Every electronic device you use depends on parts manufactured by the same handful of companies. Almost all PC and smart-device memory (or DRAM) is made by Samsung and SK Hynix, which are both based in South Korea, and Micron, which is based in the US. The same companies, plus a few others, are responsible for virtually all production of NAND (a kind of non-volatile storage technology), and NAND chips are used in solid-state drives (SSDs), smart-device internal storage, and flash memory.
Similarly, almost all modern processors are created in a handful of semiconductor foundries owned by a small number of large companies. The biggest of these is Taiwan Semiconductor Manufacturing Company, or TSMC, followed by Samsung, with smaller organizations providing a fraction of the manufacturing.
These basic components go into the devices you use every day, including your phone and your laptop, but they’re also the same commodities that feed into the construction of AI servers and development hardware. Companies such as Samsung and SK Hynix have limits on the amount of parts they can manufacture, due to a comparatively small number of manufacturing facilities, a finite number of raw materials, and the time necessary to build NAND and semiconductors. And increasing that output takes huge amounts of money and, more importantly, time.
This interconnected web of manufacturing and business relationships was upended last fall. OpenAI kicked off a supply-chain war with the announcement that it had secured a significant percentage of Samsung’s and SK Hynix’s overall DRAM output (the two companies were responsible for about 67% of the global market combined last year) for an indefinite period of time. Analysts estimate that this arrangement, which allows for up to 900,000 wafers of DRAM per month, could account for 40% of the world’s RAM supply for the duration of the agreement.
That agreement kicked off a component rush, as other AI companies and infrastructure providers moved to lock down similar long-term guarantees for DRAM and then additional components, including NAND products (found in SSDs and flash memory for PCs, smart devices, appliances, and automobiles) and hard disk drives (for servers).
Manufacturers, such as TSMC, that are responsible for fabricating processors and simpler integrated circuits (used for CPUs, GPUs, controller chips, chips for smart devices, and the like) are also seeing increasing demand. Companies such as OpenAI, Anthropic, and Google, often referred to as hyperscalers, have almost locked down industrial-scale production, and smaller players are now snapping up off-the-shelf consumer-grade hardware for AI-development tools, driving up prices for existing products.
The result is a price explosion impacting PC components, such as memory and storage, plus the knock-on effect on devices such as laptops, where prices are already up as much as 20% year over year for comparable models. Companies are expecting that trend to worsen. And companies that thought they had gotten at least two quarters ahead of the RAM-pricing problem last fall are now facing major storage-price increases that they hadn’t anticipated.
Component manufacturers are ending long-term pricing agreements and agreeing only to shorter-term obligations, which means more frequent price increases are on the horizon for device makers. Some manufacturers are already warning that they can’t meet the demands of their “core customers.” In fact, some component manufacturers have already informed their customers that their entire inventory for 2026 is already sold out.
