Intel delivered its fastest revenue growth in nearly 15 years Thursday, posting $16.1 billion in second-quarter sales and adjusted earnings per share of 42 cents — double the 21 cents analysts had penciled in. The chipmaker's data center business, riding the AI infrastructure boom, surged 59% to $6.3 billion and single-handedly carried the quarter.
The results mark a turning point for a company that spent most of 2024 and 2025 fighting to stay relevant as Nvidia swallowed the AI computing narrative. Intel's 25% revenue jump is its biggest since 2011, and the stock has already climbed over 170% so far this year — though it has slid 28% from June highs on profit-taking and jitters about sustainability.
Intel CEO Lip-Bu Tan put the AI opportunity front and center in the earnings statement. "AI is driving unprecedented demand for compute," he said. "As we continue to execute, Intel is well-positioned to capture sustainable growth across our CPU franchise." The message is aimed squarely at investors who have treated Nvidia as the only AI chip game in town — Intel wants them to remember that every AI workload still runs through a server CPU before it touches a GPU.
Data Center Demand Is Outstripping Supply
The data center group's $6.3 billion quarter was supported by what CFO David Zinsner described as a supply-constrained environment — customers are asking for more chips than Intel can produce. The company has started signing long-term agreements with server buyers, locking in pricing and volume commitments. Zinsner said Intel has already inked ten such deals.
"Customers continue to signal a strong and sustainable spending environment," Zinsner told analysts on the earnings call. That's a deliberate contrast with the narrative floating around the broader chip sector, where memory makers and some fabless designers have warned that AI demand may cool next year. Intel is betting the opposite.
Client computing, Intel's traditional engine, grew a more modest 13% to $8.9 billion. PC sales remain solid, but the company expects the third quarter to flatten because of a memory shortage that is squeezing supply chains across the industry. The data center business is where the growth story lives now.
Intel's gross margin recovery tells its own story. The company reported 42% gross margin for the quarter, up from just 2.5% a year earlier — a swing the company attributes to higher revenue volume, a richer product mix of server CPUs that carry better margins, and a disciplined pricing stance. The margin recovery is central to Intel's turnaround thesis: the company burned cash on manufacturing ramp-up for years, and investors are now watching to see if those investments translate into consistent profitability.
Foundry: Still Waiting for a Big Name Customer
Intel's foundry business, which the company is rebuilding from scratch under Tan's leadership to compete with TSMC, reported $5.8 billion in sales — up 31% year over year. That division will determine whether Intel can become a services company that manufactures chips for rivals or stays a captive foundry that makes only its own designs.
The foundry landed its first named external customer under Tan this week with Fortinet, the cybersecurity hardware maker. But the deal uses an older manufacturing node, not the bleeding-edge 14A process that Intel hopes will win business from AMD, Apple, and the hyperscalers. Speaking to CNBC, Zinsner said Intel's 14A process is ahead of where previous nodes were at the same stage of development, and that capital spending would increase "meaningfully" next year — most of it for factory tooling.
Intel has not yet announced a marquee foundry customer, and that silence continues to weigh on the stock. The company has the U.S. government as a 10% shareholder — a stake taken as part of the CHIPS Act push — but needs external commercial orders to prove its foundry model works. Analysts at Bernstein and Morgan Stanley have noted that Intel's foundry business needs at least two more major external commitments within the next 12 to 18 months before the unit can be valued as a standalone foundry rather than an internal manufacturing arm.
SMIC N+3 Shows China's Chipmaking Progress — With Big Caveats
On the other side of the Pacific, a teardown of Huawei's Kirin 9030 smartphone processor by SemiAnalysis revealed that SMIC's third-generation 7nm-class N+3 node has achieved a minimum metal pitch of 32.5 nanometers — narrower than the 36nm used in Intel's 18A process for some high-performance cells, and on par with TSMC's N6 transistor density of 113.4 million transistors per square millimeter.
The achievement is notable because SMIC does it without EUV lithography, relying instead on DUV multi-patterning and extensive design-technology co-optimization. But the caveats are large. The Kirin 9030's performance is roughly comparable to flagship phone processors from three years ago, and its energy efficiency trails Apple, Qualcomm, and MediaTek designs by a wide margin. Intel 18A also offers gate-all-around transistors, backside power delivery, and a much higher overall transistor density — roughly 38% denser than the SMIC node, according to SemiAnalysis.
The analyst firm projected that SMIC could continue scaling, with an N+4 node approaching TSMC N5-class density and an eventual N+5 with backside power matching Intel 18A. But density alone doesn't guarantee competitive performance or power efficiency, especially given the yield and cost penalties SMIC faces by squeezing density from 20-year-old DUV tools. The Chinese foundry's achievements are real, but they come with trade-offs that limit their practical value.
For the broader semiconductor industry, the two stories tell the same truth from opposite ends of the market: AI demand is reshaping manufacturing priorities, and the race to build denser, faster chips is accelerating everywhere — from Santa Clara to Shanghai.
What's Next for Intel
Intel guided for third-quarter revenue between $15.8 billion and $16.8 billion, above the $15.1 billion consensus, with adjusted EPS of 38 cents. "We are supply constrained," Zinsner repeated, signaling that the data center business has room to run.
The company is boosting capital expenditures, targeting a "meaningful increase" next year, as it aggressively tries to morph into a manufacturer of chips for other companies. Zinsner told CNBC's Kristina Partsinevelos that most of the spending would go toward factory tooling for the 14A process and beyond.
The open question is whether Intel can turn its foundry ambitions into real external revenue before the AI spending cycle peaks. Fortinet is a start, but investors want to hear that an AMD, an Apple, or an AWS will commit to Intel's 14A. Until that happens, the foundry story is a promise backed by government money — not a self-sustaining business.
Still, the Q2 numbers give Tan credibility. A 25% revenue jump and earnings that beat by a factor of two are hard to argue with, and Intel's product cycle is the strongest it has been in years. The company has customers lining up for server CPUs they can't get enough of, a foundry roadmap that the CEO says is ahead of schedule, and a balance sheet that is finally healing. The stock may be down 28% from its June peak, but the fundamentals are pointing in a direction the company has not seen since the data center boom of the late 2010s.
For more analysis of the companies reshaping semiconductor manufacturing, check our Semiconductors coverage for ongoing coverage of TSMC, Intel, and the global foundry race. Detailed financial filings are available at Intel's investor relations page. The SMIC teardown analysis was reported by SemiAnalysis and Tom's Hardware.