SoftBank Group reported net profit of ¥5 trillion ($32 billion) for the fiscal year ended March, more than quadrupling from ¥1.15 trillion a year earlier, in a result that was almost entirely attributable to the rising paper value of a single investment: OpenAI.
The Vision Fund recorded a $46 billion gain for the year. Of that, $45 billion came from OpenAI. The remaining portfolio, including stakes in Coupang, DiDi Global and Klarna, collectively lost money. In the March quarter alone, the Vision Fund posted gains of approximately $20 billion, with OpenAI accounting for virtually all of the increase while other holdings dragged performance down.
Put plainly: 98% of SoftBank's Vision Fund returns for the fiscal year came from one company.
Masayoshi Son, SoftBank's founder and chief executive, has wagered his company's future on this concentration. SoftBank invested $32.4 billion into OpenAI during the fiscal year, bringing cumulative holdings to $34.6 billion by the end of March. The fair value of the stake was $79.6 billion at year-end, a gain of $45 billion on paper.
Since March, SoftBank has invested a further $10 billion and plans another $20 billion in two equal tranches in July and October, which would bring cumulative investment to $64.6 billion for approximately 13% of OpenAI.
The gains, combined with significant appreciation in SoftBank's holdings in Arm and Intel, pushed net asset value to a record ¥40.1 trillion ($250 billion) at the fiscal year end, rising above $300 billion by Tuesday. SoftBank called the results evidence of "progress we are making to become the No.1 platform provider in an Artificial Super Intelligence world."
The bull case is straightforward. OpenAI's valuation surged from $157 billion in October 2025 to $852 billion by March 2026, a 5.4-times increase in five months. Monthly revenue reached a $2 billion annualised run rate. SoftBank invested at the right time, in the right company, and the mark-to-market gains have been spectacular.
The bear case is equally straightforward, and S&P Global Ratings has already articulated it. In March, the agency revised SoftBank's outlook from stable to negative, warning that "asset liquidity and quality of its portfolio, and its financial capacity are likely to deteriorate because of its additional huge investment in OpenAI."
The ratings agency said SoftBank could mitigate the risk by selling assets, which it has been doing aggressively: $16.25 billion from T-Mobile shares, $5.83 billion from Nvidia, and $2.74 billion from Deutsche Telekom-related transactions, all used to fund the OpenAI commitment.
The financing architecture supporting the bet is complex and increasingly leveraged. SoftBank arranged a $40 billion bridge loan agreement in March. It has issued ¥1.12 trillion in domestic bonds, $4.2 billion in foreign currency notes and ¥200 billion in hybrid bonds during the fiscal year. It has taken out a $10 billion loan backed by its OpenAI stake. Every layer of this structure depends on OpenAI's valuation continuing to rise, or at least not falling significantly.
The comparison with SoftBank's previous era of concentrated bets is unavoidable. The original Vision Fund, launched in 2017 with $100 billion from Saudi Arabia's Public Investment Fund and other sovereign backers, invested across dozens of companies, including WeWork, Uber, DoorDash and Coupang.
The diversification was supposed to be the strategy. When WeWork collapsed and several other bets soured, the portfolio's winners, principally Coupang, were large enough to offset the losses.
The current strategy is the opposite of diversification. It is a single bet, at an unprecedented scale, on a company that is simultaneously the most valuable private technology firm in the world and one that has never reported a profit. OpenAI is projected to lose roughly $13 billion in 2026 on revenue of approximately $26 billion.
Its competitive position, while strong, is being challenged by Anthropic, Google and an increasingly capable DeepSeek. The cost of training and running frontier models continues to rise. And the company is currently defending itself against a $134 billion lawsuit from its co-founder that could, in an adverse outcome, result in the removal of its chief executive.
None of this means the bet will fail. OpenAI may be worth $852 billion or more. It may go public at an even higher valuation and deliver liquid returns that validate Son's conviction. But the fiscal year results make the structure of the wager impossible to ignore: SoftBank is selling proven, cash-generating assets to fund an increasingly leveraged position in a single unprofitable company, and the profit it is reporting is almost entirely unrealised.
Analysts noted that the March quarter's entire net income could be attributed entirely to booking $25 billion in valuation gains on OpenAI.
Related reading
- Anthropic finalising $1.5bn joint venture with Blackstone, Goldman Sachs and Hellman & Friedman
- Robinhood Ventures invests $75m in OpenAI
- MARA's $1.5bn Bitcoin fire sale marks the moment a crypto miner became a power company
When a ¥5 trillion annual profit can be explained by the mark-to-market movement of one holding, the profit figure tells you less about the company's financial health than it does about its exposure.
SoftBank shares closed 0.4% higher at ¥6,012 in Tokyo, a muted reaction that suggests investors have already priced in the OpenAI gains and are now waiting to see whether the next $30 billion of planned investment produces returns, or whether concentration risk eventually extracts the price it usually does.
The recap
- SoftBank reports net profit of JPY5.003trn for year
- OpenAI investment fair value now estimated at $79.6bn
- Plans to invest additional $20bn in two equal tranches
