Overview

• Ironwood, Google’s seventh-generation TPU, is designed specifically for inference computing, enabling rapid AI model execution for tasks like chatbot responses and recommendation engines.
• The chip delivers up to 4,614 TFLOPs of peak performance, features 192GB of dedicated RAM, and achieves bandwidth speeds of up to 7.4 Tbps.
• Ironwood offers double the performance per energy unit compared to Google’s previous Trillium chip, underscoring its energy efficiency advancements.
• The TPU will be available later this year in two configurations: a 256-chip cluster and a 9,216-chip cluster, tailored for scalable AI workloads via Google Cloud.
• Google’s development of Ironwood strengthens its AI and cloud service offerings as it competes with industry leaders like Nvidia, Amazon, and Microsoft in the AI hardware space.