A memory kit that cost about $191 a year ago now runs $1,118, up 485%, according to the website Tom's Hardware, and that's just one data point behind a brutal October: PC shipments tumbled more than 20% in the third quarter as laptop and desktop makers got priced out of the same memory that data centers are now hoarding.
That's what a memory shortage actually looks like in practice. You hear a lot of talk about such a shortage on earnings calls or in tech news segments, but what does that really mean?
It means real tech products are getting more expensive or simply not being built. Whatever chipmaker controls that scarce supply gets to name the price, and right now, it seems like it's the artificial intelligence (AI) companies that can do it.
The question worth asking isn't which AI company has the flashiest model. It's which one owns something scarce when demand exceeds what factories can produce.
Image source: Getty Images.
How prices are moving, and why
According to the website TrendForce, conventional DRAM contract prices jumped roughly 90% to 95% in the first quarter of 2026, with NAND flash memory climbing 55% to 60% over the same stretch. And suppliers have been warning customers to brace for further monthly increases heading into year-end.
SK Hynix (SKHY -6.05%) told investors that its memory capacity is essentially sold out for 2026. That's not a glitch in the supply chain; it's a deliberate choice.
Micron Technology (MU -4.08%), SK Hynix, and Samsung Electronics (SSNLF +0.00%), which together control more than 95% of global DRAM production, have been redirecting wafer capacity away from ordinary memory chips and toward high-bandwidth memory (HBM).
NASDAQ: SKHY
Key Data Points
AI talk can get really confusing and jargon-heavy. Basically, the three companies that make almost all the world's memory chips looked at two options: either build dumb, everyday memory chips cheap and in bulk, like they always have, or build the fancier, pricier memory chips that AI data centers are desperate for.
And they chose to lean hard into the second option instead of racing to build more factories, like they've done in past boom cycles. It's a smart financial choice because AI is where the money is these days.
Who holds the pricing power?
Micron sits at the center of this story as the only major U.S.-based memory maker, which is exactly why its own numbers now read like the rest of this sector's. Sandisk (SNDK -5.33%) tells perhaps the starkest version: Its gross margin hit 78.4% last quarter, a figure that belongs to a software company, not a chip manufacturer.
NASDAQ: MU
Key Data Points
It seems this is a direct result of NAND scarcity rather than any special differentiation in what Sandisk actually makes. SK Hynix now trades in the U.S. as well, with its American depositary receipts beginning trading on the Nasdaq stock exchange in July. All three are, for now, in the rare position of selling a commodity that behaves like a luxury good.
Western Digital (WDC -4.36%) deserves a spot here, but with somewhat of an asterisk. Storage demand swings more with the broader economy than pure memory does, so even amid today's shortage, Western Digital's fortunes remain more tied to big-company budgets than do those of Sandisk or Micron.
NASDAQ: SNDK
Key Data Points
Nvidia (NVDA -3.16%) earns its own asterisk for the opposite reason: It benefits enormously from AI demand, but it also needs a steady, affordable supply of HBM to build its own chips, which means the same shortage lifting memory makers' margins is simultaneously a cost and availability risk sitting inside Nvidia's own supply chain.
Who's getting squeezed?
Apple (AAPL +0.49%) is where this shortage turns from an abstraction into a real number. DRAM costs for its next flagship iPhone are projected to roughly triple, according to Fortune, with memory potentially climbing from around 10% of an iPhone's component cost today to more than 45% by 2027.
The company has already guided for a lower gross margin for a recent quarter, specifically citing rising memory costs, and estimates suggest it would need to raise iPhone pricing by hundreds of dollars just to hold its margin steady, a trade-off with no comfortable answer, according to the British management consultancy EBC Research.
Super Micro Computer (SMCI -8.42%) is feeling this even more acutely since it builds AI servers packed with exactly the memory everyone's fighting over, with thinner purchasing power than larger rivals to cushion the blow. Its gross margin swung wildly through 2026 as memory and tariff costs moved in and out of its favor.
So what does this actually mean for an investor watching from the sidelines? Mostly that the clock matters as much as the picks themselves. New chip fab capacity isn't expected to arrive in meaningful volume until late 2027 at the earliest, with real relief more likely in 2028 or 2029, according to memory executives, including Micron's CEO. Until new supply physically exists, this remains a story about which companies absorb higher costs and which profit from them.
