<h2>The Invisible Component That Is Changing What You Pay for Everything</h2>
<p>Most people have never thought about the specific type of memory chip that goes into their laptop. They know their computer has some amount of memory, they know more is generally better, and they know the price they paid for the machine. The specific name of the component — DRAM, which stands for Dynamic Random-Access Memory — is the kind of technical detail that only hardware enthusiasts and procurement managers have historically needed to care about. In June 2026 it became something that every consumer in every developed economy needed to understand, because the global supply of DRAM has been dramatically affected by the artificial intelligence infrastructure buildout, and the price consequences of that shortage are now appearing on the price tags of MacBooks, iPads, Xbox consoles, PCs, tablets, and every other device that uses the component.</p>
<p>Apple's announcement on June 25 was perhaps the clearest statement yet of what has happened. "We have never seen a component price increase this much, this quickly," the company said. That statement, from one of the world's most sophisticated supply chain managers — a company that sources components globally with extraordinary planning and negotiating sophistication — should be taken seriously as a measure of how unusual the current environment is. DRAM contract prices rose as much as 98 percent in the first quarter of 2026 alone. Tarun Pathak of Counterpoint Research told TechCrunch that memory prices had increased more than fourfold since the fourth quarter of 2025. Gartner projects a 130 percent surge in combined DRAM and SSD prices by the end of 2026. These are not modest adjustments. They represent a structural disruption in the economics of the consumer electronics industry that traces directly to one cause: the AI boom.</p>
<h2>What Memory Chips Do and Why Computers Need Them</h2>
<p>Before understanding why AI is consuming memory chips, it helps to understand what they do. DRAM — Dynamic Random-Access Memory — is the short-term working memory of a computer. When you open an application, the relevant data and code are loaded from your storage drive into DRAM so that the processor can access them quickly. DRAM is fast but volatile: it loses its data when power is removed, which is why you need to save files before shutting down your computer. The more DRAM a system has, the more data it can hold in fast-access working memory simultaneously, which generally means faster performance and the ability to run more applications or handle more complex tasks without slowing down. A typical laptop in 2026 contains between 8 and 32 gigabytes of DRAM. A typical desktop workstation might have 32 to 128 gigabytes. These are the familiar scales of consumer computing.</p>
<p>NAND flash storage — the type of memory used in solid-state drives rather than in working memory — is a different but related component that is experiencing its own shortage for similar reasons. NAND retains data when powered off, making it suitable for long-term storage. Modern laptops use NAND-based solid-state drives for their operating system and files rather than the older mechanical hard drives that preceded them. The AI data centre buildout requires enormous quantities of both types of memory: DRAM for the working memory of the servers running AI computations and NAND for the storage systems that hold the training datasets, model weights, and intermediate computational results that large AI systems require. The distinction matters because both shortages are contributing to the consumer price increases of June 2026, though DRAM is the more acute bottleneck.</p>
<h2>How AI Data Centres Devoured the World's Memory Supply</h2>
<p>The AI infrastructure buildout began in earnest in 2023 when generative AI applications demonstrated commercial viability at scale and the major technology companies began racing to build the data centres capable of training and deploying frontier AI models. The computational work of AI — specifically the matrix multiplication operations that underlie neural network training and inference — is performed primarily by graphics processing units made by Nvidia, AMD, and others. Those GPUs require a specialised variant of memory called High-Bandwidth Memory, or HBM, which provides the extremely fast data access rates that allow GPUs to process the enormous datasets used in AI training. HBM is manufactured using the same fabrication processes and capacity as conventional DRAM, produced by the same three companies — Samsung, SK Hynix, and Micron — that dominate the global memory market.</p>
<p>As hyperscalers — Microsoft, Amazon, Google, Meta, and others — committed to spending hundreds of billions of dollars on AI infrastructure, they negotiated long-term supply agreements with Samsung, SK Hynix, and Micron for HBM production. Those agreements locked up capacity — the physical manufacturing facilities and the production time — that would otherwise have been available for conventional DRAM production. The memory industry has a relatively fixed total capacity at any given time: building new fabrication plants takes years and costs billions of dollars, and cannot respond quickly to sudden changes in demand. When a large portion of that capacity is committed to producing AI-grade HBM under multi-year contracts, less of it is available to produce the conventional DRAM that goes into your laptop, your smartphone, and your gaming console. Supply contracts, Micron's blowout Q3 2026 earnings — reporting 41.46 billion dollars in revenue, quadruple a year earlier — illustrate the other side of the same story: for the memory makers, the AI boom has been extraordinarily profitable. The customers who purchase their AI-grade products are paying extremely high prices and the memory companies are generating margins and cash flows they have never previously achieved. The cost of that profitability is borne downstream by the consumers who buy the devices whose manufacturers cannot get conventional memory at the prices that prevailed before the AI boom began.</p>
<h2>When Does the Shortage End?</h2>
<p>The honest answer is: not soon. Gartner's 2026 forecast of a 130 percent combined DRAM and SSD price surge by year-end, with no meaningful relief until late 2027, reflects the lead times involved in adding new memory manufacturing capacity. Building a modern DRAM fabrication facility — called a fab — takes approximately three to four years from groundbreaking to production, costs five to ten billion dollars, and requires specialist equipment that is itself subject to long delivery timelines. Samsung, SK Hynix, and Micron are all investing in new capacity, but those investments will not produce meaningful additional supply until 2027 and 2028 at the earliest. In the meantime the AI infrastructure buildout shows no signs of slowing: Microsoft has committed to 190 billion dollars in AI infrastructure in 2026. Amazon has signalled 200 billion dollars. Google and Meta are also accelerating. The demand side of the equation is growing faster than any realistic assessment of the supply side's ability to respond.</p>
<p>For consumers the practical implication is straightforward: the devices you buy in 2026 will cost more than the equivalent devices cost in 2025, and the devices you buy in 2027 may cost more still. The iPhone, which Apple has so far spared from price increases, will be the most watched single product announcement of the autumn technology season. If Apple raises iPhone prices, the signal will be that no consumer electronics product is immune to the AI memory crunch. If it holds firm, the question will be how long it can continue to absorb costs that are by its own admission unprecedented in the company's history. The deeper question — whether the AI infrastructure buildout that is causing the shortage will ultimately produce benefits for consumers that justify the higher device costs it is generating — is the kind of question that will be answered over years and decades rather than months. For now the answer visible in the Apple Store price list is simply: more.</p>