Google broadens its AI hardware portfolio with two new specialized chips

Google has developed two new tensor processing units (TPUs): TPU 8tx and TPU 8i Max, which are designed for confidential model training and more memory capacity, respectively. TPU 8tx is built with AMD and provides a custom inferencing network for running reasoning alongside training, and TPU 8i Max combines two TPU 8i chips to double memory capacity, processing power, and bandwidth. This move expands Google's chip portfolio and reduces its reliance on chips made by other companies for its cloud computing business. The new chips are part of Google's eighth generation of TPU chips, introduced in April, which offered more inference capacity and support for large training requirements. This expansion comes as hyperscalers like Microsoft, Amazon, and AI labs like OpenAI and Anthropic are working on their own AI-specific semiconductors to reduce their reliance on Nvidia.