$DATACENTER - Data centers are becoming increasingly scarce as a bottleneck for AI training, with hyperscalers struggling to find facilities large enough to train models with sufficient GPUs.
$INFRA - Better networking infrastructure between large data centers will be needed in the future as GPU scarcity and data center constraints become more acute for AI model training.
$ALTDATA - Alternative data combined with AI is enabling investment firms to generate alpha by synthesizing unstructured information from podcasts, broker research, and other non-traditional sources that were previously difficult to access at scale.
$AIMODELS - AI models are demonstrating exponential capability growth with agentic workflows doubling in task duration every seven months, fundamentally changing how investment firms operate and enabling 86x growth in token consumption since January.