OpenAI's latest funding round, reportedly the largest in startup history, has renewed serious questions about whether the company will ever reach profitability before its cash reserves run out. Meanwhile, a Fast Company analysis of critical minerals demand lays out exactly where that money goes: into data centers requiring lithium, cobalt, nickel, and rare earths at scale that current supply chains cannot sustain. The VC community is funding a machine that is simultaneously eating capital at the top and eating the earth at the bottom.

The Real Cost Structure of Generative AI

The standard cash-burn critique of OpenAI focuses on compute costs and model training runs. That is the visible burn. The invisible burn is the minerals extraction and energy infrastructure required to run inference at scale. A 2026 paper by Richard J. Mitchell in arXiv CS.CY frames AI-driven displacement as requiring genuine human oversight architectures. That framing applies to supply chains as much as labor markets. Nobody has built a governance layer for the cobalt that goes into the GPU that runs the model that generates the avatar that sells the product. The chain is long and almost entirely unaccountable. .

Carson Block's Short and What It Signals

Carson Block is betting against credit ETFs as a hedge against an AI meltdown, specifically HYG and LQD, on the thesis that AI-driven job losses will cascade into credit defaults. That is a downstream bet on the same phenomenon. The minerals get extracted, the data centers get built, the models get deployed, the jobs disappear, the credit deteriorates, Block collects. The question OpenAI's investors have not publicly answered is: at what point does the cost of the infrastructure required to reach profitability exceed the revenue that profitability could generate. The minerals crisis is not a policy footnote. It is a line item in that calculation.