When your collateral is so cutting-edge that banks aren't sure how to price it, you get a 40% haircut on day one.
The Summary
- SoftBank slashed its planned margin loan from $10 billion to $6 billion after creditors balked at the OpenAI equity backing it
- The hesitation reveals how traditional finance still doesn't know how to value AI companies, even the most visible ones
- This is what happens when you try to use Web4 assets as Web2 collateral
The Signal
SoftBank wanted to borrow $10 billion against its OpenAI stake. Banks said sure, but only $6 billion worth. That 40% discount tells you everything about where we are in the collision between old money and new intelligence.
Margin loans are straightforward when the collateral is public stock. Apple shares have a ticker, liquidity, and a price that updates every millisecond. Banks know exactly what they're lending against. But OpenAI equity? There's no exchange, no daily mark-to-market, and no consensus on whether the company is worth $80 billion or $300 billion depending on which rumored valuation you believe.
"When your collateral is a private stake in the company building AGI, loan officers start sweating."
The creditors' hesitation isn't about OpenAI's technology or revenue. It's about:
- Valuation volatility in private AI companies with no public comps
- Illiquidity if SoftBank defaults and banks need to sell the stake
- Regulatory uncertainty around AI assets as collateral
Bloomberg Intelligence noted the creditor pushback, but the deeper story is structural. SoftBank is trying to unlock capital from what might be the most valuable private company in the world, and traditional finance is treating it like a subprime asset. Not because OpenAI is risky in the startup sense, but because banks don't have the frameworks to underwrite intelligence infrastructure.
This is the gap. AI companies are building the rails for Web4, the agent economy, the infrastructure that will run commerce and work for the next 20 years. But they're stuck in a pre-Web3 financing environment where ownership is opaque, liquidity is theoretical, and creditors price in massive discounts for anything they can't sell in 48 hours.
The Implication
Watch for two responses. First, AI companies will push harder toward tokenization and on-chain equity, because you can't get a 40% haircut on an asset with real-time pricing and global liquidity. Second, SoftBank and others will build shadow banking infrastructure purpose-built for AI collateral, because waiting for JPMorgan to understand transformer models is a losing strategy.
If you're building in agents or infrastructure, this is your financing landscape for the next 18 months. Plan accordingly.