Intel is undergoing a sweeping strategic overhaul under CEO Lip-Bu Tan, signaling a decisive break from legacy structures and a bold push to reposition the company at the center of the AI revolution. With revenue declining across multiple business units in recent years, Tan has launched broad-based layoffs, project rationalizations, and a reset of manufacturing priorities—moves aimed at boosting efficiency and restoring competitiveness.
Central to Tan's vision is a recalibrated AI strategy that goes beyond GPU-centric thinking. Instead, Intel is doubling down on advanced process nodes, system-level integration, and software optimization to regain relevance in a rapidly evolving semiconductor landscape.
AI blueprint anchored in advanced process Nodes
At the core of Intel's renewed ambition lies its 18A process technology, slated to enter volume production in 2025. The node is designed to power three generations of internal products from 2026 through 2028. The company is also investing in the development of its 14A node, expected to align with TSMC's A14 timeline in 2028–2029.
While Intel has left the door open to delay or cancel the 14A node if major clients do not commit to production volumes, the company remains bullish on 18A as the long-term foundation of its foundry strategy. Intel expects peak 18A capacity to arrive by early 2030, positioning the node as a magnet for external customers and a cornerstone of its future competitiveness.
Moving beyond the GPU battlefield
During Intel's recent earnings call, analysts inquired whether the company plans to compete directly with Nvidia and AMD in the GPU market. In response, Tan indicated that the company will release a comprehensive AI roadmap in the coming months. The roadmap will emphasize inference workloads and advancements in agentic AI, a technology that prioritizes speed, accuracy, and autonomous decision-making.
Rather than centering its strategy on GPU supremacy, Intel plans to harness the combined strengths of its x86 CPUs and AI accelerators to build next-generation computing platforms. The company is also expanding its pool of software engineers to bolster AI capabilities and enhance performance across the full technology stack.
Open to ASIC, committed to customization
Pressed further on whether Intel would consider ASICs to boost its AI market standing, Tan emphasized a flexible, full-stack philosophy. Intel will capitalize on the strength of its x86 architecture to optimize workloads while remaining open to alternative architectures. He added that Intel plans to co-develop custom AI platforms with systems integrators to address specific performance needs.
Although traditional segments remain under pressure in the third quarter of 2025, Intel is signaling a profound strategic pivot anchored in three focus areas:
• Advanced nodes centered on 18A and its performance-enhanced sibling, 18A-P
• Continued development of the 14A node to support internal IP and advanced packaging collaboration
• Revival of simultaneous multithreading (SMT) to boost system-level integration in AI applications
Intel has already begun wafer production on 18A at its Fab 52 facility in Arizona. Meanwhile, 14A development will support in-house IP and integrate seamlessly with Intel's EMIB and Foveros packaging technologies, enabling a "logic-plus-packaging" platform designed for differentiated performance in data center and AI training environments.
Software-driven stack for AI synergy
Intel is working to integrate its hardware and software into a unified platform that combines chips, systems, and development frameworks. The company aims to tighten the alignment between its Gaudi, Xeon, and forthcoming Panther Lake processors with software tools such as OpenVINO and OneAPI, targeting more efficient workload management and faster AI inference.
This system-level approach, coupled with advances in process technology and AI-focused packaging, underpins Intel's broader AI strategy. The company is moving beyond its legacy role as a CPU vendor toward becoming a full-stack AI solutions provider, as it faces mounting competition from Nvidia, AMD, and the growing adoption of Arm-based architectures.
Source: Intel; compiled by DIGITIMES, August 2025
Source: Intel; compiled by DIGITIMES, August 2025
Article edited by Jack Wu
