Tensor Processing Units
Coverage of Tensor Processing Units in the Nexus archive.
- DeepMind Chief Demis Hassabis says Google’s still winning AI talent
Google DeepMind CEO Demis Hassabis stated the company remains competitive in attracting AI talent despite recent departures of top researchers like Noam Shazeer and John Jumper. He emphasized Google's structural advantages, including data, hardware, and computing power, to retain researchers in the fiercely competitive AI job market.
- Google to sell its TPUs to some customers, who also fancy big-G GPUs
Google Cloud is selling its custom Tensor Processing Units (TPUs) to select customers to meet demand and diversify revenue, driven by AI's role in enhancing searches and ads.
- TorchTPU: Running PyTorch Natively on TPUs at Google Scale
Google has introduced TorchTPU, a framework that allows PyTorch to run natively on Google's Tensor Processing Units (TPUs). This collaboration between Google and the PyTorch team aims to enable large-scale machine learning workloads on TPUs, expanding their accessibility beyond TensorFlow.
- Google unveils two new TPUs designed for the "agentic era"
Google has introduced two new eighth-generation Tensor Processing Units (TPUs) tailored for the 'agent era' of AI, with TPU8t for training and TPU8i for inference. The chips aim to reduce training times for frontier AI models and offer a faster, more efficient alternative to competitors' hardware.
- Our eighth generation TPUs: two chips for the agentic era
Google Cloud has launched its eighth-generation Tensor Processing Units (TPUs), designed to optimize performance for the 'agentic era' of AI. The new TPUs feature two specialized chips aimed at enhancing machine learning workloads and infrastructure scalability.