Bloomberg: 'Why Fears of a Trillion-Dollar AI Bubble Are Growing' (finance.yahoo.com)

Bloomberg warns the AI boom may be turning into a trillion-dollar bubble as tech giants and startups race to build vast compute and data-center capacity. Headlines like OpenAI’s $500 billion “Stargate” plan and Nvidia’s up-to-$100 billion support for OpenAI sit alongside massive debt raises (Meta’s $26B loan, a $22B facility for Vantage) and venture financing. Analysts estimate AI infrastructure could demand $2 trillion in annual revenue by 2030, while Bain predicts an $800 billion shortfall; OpenAI reportedly expects to burn ~$115 billion by 2029. The funding mix—chips, accelerators, hyperscale centers—has spawned unconventional deals that some fear primarily prop up hardware suppliers rather than profitable AI businesses. The concern is not just scale but payback. Academic studies (MIT; Harvard/Stanford) show poor ROI or “workslop” from many AI deployments, and developers report diminishing returns from scaling laws that once promised steady improvements with more compute and data. Releases like GPT-5 drew mixed reviews and admissions from leaders that “we’re still missing something” on the path to AGI. Add rising low-cost competition from China, strained power grids for data-center growth, and opaque financing structures, and the result could be significant capital destruction even as proponents (OpenAI, Anthropic, Meta) argue models already boost productivity and will eventually justify the spending.