Google launches dedicated training and inference AI chips in challenge to Nvidia

The TPU 8t targets model training while the TPU 8i handles inference workloads, a split from previous generations that handled both tasks

Go to News Site