SoftBank’s 8.2% Yield Is the Real Verdict on Its OpenAI Bet

Masayoshi Son has placed a bet so large that the credit market has stopped treating it as a portfolio decision and started treating it as a credit event. SoftBank is seeking to issue $10 billion of dollar securities across three tenors and €1 billion of euro notes across two maturities, in what would rank among the largest junk bond deals a single issuer has ever attempted. The proceeds go almost entirely to one position.

The Bull Case: Concentrated Conviction, Not Recklessness

SoftBank’s total commitment to OpenAI is approaching $65 billion for a targeted 13% stake, a rough calculation that implies a valuation around $730 billion. That is an extraordinary number, but the business underneath it is moving. SoftBank itself has highlighted OpenAI’s rapid commercial momentum in recent investor materials, and if that continues accelerating, the upside to SoftBank’s stake can outweigh higher interest costs.

The financing structure itself has improved. The bond issue pushes repayment out across 3.5- to 7.5-year maturities, replacing short bridge funding with longer runway. Proceeds will cancel a previously secured $10 billion bridge loan facility. Term debt replacing bridge debt is not inherently alarming; it is how large commitments get refinanced once conviction hardens. The bull reads today’s deal as Son locking in his position before OpenAI approaches any kind of public liquidity event, at which point the equity value would dwarf the cost of carry.

The bond deal is part of a broader financing push that has seen the Japanese conglomerate raise tens of billions of dollars year-to-date through a combination of loans and bond sales. Scale alone does not make something irresponsible.

The Bear Case: AI Spending Has Become a Credit Problem

The bond market is not buying the conviction argument at a discount. The yield on SoftBank’s dollar bond due in 2031 climbed to about 8.2% earlier this month, up from a low near 6.7% in January, reflecting wider credit spreads and rising Treasury yields. That roughly 150-basis-point deterioration happened in under nine months, while SoftBank was actively issuing. Demand has not disappeared, but its price has risen sharply.

The cost of insuring SoftBank’s debt against default has also risen to its highest level in three years. SoftBank has sold almost $15 billion of bonds across currencies in 2026, making it one of the largest junk-rated corporate borrowers in bond markets so far this year. Add today’s deal and the total approaches $26 billion in bond issuance alone this year. S&P and Fitch both rate the group BB+, one notch below investment grade, meaning any further deterioration in the credit outlook raises the risk of forced selling by some constrained investors.

The equity collateral story has complications too. SoftBank has disclosed that it sold its entire Nvidia stake for about $5.8 billion. It has also continued to monetize other listed holdings, including T-Mobile shares, and has a large margin loan facility secured by Arm shares. The group puts loan-to-value at 13% on its own method as of June 30; S&P assessed it at 33%. A 20-percentage-point gap in LTV methodology is not a rounding error.

Where the Evidence Leads

The bullish case rests on OpenAI’s continued commercial growth and an eventual liquidity event that returns capital at a multiple of today’s cost. That outcome is possible. It is not yet provable. The bearish case rests on observable data: yields that have widened roughly 150 basis points in nine months, default-insurance costs at a three-year high, and a balance sheet whose leverage ratio depends heavily on which methodology you trust.

SoftBank’s retail bond in Japan, set on September 3 with a 4.75% coupon, sits roughly 350 basis points below what offshore institutional investors are demanding for the same credit. That gap describes exactly how wide the uncertainty range has become.

Final Verdict

The bear case is better supported today. An 8.2% yield on five-year paper from a BB+ borrower is not a market blessing of the OpenAI bet; it is a market demand for significantly more compensation to hold it. SoftBank may be right about OpenAI. But the credit market has decided that being right will take longer and cost more than the original bridge financing assumed. Watch the September 24 pricing: if the deal clears at yields materially above the existing 2031 curve, the credit problem is getting worse, not better. If books are oversubscribed and spreads tighten, the bull case gets a legitimate data point.