Overview
- Hayes published an essay and discussed the thesis on podcasts arguing the AI buildout is a credit story like 2008 rather than an earnings bubble, and he links a policy-driven liquidity response to a potential bitcoin surge to $1 million.
- He cites large exposures tied to AI infrastructure, including roughly $1.09 trillion in data-center lease commitments and about $1.5 trillion of AI-related debt issued through mid-2026, figures reported with limited sourcing.
- Hayes highlights a financing mismatch where GPU loans are amortized over five to six years while GPUs materially lose value in about two years, which he says raises the risk of creditor losses and fire sales.
- He acknowledges great timing uncertainty for any unwind, saying it could happen quickly or take years, and says Maelstrom is positioning by building ETH exposure and selling out-of-the-money puts to manage downside risk.
- Market context remains cautious with bitcoin rangebound near $64,000 and other analysts warning this is a speculative scenario; the main near-term effects to watch are signs of stress in AI lending, policy backstop moves from Washington or Beijing, and shifts of capital into crypto and other nonbank assets.