SoftBank earnings surge on OpenAI valuation gains

The headline numbers are real, but they collapse to a single position. SoftBank's full-year net income jumped to ¥5.002 trillion from ¥1.153 trillion, and Q4 alone tripled to ¥1.83 trillion — beating the consensus ¥295 billion estimate by more than 6x ^[1]. Inside that result, the Vision Fund posted roughly $46 billion of fiscal-year gains, with nearly $45 billion tied to OpenAI ^[3]. Bloomberg Intelligence's Kirk Boodry put it most bluntly: 'This quarter's profit basically comes down to the OpenAI stake, which SoftBank marked up by around $25 billion' ^[1]. In other words, almost every other portfolio company netted to roughly nothing this quarter.

That is not a diversified investment conglomerate any more; it is a single-name vehicle wearing a portfolio costume. The mark on OpenAI alone overwhelms the rest of the Vision Fund's book ^[10]. SoftBank's reported profit cycle is now structurally tethered to whatever OpenAI's next primary funding round prints — a level of single-asset concentration unusual even by Masayoshi Son's historical standards.

The mark is large precisely because OpenAI's primary-market valuation re-rated violently: from $157 billion in October 2025 to $852 billion by March 2026, a 5.4x move in five months ^[3]. Under fair-value accounting, SoftBank re-prices its private stake to the most recent transaction; the cumulative $34.6 billion invested was marked to $79.6 billion as of March 31 ^[2].

That accounting honesty produces accounting volatility — and, critically, no cash. SoftBank cannot sell into a public market, so the reported earnings number translates poorly into deployable liquidity. The skeptical thread in independent finance coverage and community discussion has fixed on exactly this paper-versus-pocket distinction: the $46 billion is an unrealized fair-value entry, and the same accounting that produced the upside produces a symmetrical downside if the next funding round prices flat. Nothing about this quarter changes that mechanic.

To fund a commitment now scheduled to reach $64.6 billion, SoftBank has been running a continuous asset-sale and leverage program. It exited Nvidia entirely for $5.83 billion in November 2025 ^[9], sold down T-Mobile, lifted margin loans against Arm shares to roughly $20 billion from $13.5 billion, and arranged a $40 billion bridge facility ^[3]. Reported total debt climbed to about $113 billion as of March 31, 2026, up from $77 billion a year earlier, while Q4 financing expenses rose to ¥229.4 billion from ¥148.9 billion ^[2][4].

Each financing line is individually rational; collectively they describe a conglomerate that has converted liquid public-market assets and incremental debt capacity into one illiquid private position. Jefferies' Atul Goyal labels the broader arrangement 'circular funding dynamics' — a reference to the way SoftBank, OpenAI and related compute-infrastructure deals increasingly recycle capital among themselves ^[1]. Son himself signals he intends to plow further gains back into AI and robotics rather than de-risk ^[11].

The most underrated datapoint in this earnings cycle is not the $46 billion gain — it is the $4 billion that didn't get raised. SoftBank reportedly trimmed a planned OpenAI-backed margin loan target from $10 billion to $6 billion in early May as banks pulled back on accepting OpenAI shares as collateral ^[8]. That is the first concrete instance of professional lenders, who get paid to underwrite exactly this kind of structure, declining to fully extend the leverage stack.

S&P had already moved first, revising the outlook to negative in March on the view that 'asset liquidity and quality of its portfolio, and its financial capacity are likely to deteriorate because of its additional huge investment in OpenAI' ^[5][6]. Equity markets are celebrating the mark; credit markets are pricing the funding gap. When the two disagree this sharply on the same balance sheet, the credit side has historically been the one to watch — and the social-media reception, particularly from the more skeptical finance corners, has zeroed in on this gap as the structural tell rather than the headline.

Critical commentary skews heavily toward one comparison: the same operator who took a high-profile loss on WeWork is now running a far larger, more concentrated, more leveraged bet on a single private company. The most-engaged independent-finance reviews of this quarter's print draw an explicit parallel — high-conviction concentration, debt-financed top-ups at rising marks, and a narrative-driven thesis untethered from realized cash flows.

Son himself signals he intends to plow the gains back into AI and robotics rather than de-risk ^[11], framing the strategy as ideologically aligned with OpenAI's mission of broad benefits from advanced AI. The bull case is that this time the underlying asset is genuinely transformative and the mark is conservative relative to a probable IPO. The bear case is structural: SoftBank's edge has historically been capital scale, not security selection, and capital scale becomes a liability when one position dominates the book, debt is rising, and the next funding round determines whether ¥5 trillion of reported earnings was a peak or a floor.