The investment conversation around artificial intelligence (AI) has been dominated by sophisticated software and the powerful graphics processing units (GPUs) that act as the brains of the operation. While these components are essential, a critical and potentially more durable investment opportunity is emerging from the physical layer of technology. A new bottleneck has appeared, not in processing power, but in the network's ability to connect thousands of these processors so they can function as a single, cohesive supercomputer.
Modern Generative AI and Large Language Models require a level of inter-processor communication that is unprecedented. The massive datasets involved in training these models mean that the speed of the network, the nervous system, is now a primary driver of performance.
