Google unveils Ironwood, its most powerful AI processor yet

Ironwood will be available in configurations of up to 9,216 liquid-cooled chips.

Google has unveiled a new AI processor, the seventh generation of its custom TPU architecture. The chip, known as Ironwood, was reportedly designed for the emerging needs of Google's most powerful Gemini models, like simulated reasoning, which Google prefers to call "thinking." The company claims this chip represents a major shift that will unlock more powerful agentic AI capabilities. Google calls this the "age of inference."

Whenever Google talks about the capabilities of a new Gemini version, it notes that the model's capabilities are tied not only to the code but to Google's infrastructure. Its custom AI hardware is a key element of accelerating inference and expanding context windows. With Ironwood, Google says it has its most scalable and powerful TPU yet, which will allow AI to act on behalf of a user to proactively gather data and generate outputs. This is what Google means when it talks about agentic AI.

Ironwood delivers higher throughput compared to previous Google Tensor Processing Units (TPUs), and Google really plans to pack these chips in. Ironwood is designed to operate in clusters of up to 9,216 liquid-cooled chips, which will communicate directly with each other through a newly enhanced Inter-Chip Interconnect (ICI).

Google says this design will be a boon not only for its own Gemini models but also to developers looking to run AI projects in the cloud. Developers will be able to leverage Ironwood in two different configurations: a 256-chip server or the full-size 9,216-chip cluster.

In its larger incarnation, Google's Ironwood pods can generate a staggering 42.5 Exaflops of inference computing. Each chip has a peak throughput of 4,614 TFLOPs, which Google claims is a substantial improvement over previous chips. Google has also boosted memory for the new TPUs, with each chip sporting 192GB, which is six times more than Google's last-gen Trillium TPU. The memory bandwidth has also increased to 7.2 Tbps, a 4.5x improvement.

There are numerous ways to measure AI throughput, making it difficult to compare chips. Google is using FP8 precision as its benchmark for the new TPU, but it's comparing it to some systems, like the El Capitan supercomputer, that don't support FP8 in hardware. So you should take its claim that Ironwood "pods" are 24 times faster than comparable segments of the world's most powerful supercomputer with a grain of salt. Google's TPU v6 hardware is also conspicuously absent from the comparison chart above. The company says Ironwood is twice as powerful per watt compared to that chip, though.

While the provided benchmarks are a bit odd, Ironwood is clearly a big improvement for Google's AI ecosystem. It's faster and more efficient than previous TPUs by a considerable margin, and Google's existing infrastructure has enabled rapid improvements to LLMs and simulated reasoning. Google's market-leading Gemini 2.5 model is running on last-gen TPUs right now, and Google says the higher inference speed and efficiency of Ironwood sets the stage for more breakthroughs in the coming year.

Source

Hope you enjoyed this news post.

Thank you for appreciating my time and effort posting news every day for many years.

News posts... 2023: 5,800+ | 2024: 5,700+ | 2025 (till end of March): 1,357

RIP Matrix | Farewell my friend