Issue Chips optimised for immersion rather than for training

Summary

Next-generation digital realities aim to feel seamless, realistic and responsive, blurring the line between physical and digital, and that requires processors designed for the task. Tensor processing units have drawn attention because they use energy efficiently and depend less on critical minerals than conventional graphics processors. TPU production is projected to more than double by 2028, and AI companies such as Anthropic, which alone is buying about a million units, are considering them. Yet TPUs are better suited to training AI models than to running real-time immersive experiences, which need varied inputs and latency below 20 milliseconds. This leaves an opening for advanced GPU chips that combine low latency, the ability to interpret neural signals and hardware-based security to keep immersion intact.