Arthur Hayes: The AI Crash That Could Send Bitcoin to $1M

The announced 60‑day window to finalize the Iran‑related deal is largely a near‑term diplomatic signal—critical specifics such as nuclear‑material handling, timing for reopening the Strait of Hormuz, and environmental fee/enforcement remain TBD—which leaves oil market risk elevated and supports the view that geopolitical shocks could quickly push prices higher.

Major drawdowns of strategic oil reserves by the US, China and Japan create a sustained physical bid as those buyers rebuild inventories, and speakers flag a scenario where restocking plus geopolitical stress could drive WTI toward roughly $120/barrel within months to a year, a level likely to feed through to inflation with about a three‑month lag.

AI economics are fundamentally energy‑driven: differences in power costs (notably China's lower cost to build electricity capacity) and the ability to self‑host open‑source Chinese models on bare metal create a structural cost advantage that risks commoditizing premium US AI offerings.

Current subscription models that hide marginal compute costs are effectively subsidizing usage; a shift to API or pay‑per‑token pricing will raise unit costs, reduce usage, and force labs and customers to focus on fundamentals—exacerbating credit strain when coupled with long GPU financing/amortization schedules.