SoftBank's 9.75% Coupon Is the Real Test of Its AI Ambitions

SoftBank Group's shares came under pressure as investors turned a more skeptical eye on how the Japanese conglomerate intends to pay for its sprawling artificial intelligence buildout. The stock's decline unfolded against a broader pullback in AI and semiconductor names, but the funding question sits closer to home for SoftBank than any sector-wide mood swing.

Higher-than-expected US inflation data revived worries about interest rate hikes, Reuters reported, while investors simultaneously questioned the financing and sustainability of AI infrastructure spending. Concerns over the financial health of CoreWeave added to the sector's burden. None of that amounts to a proven problem at SoftBank itself — yet it sharpens where the doubt is aimed. Grand investment plans carry less conviction when their financial viability is in question.

The price of capital, not the size of the raise

The most concrete evidence of that tension arrived with SoftBank's record high-yield bond sale. The $11 billion issuance carried a coupon of as much as 9.75%. Being able to raise capital is not the same as raising it cheaply, and the higher the funding cost, the more demanding the economic case for whatever that money finances. A successful placement, on its own, does not settle the skeptics.

Nor is the bond the only thread. The Financial Times reported on October 9 that CEO Masayoshi Son is courting Gulf-region investors for as much as $100 billion to fund AI investments. That figure is explicitly not a secured inflow. SoftBank declined to comment, and Reuters could not independently confirm the report at the time. Anyone treating it as a closed financing round is skipping over the central caveat.

What matters, then, is not the maximum target size but the binding nature of any commitments. A confirmed participation from capital providers would carry a different weight than talks about an aspirational volume. Investors would need to distinguish whether a future announcement merely describes interest or actually documents pledged capital.

A Malaysian pact and a euro bond in the mix

A framework agreement closed on October 8 with Grab Holdings and Petroleum Sarawak, known as PETROS, offers another possible entry point. The partners plan to examine a platform for AI and digital infrastructure in the Malaysian state of Sarawak. What has been agreed are studies and the drafting of a roadmap — not binding implementation. A shift from review to firmer arrangements would be a positive step, and only then could the project's economic significance for SoftBank be judged. Until that happens, the deal points to a possible direction of travel without justifying any earnings taken in advance.

Separately, SoftBank announced on October 9 the full repayment of an outstanding euro-denominated bond at par. That bond carries a coupon of 2.875% and was originally due on January 6, 2027. The early redemption has been announced but not yet completed, making its actual execution a clearly verifiable next step for investors. It does not, however, confirm that fresh capital is already in place for new AI projects — different financing transactions should not be conflated.

Insurance costs and the OpenAI exposure

On the less favorable side, Reuters also reported rising costs to hedge against a SoftBank default, linking the move to the company's heavy exposure to OpenAI. That is a serious counterweight to the growth plans. If capital grows more expensive and investors grow more critical of long timelines to economic results, even ambitious projects could find less support. A potential fund sum alone would not dispel that risk.

The next hard checkpoint comes on November 10, 2026, with the scheduled earnings presentation for the second quarter of fiscal 2026, covering the six-month period through the end of September of that year. Media reports of a recovery in Asian markets today do little to change the underlying examination — a friendlier market mood alone would not justify a reassessment of SoftBank. More relevant would be progress that makes the financing of future ventures more tangible.

For now, the possible fund volume should be classified as an option rather than available financing. Should commitments become binding, the more favorable interpretation could gain ground. If funding worries instead intensify, the viability of further AI investments is likely to become an even more central yardstick for valuation. The announced obligations sit on one side of the ledger; the investment capital still awaiting confirmation sits on the other.