OpenAI's $30B at $1.4T: A Bridge Round SoftBank Is Funding With High-Yield Paper

At $1.4T pre-money and at least $30B coming in, OpenAI dilutes existing holders by roughly 2.1% — small in absolute terms. What matters is the implied revenue multiple: against the most recent annual recurring revenue figure of about $70 billion, $1.43T post-money is ~20× ARR, versus ~34× at the March round's $852B valuation on the then-$25B run rate. The headline number looks larger; the multiple has actually compressed because revenue is growing faster than valuation. Investors are paying a higher absolute price but a lower multiple of sales, and that is the only way the IPO narrative survives.

Bloomberg, citing people familiar with the matter, says demand for the round is being led by investors, not by OpenAI's bankers — a meaningful inversion of normal private-market dynamics. The check-writers fall into three buckets.

On a cumulative basis, SoftBank has committed roughly $64.6 billion to OpenAI for about an 11.7–13% stake, per its Feb. 27, 2026 announcement and a reconstructed cap table that puts its paper gain at ~$50 billion even before this new round. Microsoft's $13B invested position is now valued at ~$228B (≈17.6×). That history is what makes the bridge structure possible — SoftBank alone has the unrealized gains to keep issuing paper against its position.

The $30B is small relative to OpenAI's $1.4T in cumulative spending commitments through 2030, but it is precisely what keeps the chip-and-power supply chain funded. The named downstream captures are already contracted.

Every dollar of fresh OpenAI equity converts, with high probability, into orders across this stack. NVIDIA's data-center segment reported Q1 FY27 (quarter ended Jul 31, 2026) revenue of roughly $303B run-rate and trades at ~28.6× trailing earnings and ~18× trailing sales — a multiple that already discounts aggressive AI capex. CoreWeave is the cleanest single-name exposure: TTM revenue of $7.6B against a $48.5B market cap, with OpenAI as its anchor customer. The risk in both names is not demand; it is concentration. If OpenAI's $1.4T ceiling cracks, the secondary market repricing travels down the supply chain on the same day.

CEO Sam Altman told Fortune on Sept. 12 that OpenAI will not go public in 2026, citing safety concerns. Advisors are pushing for a 2027 listing at a $1 trillion valuation — a figure that already looks modest against the $1.4T now on the tape. The 2027 IPO has to defend a price that grew ~64% in six months on private marks, against a public-market backdrop where Anthropic is reportedly targeting a mid-October 2026 listing at ~$2 trillion on $74B of ARR (vs. OpenAI's ~$70B). The competitive set is no longer tolerating multiple compression.

The headline that matters is not \"$30B at $1.4T.\" It is who pays, and the answer is concentrated: SoftBank is funding its share with high-yield paper, NVIDIA is funding its share with chip orders that recycle into OpenAI revenue, and the public-market listing that was supposed to validate the cap table has slipped a year. For every $1 of OpenAI equity, roughly $14 of compute spend must follow; the supply chain — NVIDIA, Oracle, CoreWeave, and the Microsoft-Amazon-Azure side — is the cleaner expression of the trade than the private round itself. The single biggest risk is a SoftBank-specific event (refinancing cost spike, Tokyo mark-down, or a missed $10B tranche close on Oct. 1) because SoftBank is now the marginal buyer in the AI private market.