Overview

• Intel will create custom x86 CPUs for Nvidia’s AI infrastructure and develop x86 PC system-on-chips that embed Nvidia RTX GPU chiplets.
• Joint products will integrate Nvidia’s NVLink into Intel CPU designs to deliver higher bandwidth and lower latency than PCIe connections.
• Nvidia will continue manufacturing its GPUs at TSMC and made no commitment to use Intel Foundry, though Intel will provide packaging and CPUs for the joint systems.
• Engineering teams have worked together for about a year on multiple product generations, and Intel says the collaboration complements its roadmap as it maintains its own GPU offerings.
• Industry analysts say the partnership could improve customer commitments and economics for Intel’s future nodes, notably 14A, while first integrated client chips are estimated to be three to four years from volume.