The AI Memory Crunch Now Runs to 2028

The shortage of memory chips that feed artificial intelligence is no longer a passing squeeze. On its latest earnings call, Micron told investors the scarcity is structural rather than cyclical, and that meaningful new industry supply will not arrive until 2028. In the meantime the company says it can satisfy only half to two-thirds of customer demand for high-bandwidth memory and DRAM, the chips that keep AI accelerators fed with data.

The numbers behind that claim are striking. Micron's gross margin reached a company record of about 85 percent, more than double a year earlier, and management guided the next quarter higher still. DRAM prices rose by roughly 60 percent from the previous quarter. When a supplier can raise prices that fast and still sell everything it makes, it is not describing a soft market.

Booked solid into 2028

What makes the crunch likely to last is who is doing the buying. Cloud providers have been signing long-term agreements that lock up capacity two years out, which means a large share of 2026 and 2027 production is already spoken for before it is manufactured. As Tom's Guide reported, Micron does not expect the imbalance to ease until new fabrication lines come online toward the end of the decade. Building those lines takes years and tens of billions of dollars, so there is no quick fix even if demand cooled tomorrow.

That is the part worth sitting with. The AI build-out has run on an assumption that the picks and shovels, the chips and the memory around them, could scale as fast as the ambition. Memory is now the pinch point that says otherwise.

The bill reaches the rest of us

For most of this year the memory boom was an industrial story, fought over between chipmakers and hyperscalers. It is starting to spill into consumer products. Huawei has warned that memory inflation is adding roughly 200 dollars to the cost of building a smartphone, a figure that eventually shows up in the price shoppers pay. Anything with RAM in it, from phones to laptops to game consoles, sits downstream of the same shortage.

The financial backdrop is tense too. Analysts at Goldman Sachs have counted close to half a trillion dollars in AI-related debt, and warned that the spending assumes returns will arrive faster than the technology's commercial payoff plausibly allows. A memory shortage that props up chipmaker margins while squeezing everyone who buys their parts is exactly the kind of imbalance that looks great on one earnings call and uncomfortable on the next.

It also rhymes with other recent signs that the physical limits of AI are starting to bite. Chipmakers have been pushing through price hikes on the strength of AI demand, while some infrastructure players have pulled back from their biggest power bets. The compute race is increasingly a race for the raw materials to run it, and right now the memory is not there.