Particle.news
Download on the App Store

OpenAI, Foxconn Strike U.S. Manufacturing Pact for AI Data-Center Hardware

The collaboration targets localized supply chains to support OpenAI's massive compute buildout.

Overview

  • Under the agreement, the companies will co-develop multiple generations of AI servers and produce power, networking, cooling and cabling components at Foxconn’s U.S. facilities.
  • OpenAI will get early access to evaluate the systems with an option to purchase, and the initial pact contains no purchase commitments or financial obligations.
  • Foxconn says it can currently manufacture about 1,000 AI server racks per week and expects output to increase next year.
  • Chairman Young Liu outlined plans to invest $2 billion to $3 billion annually in AI for the next several years as the cloud and networking unit surpasses consumer electronics.
  • OpenAI has signaled roughly $1.4 trillion for about 30 gigawatts of compute capacity and is expanding hardware efforts alongside partners including Nvidia and Broadcom.