SoftBank's proposed OpenAI margin loan has been scaled back from $10 billion to roughly $6 billion. The headline reads as a minor restructuring. It isn't.
Lenders are hesitating because private AI stakes are genuinely difficult collateral. There is no enforceable mark. There is no clear exit path. OpenAI CFO Sarah Friar responded to concerns by saying the company faces "a vertical wall of demand" — which is exactly what you say when the other side of the table is asking you to show them the conversion, not the pipeline. S&P Global has already moved SoftBank's outlook to negative, citing OpenAI-linked exposure as a potential drag on liquidity and asset quality. Internal reports flagged missed monthly sales targets earlier this year and a failure to hit one billion weekly ChatGPT users by end of 2025. None of that is disqualifying. All of it matters to a creditor pricing two-year paper with a one-year extension option.
The broader shift is real: AI's next funding phase runs through leverage, not just venture. SoftBank secured a $40 billion loan in March and committed another $30 billion to OpenAI. That is a lot of debt service resting on cash flow visibility that does not yet exist at the required resolution. What the lenders are doing right now — demanding tighter guardrails, shrinking the facility, asking for durable conversion evidence — is exactly what VCs have never had to do. The instrument has changed. The collateral hasn't caught up. The number I want to see is whether the final loan structure includes any cash flow triggers. That's where you find out what the lenders actually believe.