SoftBank Group has scaled back plans for a margin loan backed by its stake in OpenAI, cutting the target size from an original $10 billion to as little as $6 billion after lenders raised concerns about the structure of the deal and the challenge of valuing the privately held ChatGPT maker, Bloomberg reported on May 8, 2026.
The 40% reduction is one of the clearest signs yet that creditors are growing cautious about underwriting debt secured by paper gains in unlisted AI companies, even as the sector continues to attract record primary investment. Discussions remain ongoing and the eventual size of any borrowing could still change.
Lenders push back on private valuations
According to Bloomberg, banks and private credit funds raised concerns about how to assign a reliable valuation to OpenAI, given that the company remains unlisted. The proposed margin loan would use SoftBank's OpenAI shares as collateral, a structure that depends on a defensible mark on those shares to size the borrowing and set covenants.
The loan was described as carrying a two-year term with an option to extend by one additional year. SoftBank reportedly held an approximately 11% stake in OpenAI at the end of the prior year and has committed roughly $30 billion to the company through Vision Fund 2.
A separate $40 billion bridge
The scaled-back margin loan is distinct from a $40 billion bridge loan that Bloomberg previously reported SoftBank had secured in March, backed by a syndicate of global banks, to help fund its OpenAI commitments and general corporate needs. Together, the two facilities reflect the unusual scale of leverage SoftBank is using to participate in OpenAI's funding cycle.
Why it matters
For SoftBank, a smaller margin loan means either accepting less cash up front or seeking to plug the gap from other sources. For lenders, the pushback signals that even at the top of the AI cycle, banks are no longer willing to treat private AI valuations as collateral-grade without scrutiny, particularly when the underlying asset is concentrated in a single hard-to-mark stake.
The decision lands at a moment when private AI valuations are being tested in other ways. Anthropic is reportedly pursuing a primary round at roughly a $900 billion valuation, and OpenAI itself has been the subject of conflicting signals about IPO timing, with reports that CFO Sarah Friar has privately favored a 2027 listing rather than late 2026.
Implications for the AI funding stack
If credit markets continue to balk at debt secured by private AI shares, the largest AI bets may need to lean more heavily on equity, sovereign capital, or compute-linked financing structures. That, in turn, increases the importance of OpenAI's revenue trajectory and any move toward public-market disclosure — the kinds of data points that would let lenders price OpenAI risk with something closer to confidence.
