AI slowdown calls justified but collapse of bubble may be more immediate threat
Scale of debt issuance being used to fund the hectic pace of the datacentre rollout by tech firms is a worry
TL;DR
- AI companies are raising concerns about the immense power of their products, alongside calls for regulation and limits on technological advancement.
- The AI industry's financial stability is questioned due to massive debt issuance for datacentre construction ($132 billion in one estimate) amidst fragile bond markets.
- The "unit economics" of AI are problematic, with falling prices for AI services (like per million tokens) and elevated costs for components like semiconductors.
- A "compute commencement wall" of $1.5 trillion in future obligations for AI labs is compared to the teaser mortgage rates that preceded the 2008 financial crisis.
- Datacenters are often built on "take or pay" contracts, creating significant future payment obligations for AI labs once services are utilized.
- The focus on AI safety and existential risks should not overshadow the potential for a financial collapse in the AI sector.
- Hyperscalers like Google, Amazon, Microsoft, Meta, and Oracle are heavily involved in datacentre rollouts funded by debt.