NextFin News - SoftBank Group is meeting with investors as it weighs a potential jumbo bond offering, a fresh turn to the debt markets to cover its more than $60 billion bet on OpenAI. The discussions come as the Japanese conglomerate prepares to price a record ¥1 trillion retail bond this week and faces a $40 billion bridge loan that matures in March 2027.
The meetings signal that SoftBank's funding needs are outrunning the cash it can raise from asset sales. The group has pledged $30 billion to OpenAI's $110 billion February funding round alone, and it has financed that commitment with a mix of bridge debt, margin loans and bond issuance rather than realized investment gains. The question for the investors it is courting is whether SoftBank is buying time until an OpenAI listing — or building a permanent, sub-investment-grade funding model around an exit date that has not been set.
The Funding Push: What SoftBank Needs and Why
SoftBank's capital requirements have grown faster than the balance sheet can fund them internally. The group's commitment to OpenAI now totals more than $60 billion, according to company statements and filings. To cover the $30 billion follow-on it made in February, SoftBank executed a $40 billion unsecured bridge facility in March, underwritten by JPMorgan Chase, Goldman Sachs, Mizuho Bank, Sumitomo Mitsui Banking Corporation and MUFG Bank. That bridge matures in March 2027. The company has also sought a further $10 billion margin loan secured against its OpenAI shares.
The bond market is the next outlet. In late August, SoftBank outlined a record ¥1 trillion (about $6.3 billion) retail bond — the largest single issuance by any borrower in Japan — with seven-year notes expected to price on 4 September at an indicative coupon of 4.3% to 4.9%. The company said it expects an A rating from Japan Credit Rating Agency for that tranche. The new investor meetings point to a possible follow-on institutional sale, though the size, currency and timing of any such deal have not been disclosed.
The credit profile sits uneasily beside the ambition. S&P Global rates SoftBank at BB+, one notch below investment grade, having revised its outlook to stable from negative in July. That label is why banks are reluctant to warehouse the risk, according to Yuuki Fukumoto of NLI Research Institute.
"Banks are finding it difficult to take on the risk given weak deposit growth, the credit rating and the seven-year duration, leaving the deal more reliant on retail investors," Fukumoto said.
The pattern is becoming routine. The September retail deal is SoftBank's third bond aimed at Japanese households this year, after ¥418 billion in April and ¥260 billion in June. Each tranche pushes more of the group's funding onto individual savers — the lenders of last resort for an AI bet that institutional lenders are pricing cautiously.
This Is a Structural Shift, Not a Cyclical Gap
The first question to settle is whether SoftBank's funding strain is cyclical — a temporary timing mismatch that mean-reverts — or structural, a permanent change in how the group must finance itself. The evidence points to structural, and the distinction matters because it determines whether this bond cycle is a bridge to an exit or the start of a new, permanent funding model.
A cyclical shortfall would show up as working-capital timing, a temporary market dislocation, or a one-off acquisition bridge that gets repaid from operating cash flow. SoftBank's situation is different on all three counts. Its legacy cash engines — the domestic telecom carrier SoftBank Corp, the Arm stake, and the Vision Funds' realized gains — cannot generate cash fast enough to match contractual commitments to a private company that is burning capital at a frontier-AI pace. The proof is in the sequence of funding: a $40 billion bridge, a $10 billion margin-loan request, a ¥1 trillion retail bond, and now investor meetings for more. Each instrument is shorter-dated, more expensive, or more retail-dependent than the last.
History sharpens the call. SoftBank has lived through funding squeezes before. In the early 2000s, after the broadband buildout and the dot-com collapse left it overlevered, the group survived by selling assets — including its stake in Yahoo Japan and stakes in E*Trade Japan — and by leaning on its relationship banks. During the Vision Fund drawdown after 2019, it again turned to asset sales, including T-Mobile US and Nvidia shares, and to loans backed by its Arm and SoftBank Corp holdings. Those were cyclical episodes: asset sales closed, the bridges were repaid, and the leverage cycle reset.
What is different now is the scale and the counterparty. The OpenAI commitment is not a portfolio position that can be trimmed quietly; it is a contractual pledge to a single private company whose capital needs are set by the global race for compute, not by SoftBank's cash flow. The transmission mechanism is unforgiving: debt-funded stakes in a private company create a maturity wall with no internal cash flow to meet it. If OpenAI lists in 2027, the bridge repays itself from the proceeds. If the listing slips again, SoftBank must refinance into a higher-rate environment with a weaker credit, or sell assets at depressed prices. Either way, the cost of capital compounds with every delay.
This is not a company that lacks valuable assets. SoftBank overtook Toyota Motor to become Japan's most valuable listed company in June, after its shares jumped 10.3% in a single session and contributed 618 of the Nikkei 225's 709-point rise. For the fiscal year ended in March, the group reported net profit of 5.003 trillion yen ($31.74 billion), more than quadruple the ¥1.153 trillion posted a year earlier, driven by $43.9 billion of gains tied to its OpenAI stake. In the June quarter alone, it booked a 1.3 trillion yen gain on its Intel holding, and its Vision Funds rose $1.7 billion in value.
But paper gains do not repay bridges. Unrealized appreciation in OpenAI, ByteDance and Intel cannot be wired to lenders until those positions are monetized. That is the structural bind: SoftBank's balance sheet looks strong on a mark-to-market basis and tight on a cash basis at the same time. A cyclical funding gap is closed with cash flow; a structural one is closed only with a monetization event, and SoftBank's monetization event is now on someone else's timeline.
The Yen Angle: Why Domestic Savers Are the Only Willing Lenders
The currency dimension of this funding push is as important as the size. SoftBank could, in theory, issue dollar bonds to match its dollar-denominated OpenAI commitment. But a BB+ issuer pricing dollar debt into a market where investment-grade technology giants are absorbing enormous order books would pay a steep concession. Instead, the group is turning to the one pool of capital that is both deep and captive: Japanese households.
Japan's retail savings market remains one of the largest pools of patient domestic capital in the world, and it has been starved for yield after decades of near-zero rates. A seven-year note paying 4.3% to 4.9% in yen is attractive in that environment — attractive enough that banks, which are themselves grappling with weak deposit growth and reluctance to warehouse sub-investment-grade credit risk, are happy to distribute rather than hold. The structure transfers the credit risk from bank balance sheets to household portfolios, and it does so at a coupon that would be hard to achieve in an institutional dollar deal.
There is a precedent for this migration, and it is not comforting for the issuer. Oracle, facing similar skepticism from banks about the scale of its AI data-center commitments, was forced to find anchors outside the banking system for its financing. When the traditional intermediaries step back, the marginal issuer must go directly to the end investor — and the end investor prices risk differently from a relationship bank. The retail channel is more stable in quiet markets but less forgiving in stress: households can hold to maturity, but they cannot be relied upon to anchor a follow-on deal if the story turns.
The Second-Order Read: What the Bond Market Is Really Pricing
The first-order read is simple — SoftBank needs cash. The second-order question is what its pricing says about the wider AI capital race. The biggest technology issuers are flooding the bond market to fund AI infrastructure, and the order books are deep. Alphabet drew about $115 billion of orders for a $20 billion to $25 billion offering in early August. Amazon sold at least $25 billion of bonds in July. Oracle raised $18 billion in a 2025 jumbo sale and has said it may seek $45 billion to $50 billion in debt and equity.
Those issuers carry investment-grade balance sheets. SoftBank does not. That gap is the story. When investment-grade tech giants can tap order books at tight spreads, the marginal sub-investment-grade issuer pays a premium that compounds with every refinancing. The 4.3% to 4.9% indicative coupon on SoftBank's seven-year retail notes is not cheap for a BB+ borrower in yen terms, and it comes with a duration that banks declined to warehouse. The market is effectively charging SoftBank for the optionality embedded in its OpenAI stake — and that charge rises every time the IPO timeline moves.
There is also a dilution dynamic that existing bondholders should not ignore. Every new tranche adds to a stack that is increasingly dependent on a single, unlisted asset for repayment. The retail bond is structured to be attractive to households — seven years, a fixed coupon, a domestic rating agency's A — but the credit underneath is the same BB+ that S&P assigns to the group. Japanese savers are being asked to take the risk that banks would not, and the divergence between a domestic A rating and a sub-investment-grade global rating is itself a signal that the risk is being shopped to the investor least equipped to price it.
The Counter-Thesis: SoftBank Is Right About AI, and the Timing Will Work Out
The strongest case against a negative read is that SoftBank's balance sheet is far from distressed and its thesis on AI is correct. The group sits on paper gains from OpenAI, ByteDance and Intel that exceed its near-term funding needs. Masayoshi Son, the founder and chief executive, has dismissed bubble talk in the strongest terms — calling it "an insult to AI" at the June shareholders meeting and "absurd" at SoftBank World in July, where he said artificial intelligence will require $5 trillion of investment a year by 2040. If OpenAI lists in 2027 at a rich valuation, today's bridge loans and bonds will look like cheap, well-timed leverage, and the funding scramble will be remembered as a footnote rather than a warning.
That case is coherent, and it is backed by the most successful track record in technology investing. But it rests on one assumption: a timely, high-valuation listing. The falsifying signal is specific and observable. If OpenAI has not filed for an initial public offering by the end of the first quarter of 2027 — six months before the bridge matures — or if it prices below the valuation implied by SoftBank's $30 billion stake, the refinancing thesis breaks. At that point the maturity wall meets a market that has already absorbed a trillion yen of retail paper, and SoftBank would need to refinance the bridge at wider spreads or sell assets into a risk-off market.
What Comes Next: Beneficiaries, the Exposed, and Three Scenarios
The near-term path is clear. SoftBank will keep accessing debt markets until the OpenAI exit converts paper gains into cash. The retail bond pricing on 4 September is the first test; the investor meetings now underway are the groundwork for a larger institutional tranche.
Who benefits and who is exposed. Japanese retail investors gain a 4.3% to 4.9% seven-year yield in a market starved for income, but they take the credit risk that banks declined. Global bond investors get a liquid, higher-yielding yen or dollar name — at a spread that must compensate for the IPO dependency. The exposed are existing SoftBank bondholders, whose claims are being diluted into a larger, longer-dated stack, and minority shareholders, whose equity is the residual claim on an increasingly levered AI bet.
The verdict splits by time horizon. In the short term, sentiment can carry the stock if the retail deal prices smoothly and OpenAI IPO chatter returns. Over the medium term, the March 2027 bridge maturity is the hard constraint — refinancing risk is real if capital markets are risk-off or if the IPO timeline slips. Structurally, the question is whether SoftBank has become a permanent sub-investment-grade issuer funding a private AI stake, a model that only works if the exit arrives on schedule.
Three scenarios frame the path. In the base case, the retail bond prices, an institutional deal follows, and OpenAI files in 2027 — SoftBank refinances the bridge and the leverage cycle resets. In the upside case, OpenAI lists early at a rich valuation, turning the bridge into a profitable trade and the bonds into a non-event. In the downside case, the IPO slips past 2027, the bridge must be refinanced at wider spreads, and SoftBank is forced into asset sales that crystallize losses.
SoftBank is no longer funding an AI bet; it is funding the market's patience with that bet. The investors it is courting this week are being asked to underwrite the gap between a $60 billion promise and an IPO date that has not been set — and that gap, not the AI thesis, is what will determine whether this bond cycle ends in a refinancing or a reckoning.
Explore more exclusive insights at nextfin.ai.
