Intel’s data center revenue climbed 59 percent last quarter, the clearest sign yet that the AI infrastructure buildout is now reaching server CPUs, not just GPUs and other specialized accelerators. The chipmaker booked $16.1 billion in overall second-quarter revenue, beat consensus on both sales and profit, and posted total growth of 25 percent, the fastest quarterly expansion Intel has produced since 2011. Investors did not reward it: the stock, up more than 170 percent so far this year, sold off on Friday following the Thursday report, extending a 28 percent slide in July alone.

Adjusted earnings per share came in at 42 cents, roughly double the 21 cents Wall Street had modeled, according to LSEG consensus estimates. For the current quarter, Intel guided to adjusted EPS of 38 cents on revenue between $15.8 billion and $16.8 billion, above the $15.1 billion analysts had penciled in.

The growth is uneven across Intel’s business. Revenue in the data center group rose to $6.3 billion, while the older client computing unit, which still supplies most of Intel’s total revenue through PC chips, grew a comparatively modest 13 percent to $8.9 billion. Intel said it expects flat PC sales next quarter because of a memory shortage.

The more consequential disclosure was about how Intel is selling its server chips, not how many it sold. The company said it has signed 10 long-term agreements with data center customers, some locking in pricing and others locking in volume, years ahead of delivery. Chief financial officer David Zinsner said Intel is supply constrained: it cannot manufacture enough server processors to satisfy current customer demand. “Customers continue to signal a strong and sustainable spending environment,” Zinsner told analysts on the earnings call.

Multi-year supply contracts are standard practice in the GPU market, where major buyers book capacity a year or more in advance to guarantee allocation. Seeing that same behavior show up in general-purpose server CPUs is the more telling signal in this earnings report. The AI-driven compute shortage is not confined to accelerators; it now reaches into the conventional processors running the rest of a data center.

Intel’s foundry business, the unit that manufactures chips for outside customers, posted $5.8 billion in revenue, up 31 percent year over year. Fortinet became the foundry’s first publicly named customer under chief executive Lip-Bu Tan earlier in the week, though that deal uses an older process for security chips rather than Intel’s leading-edge manufacturing. Zinsner said the 14A node, Intel’s newest, is progressing faster at this stage of development than prior process generations did.

Gross margin recovered to 42 percent, up sharply from 2.5 percent a year earlier, which Intel attributed to higher volume and a richer mix of higher-margin chips. The company also plans a “meaningful increase” in capital spending next year, mostly for factory tooling, a bet that current demand holds.

None of this fully explains why the stock fell after the report. Intel shares are up more than 170 percent this year, after climbing 84 percent last year when the U.S. government acquired a 10 percent stake in Intel to support domestic chip manufacturing. A run of that size already prices in a great deal of successful execution on the foundry ramp, the 14A node, and landing outside customers Intel has so far failed to name. Being supply constrained cuts both ways. It confirms real demand, but it is also an admission that Intel’s manufacturing has not kept pace with the order book it now says it holds.

For buyers of server compute, capacity planning now matters as much as pricing. Any team assuming it can source Intel server chips on short notice should ask its supplier whether Intel’s 10 long-term agreements have already claimed the volume it is counting on.

Reported by CNBC on July 23, 2026.