Google reveals Ironwood: new AI chip brings huge leap in computing power

Google finally released his new AI chip called Ironwood. This is a TPU (Tensor Processing Unit) which is mainly intended to run existing AI models at a rapid rate. The chip was announced earlier this year, but is now really available.

The Ironwood chip is part of a larger system where thousands of these chips work together, over nine thousand in total. They are all connected to each other via a huge fast connection and have access to huge amounts of memory. According to Google, this together produces a computing power that is many times higher than what the competition can offer at the moment.

Compared to the previous generation, the TPU 6, Ironwood is so’Four times faster. This applies not only to performing AI tasks, but also to training new models. The improvement is not only in the hardware, but also in the software that Google built around it.

A large customer of this new technology is Anthropic, the company behind the AI assistant Claude. From 2026 they have access to an enormous amount of computing power. Google talks about more than a gigawatt of capacity, which shows how big the stakes are.

Next to Ironwood, Google also comes up with new servers based on the Arm architecture, the so-called “Axion”-Instances. They are intended for more general applications and, according to Google, they prove to be much faster and efficient in practiceënter work. For example, companies like Vimeo and ZoomInfo saw noticeable improvements in performance and cost.

Google tries with all this a kind of complete “AI supercomputer” (Parliament adopted the resolution) Not only the chips, but also the infrastructure around them, such as liquid cooling and smart network connections, should ensure that everything is stable, efficientënt and reliable running. The company even says that these systems have hardly had downtime since 2020.

In short: Ironwood is Google’s latest step in scaling AI power. Faster, more efficientënter, and above all designed to meet the growing demand for AI applications from large customers.