Google has unveiled Ironwood, its seventh-generation AI chip, which the company said is designed to handle the most demanding AI inference workloads at scale.
At Google Cloud Next 25 yesterday (April 9), Google said the new Ironwood tensor processing unit (TPU) represents a “significant shift in the development of AI” and the infrastructure that powers its progress.
“Ironwood is our most powerful, capable and energy-efficient TPU yet. And it’s purpose-built to power thinking, inferential AI models at scale,” said Amin Vahdat, vice president and general manager of machine learning at Google’s Systems and Cloud AI division, in an accompanying blog post.
“For more than a decade, TPUs have powered Google’s most demanding AI training and serving workloads and have enabled our cloud customers to do the same.”
The Age of Inference
According to Google, Ironwood represents a shift from responsive AI models that provide real-time information for people to interpret to models that proactively generate insights and interpretation.
“Ironwood is built to support this next phase of generative AI and its tremendous computational and communication requirements,” the search giant said.
One of several new components in Google Cloud AI Hypercomputer architecture, Ironwood scales up to 9,216 liquid-cooled chips linked with Inter-Chip Interconnect (ICI) networking spanning nearly 10 MW.
Each chip delivers a peak performance of 4,614 teraflops. When scaled to 9,216 chips per pod for 42.5 exaflops, Ironwood is said to deliver more than 24 times the compute power of the world’s largest supercomputer, El Capitan.
Google Ironwood: Key Features
Key features of Google Ironwood include:
-
Significant performance gains, with a focus on efficiency. Ironwood’s performance per watt is 2x that of Trillium, the sixth generation TPU announced last year.
-
Increased high-bandwidth memory (HBM) capacity. Ironwood offers 192 GB per chip, 6x that of Trillium.
-
Improved HBM bandwidth, reaching 7.2 TBps per chip. This high bandwidth ensures rapid data access for memory-intensive AI workloads.
“Ironwood represents a unique breakthrough in the age of inference with increased computation power, memory capacity, ICI networking advancements and reliability,” Vahdat said.
“These breakthroughs, coupled with a nearly 2x improvement in power efficiency, mean that our most demanding customers can take on training and serving workloads with the highest performance and lowest latency, all while meeting the exponential rise in computing demand.”
Read more of the latest data center hardware news
The AI Chip Race Heats Up
Google’s Ironwood announcement is the latest in a string of next-gen chip launches aimed at powering large-scale AI workloads.
Last month, at GTC 2025, Nvidia CEO Jensen Huang outlined the chip giant’s AI vision, unveiling new supercomputers and software to power next-gen workloads. These include the new Blackwell Ultra AI chip and Vera Rubin processors.
In February, Intel expanded its family of Xeon 6 processors with new high-performance chips designed for enterprises with compute-intensive needs, such as AI, virtualization, and databases.
Microsoft, meanwhile, recently announced Majorana 1, its first quantum computing chip that’s said to mark a major step in the company’s effort to produce devices that might someday solve problems beyond the reach of modern computers.
