Janus Henderson’s Denny Fish on AI: We’ll continue to see models ‘leapfrogging each other'

Market Trends & Competition - The AI sector is characterized by a race to the frontier, with companies like Google, Meta, Anthropic, OpenAI, and Microsoft competing to achieve artificial general intelligence [2] - Model leapfrogging is expected to continue with each iteration of GPT, Gemini, Claude, and XAI [2] - Investors should be concerned if models stop leapfrogging each other, as it would suggest diminishing scaling laws and potentially slow infrastructure demand [4] Infrastructure & Investment - Remarkable Nvidia earnings and Gemini 3 imply that scaling laws are firmly intact, leading to aggressive demand for GPU/TPU accelerators during the training phase [2] - Hyperscalers continue to raise capex expectations, indicating sustained investment in infrastructure [6] - The infrastructure build is expected to be prolonged due to natural governors such as data center construction speed, permitting, and power availability [6] - Every GPU from TSMC's foundry has high demand, indicating that the deployment of these assets into data centers is just beginning [6] Risk Assessment - The market may take a breather after a strong infrastructure trade, but recent data has not changed expectations for the infrastructure build [5][6] - Short-lived assets like GPUs/TPUs are not deployed until the data center shell is up and power is available due to high carrying costs [6]