Intel will cut 15% of its global workforce and cancel planned projects in Europe, CEO Lip-Bu Tan told employees in a memo.
Tan described the restructuring efforts as “hard but necessary decisions,” emphasizing a move toward stricter financial discipline and more focused execution. In the memo, he detailed upcoming changes in how Intel invests, builds, and aligns its operations with customer demand.
Intel targets 75,000 headcount in 2025 restructure
“Thank you, sincerely, for your contributions,” Tan wrote to employees affected by the reduction.
The CEO noted that a significant portion of the layoffs had already been completed, with the remaining reductions expected through attrition and other measures. These cuts are part of a larger push to simplify Intel’s structure and sharpen accountability across teams.
As part of that effort, Intel has eliminated about 50% of its management layers. Tan said the changes are intended to accelerate decision-making and eliminate internal bottlenecks.
The tech company also plans to implement its return-to-office policy in September. Facilities are being upgraded to support full-capacity operations across Intel’s global sites, according to Tan.
Smarter spending, stronger chips, sharper AI focus
Aside from the job cuts, Tan also outlined three strategic priorities that will guide Intel’s reset: capital discipline, a streamlined x86 roadmap, and a more centered approach to emerging workloads.
Leaner, demand-aligned foundry model
The CEO indicated a move away from the company’s previous “build first” manufacturing approach. Future factory investments will be directly linked to confirmed customer demand, with capital spending tightly controlled. Projects in Germany and Poland have been canceled, construction in Ohio will slow, and packaging operations are being consolidated in Asia.
Future investments in nodes like Intel 14A will hinge on firm customer commitments, not forecasts. The immediate priority is scaling Intel 18A, with production focused on internal products and key partners like the US government. This contrasts with recent speculation that Intel might retreat from offering the node to external customers. Tan made clear that Intel will spend only where returns are certain.
Tighter grip on Intel’s core products
Tan is also putting Intel’s x86 chip business under direct scrutiny. The company is bringing back simultaneous multi-threading to address performance gaps in cloud and enterprise systems, a reversal meant to recover ground lost in recent years. Product plans are being stripped down and reset, with all major chip designs now requiring Tan’s personal approval before advancing.
Intel is betting on Panther Lake to anchor its notebook lineup, while Nova Lake remains in development for high-end desktop. On the server side, Granite Rapids is moving into production as the company works to regain share and stabilize leadership within the data center division.
Refined approach to inference and agentic AI systems
Intel is shifting away from training-centric AI strategies in favor of a more integrated approach that brings silicon, systems, and software into closer alignment. It will now prioritize areas where the company sees a competitive edge: inference and agentic computing.
Rather than leading with technology and chasing applications, Intel plans to start with customer workloads and build solutions backwards. The company says more details on the new approach will be shared in the coming months.
All eyes on Q3
Tan’s overhaul has touched nearly every corner of Intel in just months. The scope and pace mark a sharp break from years of drift.
Investors have responded cautiously. Shares are up about 13% this year, recovering slightly from a 60% drop in 2024; Intel’s worst year on record, according to CNBC.
But pressure is building. Reuters reports that the tech company is forecasting a third-quarter loss of 24 cents per share. Whether Tan’s reset delivers more than just cost-cutting is yet to be seen.
Read how a joint Intel–SoftBank venture is pushing memory chip innovation to support AI-era workloads.
