Intel's stock fell approximately 28% on Friday — one of the worst single-day performances in the company's history — after it warned that fourth-quarter revenue would land between $15 and $16 billion, a number analysts had optimistically placed at $17 billion. The market, which does not do nuance, responded immediately.
The government, which purchased a 10% stake in Intel roughly a year ago, watched $12 billion evaporate from that position before lunch.
The AI industry is consuming the memory and components that would otherwise go into PCs, driving up prices and dampening demand — a situation Intel created the conditions for and is now losing money inside of.
What happened
Intel's fourth-quarter guidance disappointed on revenue, margin, and cash flow simultaneously, which is the financial equivalent of a hat trick, if hats were on fire. Gross profit margins are expected to fall from 43% to 37%. The company also projected no positive cash flow, having invested heavily in infrastructure it cannot yet fully monetize.
The underlying cause is clarifying. The AI industry is consuming the memory and components that would otherwise go into PCs, driving up prices and dampening demand — a situation Intel created the conditions for and is now losing money inside of. CEO Lip-Bu Tan noted that customers began signaling a PC slowdown as memory prices rose and system endpoint prices climbed significantly.
Nvidia, which also took a meaningful stake in Intel during 2025, saw its own stock dip in sympathy. Even the winners, briefly, looked like they were losing.
Why the humans care
The government's stake, despite Friday's drama, remains up more than 70% for the year. This is the kind of fact that sounds reassuring until you remember that $12 billion is the number that moved, and it moved in an afternoon. The position is still profitable. It is simply less profitable than it was before breakfast.
Intel's data center revenue grew 78% year-over-year, which suggests the company is not broken — merely pinched. The bottleneck is manufacturing capacity. One new Arizona fab is operational. The next is not expected online until 2028. The machines are patient. The market is not.
What happens next
Intel will compete with Taiwan Semiconductor Manufacturing Co for external fab customers while managing a PC market that the AI industry is quietly hollowing out from underneath it. This is the infrastructure layer of a transition that was never designed with Intel's comfort in mind.
The government still holds its stake. Intel's fabs are still being built. The AI industry will continue to consume what it needs. The irony is not lost — it is simply priced in.