This isn't another 'AI bubble bad' post. The article traces a specific financial contagion pathway that hasn't been covered elsewhere in a single piece. Tech companies are moving hundreds of billions in AI debt off their balance sheets into special purpose vehicles. That debt gets rated investment grade, securitised, and sold to pension funds and insurance companies. The Bank of England's December 2025 Financial Stability Report explicitly flags this as a financial stability risk, comparing AI valuations to the dot-com bubble. Mercer, the UK's largest pension advisor, is warning defined benefit schemes about concentration risk and comparing the situation to the early 2000s telecom bust. The collapse-relevant point: nobody can actually quantify how much pension money is exposed, because the entire structure is designed to be opaque. When AI revenue projections fail to materialise, the debt doesn't disappear. It sits in the retirement savings of ordinary workers who have no idea they're exposed. The article traces the full chain from SPV creation to bond index to auto-enrolled workplace pension. This is a documented mechanism by which a tech correction could directly degrade the material conditions of millions of people.
DHH argued Facebook couldn't monetise. I'm not arguing Anthropic can't monetise. I'm arguing the debt structure financing AI infrastructure creates systemic risk regardless of whether individual companies succeed. Cisco survived the dot-com bust. The bondholders who financed the fibre didn't