Analysis
Chip stocks extended a two-day rally on September 17, with Intel shares up 7.67%, AMD gaining 6.36% and Nvidia adding 2.54%, according to reporting from TipRanks. The move followed a Wednesday session in which the same names had already climbed as investors moved past a recent bout of concern about a slowdown in AI infrastructure spending -- concern that, a week earlier, had been dragging chip names lower on fears the AI buildout was outrunning demand.
Two specific catalysts drove the Thursday leg of the rally. First, Intel's CEO flagged severe CPU supply constraints and warned that a broader memory shortage could worsen heading into next year -- a warning markets read as evidence of pricing power and tight capacity rather than pure downside risk. Second, reports surfaced of exploratory talks between Intel and SK Hynix, fueling speculation about a manufacturing or supply partnership that would strengthen Intel's position in memory and advanced packaging at a moment when AI compute demand keeps requiring more of both. A broader pullback in Treasury yields the same week added a macro tailwind, lifting rate-sensitive growth and semiconductor names across the board rather than rewarding any single company's news in isolation.
The rally is a useful data point next to this week's other big infrastructure story: Crusoe's initial close of a $3.9 billion Series F at a $30.9 billion valuation, covered elsewhere in this issue, is built on the same underlying bet that chip and compute demand keeps climbing. When public-market chip investors and late-stage AI infrastructure investors are both leaning into the same demand assumption in the same week, it's either strong corroborating signal or a sign the whole market is pricing in the same optimistic scenario simultaneously -- and those two readings look identical until demand data actually turns.
The gap in the rally's logic: a memory shortage and supply-constraint story is bullish for chipmakers' near-term pricing but bearish for the AI companies that have to buy that memory and compute at higher prices, and none of Thursday's coverage addressed how quickly rising input costs might flow through to squeeze margins at the AI labs and infrastructure operators actually building on top of these chips.