SB Energy, the SoftBank-backed AI infrastructure venture that has yet to turn on a single data center, filed for a U.S. listing on Tuesday carrying a first-half net loss of $3.2 billion and a prospectus that reads more like a disclosure of counterparty risk than a growth story. The company is seeking to raise between $5 billion and $7 billion, according to The Wall Street Journal, while its only meaningful revenue, $139 million over six months, comes from a legacy energy business it is effectively leaving behind.
The OpenAI concentration is the business model
The S-1 mentions OpenAI 306 times, nearly matching SoftBank’s 325. The filing states plainly that SB Energy is “substantially dependent” on OpenAI as both tenant and equity investor, and that near-term revenues, project financing and development plans are “significantly linked” to OpenAI’s continued performance under lease and related agreements. Sam Altman was an early personal investor. The concentration is not a risk factor; it is the capital structure.
No operating assets, $105 billion in partner financing
None of SB Energy’s data center campuses are operational. The company relies on outside financing from partners for construction, and in August Nvidia committed $105 billion to fund an OpenAI campus in Ohio that SB Energy will build. Chief Executive Rich Hossfeld told CNBC the Nvidia commitment “helps us to unlock things like investment-grade financing.” That is a polite way of saying the equity cushion is thin and the lenders need a creditworthy anchor.
SoftBank controls the vote, the market prices the risk
SoftBank remains the controlling shareholder. The dual-class structure means public investors will buy into a vehicle where the sponsor sets strategy and the anchor tenant sets cash flow, while the company bears the construction risk and the community opposition the prospectus flags as a growing threat. The ticker will be SBE on Nasdaq and Nasdaq Texas. Pricing and timing are undisclosed, though the Journal reports a debut as early as this month.
Obsolescence risk priced at zero
The filing also warns that technological advances could render its facilities “obsolete or unmarketable,” that AI adoption may stall, and that hyperscaler capex could decelerate. Those are not boilerplate disclosures, they are the variables that determine whether a $105 billion build-out generates a return or becomes the world’s most expensive stranded asset. The market will decide whether to fund that bet at a premium to the undisturbed price, or at all.