacity to HBM, there will still be enormous pressure on LPDDR. In the last quarter of 2026, Nvidia will be launching its new Vera Rubin platform, a rack-scale AI supercomputer that pairs Rubin GPUs with Vera CPUs into a single system built for large-scale AI training and inference. The Vera CPUs will be enormously hungry for LPDDR: by 2027, Vera Rubin is projected to consume more LPDDR than Apple and Samsung combined. A report from JPMorgan projected that memory could account for 45 percent of the iPhone’s component cost by 2027, against roughly 10 percent today. Within the year Apple will be forced to make a decision: either it will cut into its margins to defend market share, or dramatically increase prices on its products.

All of which is to say: things are going to get a lot worse before they get better.

We’re already at the point where marginal buyers in the poor world are getting priced out of the smartphone market. We’re rapidly approaching the point where buyers in the rich world feel the same thing. In the short term, smartphone makers might be able to cope by significantly reducing the amount of memory per device, and thus degrading equipment performance, or simply by destroying demand through large price hikes. Margins for LPDDR and DDR have soared, and may even be higher than HBM margins: but so much HBM capacity has been secured via long-term agreements that there is no pivot to commodity DRAM coming anytime soon. If there’s any hope of relief, it’s coming from China. Upstart Chinese memory makers—like ChangXin Memory Technologies, which already commands more than 30 percent of China’s LPDDR market—are scaling up rapidly and hoping to fill the gap for DDR and LPDDR.

But as long as we are facing a shortage of memory for AI data centers, the economics of the DRAM shortage will be difficult to escape. Hyperscalers are simply willing to outbid budget phone manufacturers for access to DRAM: even ChangXin is planning to convert about 20 percent of its capacity to HBM.

So it will be hard to avoid a great repricing of consumer electronics in the coming years. We’re already in a world where poor-world consumers are getting priced out; we’re rapidly approaching a world in which rich-world consumers are getting priced out as well. The last few decades of technological progress democratized computing; but that era is now over. The long trend of consumer electronics getting faster, cheaper, and more powerful every year has reversed. The people who will feel it first, and feel it worst, are the world’s poor: but it won’t be too long before we, too, feel the crunch.