Google on Wednesday unveiled its seventh-generation artificial intelligence chip named Ironwood,
a processor designed to accelerate the performance of AI applications. This new chip is particularly suited for "inference" computing — the kind of data processing needed when users interact with AI tools like OpenAI's ChatGPT. Inference involves performing quick calculations to produce responses in chatbots or generate other outputs, making Ironwood a crucial tool in the AI space.
This announcement marks a significant step in Google’s multi-billion-dollar AI chip initiative that has been ongoing for roughly a decade. It also positions the company as one of the few viable challengers to Nvidia’s powerful AI processors, which currently dominate the market. However, Google’s tensor processing units (TPUs), including Ironwood, are limited in availability — they can only be used by Google’s engineers or through its cloud platform.
In previous chip generations, Google created two types of TPUs: one optimized for training large AI models, and another that removed training features to reduce the cost of running AI applications. The Ironwood chip is specifically built for running these AI applications — also known as inference — and is designed to function in large-scale systems of up to 9,216 chips, according to Google Vice President Amind Vahdat.
Unveiled during a cloud computing conference, Ironwood combines features from previous chip types into a single design and includes more memory, making it even more capable of handling demanding AI services. Vahdat emphasized the growing importance of inference computing, saying, “It’s just that the relative importance of inference is going up significantly.”
Compared to Google's previous AI chip, Trillium, Ironwood offers twice the performance per unit of energy, which represents a major boost in efficiency. Google also uses its own chips, including Ironwood, to power its Gemini AI models. However, the company did not reveal which manufacturer is producing the Ironwood chip design.
Found this article interesting? Follow us on X(Twitter) and FaceBook to read more exclusive content we post.
