Google Launches Ironwood: The 7th-Gen TPU Powering the Future of AI Inference

At the 2025 Cloud Next conference, Google unveiled its most powerful AI chip to date — the seventh-generation TPU, codenamed Ironwood. Designed for advanced inference and generative workloads, Ironwood marks a significant leap forward in performance and efficiency, redefining what’s possible for cloud-based artificial intelligence.

Unprecedented AI Performance Gains

Compared to the first-generation TPU introduced in 2018, Ironwood delivers 3,600 times more inference performance and 29 times better efficiency. It represents a complete rethinking of Google’s AI hardware, optimized not only for training large models but also for real-time inference and intelligent decision-making.

Google claims that Ironwood’s performance exceeds that of the world’s fastest supercomputer, El Capitan, by over 24 times when deployed at full scale.

Key Specifications and Architecture

Each Ironwood TPU chip is packed with cutting-edge capabilities:

192GB High-Bandwidth Memory (HBM)

4,614 TFLOPs of peak performance per chip

1.2 Tbps ICI (Inter-Chip Interconnect) bandwidth

Compared to its predecessor, Trillium, Ironwood is twice as energy-efficient, enabling superior compute density for complex AI workloads.

Massive TPU Pods for Enterprise-Scale AI

Google is offering Ironwood in two TPU Pod configurations:

256-chip pod for mid-scale AI deployments

9,216-chip pod for ultra-large-scale AI workloads

The largest pod delivers 42.5 ExaFLOPs of compute power, enabling organizations to run the most demanding AI models with unprecedented speed and scale.

From Reactive to Proactive AI

Ironwood isn't just about raw power—it reflects a shift in AI strategy. Google is designing its TPUs to support "proactive AI agents" capable of not only answering questions but anticipating user needs and autonomously generating insights. This makes Ironwood ideal for next-gen AI use cases in enterprise analytics, cloud services, and autonomous systems.

With Ironwood, Google reinforces its position as a global leader in AI infrastructure, offering unmatched performance and scalability for cloud customers preparing for the AI-driven future.