All Eyes on Markets for AI Bubble Watch: Is It a Floater or a Popper? (www.theregister.com)

The market is flashing both fascination and fragility around AI. Recent employee share sales put OpenAI at a nominal $500 billion despite no profit, while GPU-cloud provider CoreWeave disclosed an incremental $3 billion loan tranche on top of more than $25 billion of debt and equity raised to build AI infrastructure. Oracle says it has a $455 billion spending pipeline from datacenter customers (OpenAI alone is tied to roughly $300 billion of that capacity), and analyst projections suggest companies will need to borrow tens to hundreds of billions—KeyBanc estimates OpenAI may need ~ $100 billion over four years, and others peg Oracle-style borrowing at ~$25 billion annually. Bain warns the infrastructure bill could imply $2 trillion of revenue needed by 2030, and Moody’s flagged significant counterparty risk where buyers lack the cash to fund contracted capacity. For AI/ML practitioners this matters because the wave of model-driven products hinges on massive GPU/cloud investment and complex financing, not just engineering. Banks and rating agencies are debating whether today’s froth is a benign hype cycle or a leveraged bubble; Barclays and Gartner say it’s “frothy” but not yet a bubble, while predicting consolidation and the disappearance of many small model builders. Technical implications: hyperscaler capex has risen (capex/sales ~25%), operational costs spike with a talent war, and private debt is increasingly underwriting data-center expansion—any funding shock could slow deployment, push consolidation, and reshape where and how models get built and served.