Google has introduced its new processor Ironwood, which takes its hardware power in the field of artificial intelligence to the next level. This processor was developed specifically to meet the high processing power demand of the company’s own artificial intelligence model Gemini.
Google has introduced its most powerful artificial intelligence processor
Google emphasizes that artificial intelligence has evolved into a structure that not only answers questions but also makes decisions and collects data on its own. This processor will power new generation artificial intelligence systems that can process on behalf of the user, collect information and make independent decisions.

Ironwood will have an inference processing capacity of up to 4,614 TFLOPs per second. In addition, these processors can communicate directly with each other thanks to the new Inter-Chip Interconnect (ICI) technology developed by Google.
Each of the chips, which is liquid-cooled, can work together in clusters of up to 9,216. Thus, a total of 42.5 Exaflops of computing power can be reached. This figure was recorded as the highest hardware processing capacity ever announced in the field of artificial intelligence.
Ironwood also makes a big leap forward in terms of memory capacity compared to previous generations. Each chip has 192 GB of memory. This is a 6x increase over the previous generation TPU, Trillium.
Memory bandwidth will be at 7.2 Tbps. This figure is a 4.5x improvement over the previous generation. Google stated that Ironwood is twice as efficient as Trillium in terms of performance per watt.
The introduction of Ironwood once again demonstrates that the trend towards reducing the dependence of big tech companies on Nvidia continues. What do you think about this? Share your thoughts with us in the comments section below.