Every AI system running today needs memory chips, a specific type of hardware that stores and retrieves information at speed. The three companies that produce roughly 95% of those chips globally, Samsung, SK Hynix, and Micron, have a finite amount of factory space. As demand from AI data centers surged, they shifted that space toward the more profitable chips that AI needs. That left less supply for consumer devices. Prices for the chips used in laptops and tablets then climbed sharply: DRAM prices rose 98% in the first quarter of 2026 alone, according to research firm TrendForce, with further increases projected.
Apple was among the last major device makers to pass those costs on to customers. It had more room to absorb the pressure: Apple operates on device profit margins of around 35% to 40%, compared to about 7% for Lenovo or HP. That advantage ran out on June 25, when Apple raised prices across its entire Mac and iPad range globally. Tim Cook had described the situation as a "hundred-year flood." Apple shares fell more than 6% on the day of the announcement, a signal of how significant a move this was even for a company with Apple's financial position.
For businesses, the practical question is straightforward: what do you do when the equipment you need to buy has jumped 15% to 25% in price with no near-term reversal in sight? Most organizations are extending how long they keep existing devices. The standard corporate laptop replacement cycle has been three to four years. Gartner confirmed in February 2026 that enterprise organizations are on track to extend device lifespans by 15% or more by the end of this year, driven directly by the memory crisis.
There is a real tension here. Keeping a device longer saves on purchase costs, but older machines require more IT support, run slower, and carry greater security risk. Industry data shows PC maintenance costs rise 59% between year one and year four. For businesses in regulated sectors such as insurance, healthcare, or finance, running older devices that cannot support the latest security patches is not just an inconvenience; it is a compliance exposure. The savings calculation only works if the machines being extended are relatively recent and well-specified to begin with.
One practical insight from the current situation: companies that bought well-specced devices with plenty of memory before the shortage hit are in a much better position to extend their use. Machines bought lean to save money two years ago are now the ones causing headaches. This is worth keeping in mind if and when prices normalize: buying slightly above minimum specification tends to pay off over time.
The shortage is also hitting unevenly. Education and government organizations, which tend to buy larger volumes of budget-priced laptops, are taking the hardest hit. Shipments of PCs priced below $500 dropped nearly 19% year-on-year in the first quarter of 2026. Businesses that can afford to buy in the mid-range and above have more options. Some vendors have even begun selling pre-built computers without memory included, an unusual step that shows how constrained supply has become.
The longer-term picture depends on new factory capacity. Samsung and SK Hynix announced a combined $518 billion investment to build new chip manufacturing plants in South Korea. That number is enormous, but building a chip factory takes years. Assembling SK Hynix's current main facility in South Korea was a nine-year effort. Most analysts expect no meaningful price relief before 2028. Jefferies Equity Research projects memory prices will continue rising through 2027 before new capacity starts making a dent.
For business operators right now, the sensible moves are narrow but clear. Audit your current device fleet and know which machines have the memory capacity to handle AI-assisted tools, because those are the ones worth protecting and extending. Avoid panic-buying in the current price environment if you can wait. And if you do need to buy, buy with enough memory headroom that the device can still do the job two years from now, when AI features inside standard business software will demand even more of it.