Overview

• OpenAI will design the accelerators and systems while Broadcom develops, manufactures and deploys full racks with Ethernet-based networking for scale-out clusters.
• Deployments are scheduled to begin in the second half of 2026 and run through 2029 across OpenAI facilities and partner data centers.
• Financial terms were not disclosed, and Broadcom’s shares jumped roughly 10%, adding more than $150 billion in market value following the announcement.
• The plan targets about 10 gigawatts of capacity, underscoring massive power, cooling and data‑center buildouts as analysts warn of foundry, packaging, HBM, and financing constraints.
• The deal follows recent multi‑gigawatt commitments with Nvidia and AMD, and on Tuesday Oracle said it will deploy 50,000 AMD GPUs starting in the second half of 2026.