SoftBank's Rally Has Nothing to Do With Short Sellers

There is a headline making the rounds: SoftBank's stock surge is raising "short squeeze risk" after bearish bets jumped. The premise is that short sellers are getting crowded out of a rapidly rising stock, and that the covering of those positions could fuel another leg higher.

The problem is that there are almost no short sellers to squeeze.

As of August 31, SoftBank's ADR short interest sat at 5.5 million shares - 0.05% of the public float. The days-to-cover ratio was 1.1. Even the reported 11% increase in short interest from the prior period added up to roughly half a million more shares sold short. That is not a crowded trade. It is the financial equivalent of a whisper.

SoftBank's stock price moves because of something far more consequential than a handful of contrarian bears deciding to cut losses. It moves because SoftBank is structured as an investment holding company whose net asset value is dominated by Arm Holdings, and because Masayoshi Son has leveraged the balance sheet to an extraordinary degree to bet on OpenAI. Understanding the actual mechanics of what drives the price - and what could break it - matters more than chasing a non-existent squeeze narrative.

Arm Is the Price Engine

SoftBank does not run a traditional operating business in the way a utility, chipmaker, or telecom does. It owns stakes in companies, and its share price tracks the aggregate market value of those holdings. The single most important one is Arm Holdings.

Arm accounts for more than 40% of SoftBank Group's invested assets. When Arm's stock rises, SoftBank's invested assets rise, and the market revalues the parent. When Arm falls, the reverse happens. This relationship is not theoretical - it is the reason SoftBank's ADR surged more than 25% in a single day on September 8, and the reason it had been down nearly 50% from its early-June peak just weeks earlier. Arm's own stock had moved 96% over the prior 120 days, up more than 130% year-to-date, then pulled back sharply before rebounding again.

Arm itself is a real company with real economics. Fiscal 2025 revenue reached $4.9 billion, up 22.8% year over year, with a 98.2% gross margin and 43% operating margin. Arm is pushing into selling complete chips for AI data centers - a business it projects to generate more than $15 billion annually by fiscal 2030. There is a genuine operating story here.

But SoftBank's shareholders do not own Arm directly. They own a holding company that owns a big chunk of Arm, and the value of that chunk is marked to market. The price action that triggers "short squeeze" headlines is really just Arm re-rating, transmitted through SoftBank's balance sheet.

The Debt That Matters

Where the SoftBank story gets material - and where the actual risk lives - is in how Son has financed his AI bet. The company has committed approximately $65 billion to OpenAI. A significant portion of that money was borrowed.

In March 2026, SoftBank arranged a $40 billion bridge loan - its largest-ever borrowing denominated solely in U.S. dollars - to finance the OpenAI investment. The outstanding balance was $25.9 billion. Separately, the company drew $20 billion in borrowings after fiscal year-end, also for OpenAI. In August, it secured a $10 billion two-year margin loan backed by its OpenAI stake, with lenders including Goldman Sachs, JPMorgan, Mizuho, Apollo, and Sumitomo Mitsui.

On September 9, SoftBank announced it would repay the entire $25.9 billion balance of the bridge loan on September 15 - six months ahead of its March 2027 maturity. The repayment is not free money sitting in a bank account. It is a refinancing. SoftBank is preparing to raise $10 billion to $20 billion in a junk bond sale to U.S. institutional investors, potentially with a euro tranche, and has already priced 1 trillion yen in retail bonds in Japan at interest rates between 4.31% and 4.91%. The bridge is being replaced with permanent debt that carries interest costs and maturity dates of its own.

The structure of this leverage matters. A S&P Global margin loan of $20 billion sits against SoftBank's Arm stake. The $10 billion OpenAI-backed loan carries protective clauses: if the value of OpenAI's preferred shares falls sharply, SoftBank may be required to post additional cash or repay early. S&P estimated SoftBank's loan-to-value ratio at 33% as of end-March, expecting it to drop to 20-25% by June as Arm's appreciation worked through the collateral math. But Fitch flagged that an AI market correction could threaten credit quality, citing "lofty valuations, high capital expenditure, and uncertain returns."

The point is not whether SoftBank can make the payments tomorrow. The point is that the entire construction - the holding company, the Arm mark-to-market, the leverage against private and public stakes - is designed to amplify a single thesis: AI infrastructure is going to produce enormous returns, and Son is positioned at the center of it. If that thesis holds, the leverage is a multiplier. If it does not, it is the same leverage working in reverse.

The IPO That Gets Tested Next

SB Energy - the data center and power infrastructure company backed by SoftBank, OpenAI, and Nvidia - filed for an IPO on September 1, targeting a $50 billion-plus valuation and aiming to raise $5 billion to $7 billion. Nvidia committed $3 billion in a concurrent private placement. OpenAI received warrants worth an estimated $5.5 billion.

The prospectus tells a story that investors will need to parse carefully. SB Energy has a $439 billion sales backlog based on signed contracts for data center services. But none of its data centers are currently operational. Of 8.8 gigawatts of contracted or under-construction capacity, only 0.8 gigawatt is actively being built. First revenue from data centers is not expected until the fourth quarter of this year. For the six months ending June, the company reported $139 million in revenue - most of it from changes in the value of financial instruments, not from customer sales - and a net loss of $3.2 billion.

The prospectus also flags that SB Energy is "substantially dependent" on OpenAI as both a tenant and an equity investor. The contracts with OpenAI constitute a significant share of near-term revenue. And OpenAI itself is deeply unprofitable, having recorded operating losses of $20.9 billion in 2025.

An IPO for SB Energy tests whether public-market investors are willing to value a company with no operating data center revenue, massive execution timelines, and a concentrated dependency on a privately held firm that has not yet reached profitability. It also tests whether SoftBank's controlling stake can be marked at a valuation that supports the broader leverage construction.

What Actually Moves the Stock

SoftBank's recent rallies - the 20% surge on OpenAI IPO speculation in May, the 25% single-day move in early September driven by Arm and semiconductor strength - are not the product of short covering. They are the product of mark-to-market revaluation of concentrated, AI-linked holdings, transmitted through a holding company structure, amplified by leverage.

The short interest data is straightforward: 0.05% of float is not a squeezable position. The headline about "short squeeze risk" confuses correlation with causation - the stock is rising, short interest ticked up slightly, and some analyst or terminal connected the dots without checking whether there were actually enough short positions to matter.

The real questions for anyone considering SoftBank have nothing to do with short sellers. They are:

• Can Arm sustain its growth trajectory as it transitions from a licensing model into a competitive chip business, and what happens to SoftBank's net asset value if Arm's stock re-rates downward?
• Can SoftBank service and refinance roughly $65 billion in OpenAI-linked commitments through bond markets, asset sales, and collateralized loans without triggering margin requirements or covenant breaches?
• Will SB Energy's IPO valuation provide the mark-to-market uplift that supports the broader construction, or will the gap between $439 billion in contractual backlog and zero operational data centers create a credibility test?

The leverage is the point. When a holding company borrows billions against stakes in companies whose valuations can swing 50% in a matter of weeks, the margin of safety lives in those valuations holding - not in the absence of short sellers. If Arm and OpenAI keep rising, the math works. If the AI buildout stalls, if Chinese competitors drive pricing pressure, or if OpenAI struggles to monetize at a scale that justifies its capital consumption, the same leverage that magnifies gains becomes the constraint.

That is the actual investment case. The short squeeze narrative is just noise.