Intel

Intel Corporation (NASDAQ: INTC) was founded in 1968 by Gordon Moore and Robert Noyce and for five decades dominated the PC and server CPU market. Today Intel occupies a unique and challenging position in the AI chip landscape: it is simultaneously a chip designer (competing with NVIDIA and AMD in AI accelerators), a foundry operator (Intel Foundry Services competing with TSMC), and an integrated device manufacturer (IDM) for its own CPU production. Intel's primary AI accelerator product is the Gaudi 3, announced in 2024. Gaudi 3 is manufactured by TSMC on the N5 (5nm-class) process — not Intel's own foundry — and targets AI training and inference workloads as a lower-cost alternative to NVIDIA H100/H200. It integrates 128GB HBM3 memory from SK Hynix and achieves competitive matrix multiplication throughput, though software ecosystem (Intel oneAPI vs. NVIDIA CUDA) remains the primary adoption barrier. For x86 compute, Intel's Xeon Scalable 4th and 5th generation (Sapphire Rapids, Emerald Rapids) processors include Advanced Matrix Extensions (AMX) for AI inference acceleration, positioning Intel CPUs as inference platforms for smaller models without requiring GPU add-ins. The Xeon Max series includes on-package HBM for bandwidth-sensitive workloads. Intel's packaging supply chain is significant: Ibiden and Shinko Electric supply FC-BGA substrates for Intel's high-volume Xeon CPUs, and Intel operates its own advanced packaging facilities (EMIB, Foveros) to integrate heterogeneous chiplets. This packaging capability is one area where Intel has genuine differentiation from fabless competitors. The company's dual role as IDM and foundry creates both strategic opportunities (captive manufacturing) and tensions (customers reluctant to use a competitor's foundry). Intel's foundry ambition — ramping Intel 18A (1.8nm-class) by 2025 — is the key variable that will determine whether Intel regains process leadership or remains a TSMC customer for its most advanced products.