Amazon's expanded GPU commitment to Nvidia points to deepening AI-chip and memory scarcity, but Micron, the stock most closely tied to that story, closed lower Thursday while the rest of the tech sector rallied.
Nvidia and Amazon Web Services disclosed Wednesday that AWS has signed on for two million additional Nvidia GPUs, pushing its total commitment past three million through 2028. The expansion followed the exhaustion of AWS's previous allocation. A pledge of more than a million GPUs over twelve months, announced at GTC 2026, was fully consumed in under five months.
On the same day's earnings call, Nvidia executives also pointed to a sharp rise in customer supply commitments over the past year, according to one account of the call. That characterization has not been checked against Nvidia's own transcript and is not the basis for the analysis that follows.
The AWS disclosure is the firmer data point, and it describes the condition Nvidia's executives were pointing to: demand for AI compute, and the memory that feeds it, is running ahead of what suppliers can turn around. That was the premise sell-side desks used Thursday morning to make the case for Micron, widely regarded as the most direct US-listed beneficiary of memory scarcity tied to AI infrastructure. Multiple desks argued Thursday that Micron stood to benefit directly from the same supply dynamics Nvidia and Amazon had just documented.
The stock did not cooperate. Micron shares fell 2.79% Thursday to close at $912.22, on a day when the broader technology sector was unusually strong. The Technology Select Sector SPDR fund gained 2.55%, and the S&P 500 rose 0.58%. Micron's decline against that backdrop is not a case of a stock drifting down with a weak market. It moved against a sector that was rallying, which makes the divergence real rather than incidental.
No Micron-specific negative catalyst turned up alongside the move. There was no downgrade, no guidance revision, no litigation disclosure, nothing in company filings that would explain a decline on a day its own sector was surging and its most closely associated supply narrative was being reinforced by Nvidia's own disclosures.
It would be easy to read the AWS numbers as unambiguous confirmation that memory suppliers should benefit, and much of Thursday's commentary did exactly that. Micron's stock priced something else. That could reflect valuation concerns after a run already priced with high expectations. It could also reflect positioning unwinds among traders who bought the scarcity story earlier, or skepticism about how cleanly Nvidia's GPU commitments translate into memory-specific revenue for Micron. Each explanation is plausible. None has verified support.
What is documented is narrower, and in some ways more useful to investors than a tidy explanation would be. The company at the center of the AI scarcity trade fell on the day the scarcity data got stronger, while the sector around it did not. That gap between the story and the tape sat unexplained at Thursday's close, on a stock the market has otherwise treated as the clearest read on how tight AI-driven memory supply has become.