The AI buildout is printing money for memory-chip suppliers. But device-makers and their customers are getting slammed.
Most of the artificial-intelligence story gets told from the supplier side, where the numbers are spectacular. On the other side are the companies that buy that memory and build it into things they sell: game consoles, phones, laptops and the components inside a desktop computer. They did not create the shortage, but their products are being repriced by it.
For an investor, the question is no longer whether a company is involved with AI. It is which side of the data center a company sells to. The memory and storage makers are printing money. The hardware companies that depend on them are eating the cost, and the market is only starting to tell the two apart.
The cause is a shift in who the memory industry serves. Manufacturers are reallocating wafer capacity toward server DRAM and high-bandwidth memory for AI. TrendForce expects enterprise SSD storage devices to become the largest single category of NAND flash in 2026, with cloud providers locking up the bulk of new supply through long-term agreements.
Pricing follows the allocation. TrendForce put the jump in conventional DRAM contract prices at 90% to 95% quarter over quarter in early 2026, a record, with PC memory prices set to at least double. When a hyperscaler will pay those prices - and a console or laptop maker cannot - the chips go to the hyperscaler. Micron $(MU)$ made the logic explicit in December, shutting its Crucial consumer brand after 29 years to redirect supply to larger customers. The chief executive of controller maker Phison has warned that the shortage will push some consumer-electronics companies out of business in 2026.
Prices for videogame consoles could approach $1,000 - not due to better hardware, but because manufacturers have no pricing power.
The repricing is landing first on the companies that put memory into finished products. Sony (JP:6758) $(SONY)$ and Microsoft $(MSFT)$ have long sold game consoles at near cost and earned their margins on software. But that model breaks when memory costs consume roughly 35% of the hardware bill, as TrendForce expects. Microsoft raised Xbox prices twice last year. Sony is considering delaying its PlayStation 6 to 2028 or 2029 if memory prices do not stabilize. Nintendo (JP:7974) (NTDOY) launched the Switch 2 at about $450, lifted in part by a 41% jump in its DRAM costs, and analysts expect another increase before year-end.
Some analysts now expect prices for the next console generation to approach $1,000, turning a mass-market product into a luxury purchase. None of this reflects better hardware. It reflects companies with thin margins and no pricing power over their most expensive component.
The end of the low-cost PC
Makers of personal computers face the same arithmetic, though not all are equally exposed. Memory and storage have climbed to roughly 35% of the bill of materials at HP $(HPQ)$, up from 15% to 18% a quarter earlier. A jump like that forces a vendor to raise prices or surrender margin. Dell Technologies $(DELL)$ faces the same pressure in PCs but carries a large and growing AI server business that offsets much of it, which is exactly the point: exposure, not the label "PC maker," determines who absorbs the damage. Gartner expects the squeeze to cost client vendors volume regardless, with consumer PC lifespans lengthening by about 20% and the sub-$500 laptop disappearing by 2028.
The least-protected companies sit furthest from the data center, among them the makers of the cases, coolers, power supplies and memory modules sold almost entirely to people building their own machines. They have no server business to offset the damage. For example, Corsair Gaming (CRSR), one of the few publicly traded American companies in this group, warned in a regulatory filing that high memory costs act as a barrier for builders, and its components guidance is already softening even as prebought inventory provides a cushion in the near term.
The same shortage is a windfall for the companies on the other side of the wall, and the contrast is the most useful map an investor has. Storage is the sharpest example. Western Digital $(WDC)$ has pretty much sold out its entire 2026 hard-drive production, has signed supply agreements running into 2027 and 2028, and has raised its dividend 20% on the strength of AI data demand. Seagate Technology $(STX)$ has watched high-capacity drive prices climb sharply while it sells its output well in advance. Sandisk $(SNDK)$ is riding the same NAND wave.
Priced for scarcity
The AI buildout is splitting the technology sector rather than lifting all of it.
These stocks are priced on the assumption that memory stays scarce. That is not a certainty. A genuine efficiency breakthrough, or a pause in data-center capital spending, would reprice both sides of this market fast: relief for the consumer-facing companies, and a reckoning for suppliers valued for permanent scarcity. The memory makers are guiding to tight supply into 2027, so the near-term direction is set. The risk is in how quickly it could unravel.
For now, the AI buildout is splitting the technology sector rather than lifting all of it. On one side are the companies selling memory, storage and accelerators into the data centers, with pricing power and multiyear backlogs. On the other is a long list of console, PC and component makers absorbing the cost, thinning their margins and postponing products - hoping to wait out a shortage they cannot influence and survive in a world where the memory goes to the data centers first.