Intel shifts manufacturing focus toward Xeon CPUs to meet data center AI demand.

Intel is reallocating foundry capacity from client chips to meet surging demand for Xeon processors used in AI servers, according to CFO David Zinsner's comments on the company's Q4 earnings call. Zinsner admitted that Intel misjudged demand for its datacenter products, leading to a capacity crunch during the quarter. Six months ago, "every hyperscale customer" was sending signals they planned to order a smaller number of high-core count chips, but demand for Intel's Xeon products increased considerably over the third and fourth quarters. The Xeon 6 platform is used extensively as the host CPUs in GPU systems like Nvidia's DGX B200 and B300, and in many AMD Instinct-based GPU boxes. To meet that demand, Zinsner said Intel is "shifting as much as we can over to the data center."

Intel is prioritizing higher-margin client products while deprioritizing low-end processors. "Within the client, we're focusing on the mid-and-high end, and [we're] not as focused on the low end. To the extent we have excess [capacity], we're pushing all of that into the datacenter space," Zinsner said. The company promises not to abandon its client business "completely" to chase AI revenues in the datacenter. Intel is not alone in reallocating capacity; major memory vendors including Micron, SK Hynix, and Samsung are shifting advanced manufacturing capacity to high-margin DRAM and high-bandwidth memory products used in AI servers and accelerators. Consumer memory prices have more than tripled over the past few months as a result. Zinsner noted that "industry-wide supply for key components like DRAM and substrates has come under increasing pressure due to intense demand to support the rapid expansion of AI infrastructure," and warned that "rising component pricing is a dynamic we continue to watch closely, especially relative to the client market, and could limit our revenue opportunity this year."

Zinsner expects Intel's capacity constraints to ease beginning in the second quarter of 2026, with improvements driven by higher yields and additional tooling coming online across its Intel 7, Intel 3, and Intel 18A process nodes. The ramp of the Core Ultra 3-series processors, codenamed Panther Lake, should also help, as these chips make heavier use of the latest generation 18A process node. Currently, that fab technology is only used by one Xeon product, Clearwater Forest, which is not commonly deployed in AI applications. Despite ongoing supply constraints, CEO Lip Bu Tan described Q4 as "another positive step forward," saying the company "delivered this result despite supply constraints, which meaningfully limited our ability to capture all of the strengths in our underwriting market."

Intel's financial performance remains challenging despite beating Wall Street predictions. In Q4, Intel posted a $591 million loss on revenues that declined 4 percent year-over-year to $13.7 billion. The Foundry division posted an operating loss of $2.5 billion for the quarter. However, compared to 2024's results, 2025 represented significant improvement; Intel lost $267 million on $52.9 billion in revenue, compared to an $18.8 billion loss the year prior. Intel forecast Q1 2026 revenue of between $11.7 and $12.7 billion.