Mandeep Singh covers tech infrastructure and AI capex for Bloomberg Intelligence, analyzing hyperscaler spending and custom silicon.
no scored calls yet — needs a stated position or a categorical verdict, with a matured window vs SPY
SK Hynix ADR listing timing is perfect as longer context windows for frontier AI models drive sustained HBM demand. No new memory supply entrants expected soon due to sovereign-backed incumbents managing own fabs, creating structural supply constraint.
Singh argues AI models are becoming more compute, power, and memory intensive with parameter counts rising to 2.8 trillion, requiring hyperscalers to raise capex beyond $1 trillion collectively, and no smaller-model breakthrough has emerged to change this trajectory.
Cloud revenue remains the key metric for hyperscalers; Google Cloud growing faster than Azure and AWS picking up, with positive upward revisions expected as AI workloads drive cloud acceleration.