SoftBank Group's Record ¥5 Trillion Profit—and Why Most of It Is an Unrealized Gain

On May 13, 2026, SoftBank Group (SBG) reported results for the fiscal year ended March 31, 2026 (FY2025): net profit of about ¥5 trillion, a record for any Japanese company. The figure surpasses the previous high for a Japanese firm, and on the headline alone it reads as "a flagship Japanese company posting an unprecedented result."

But it would be premature to take that ¥5 trillion at face value as "earning power." Most of the profit comes from unrealized valuation gains—the rising paper value of its OpenAI and other holdings—which is different in character from profit that actually arrived as cash.

Behind the record number, SBG has shifted its center of gravity from a "telecom-and-investment conglomerate" to an "investment company betting on AI." Below, we break down the figures to read the true nature of this profit and the structure of an all-in bet.

A record for any Japanese company: about ¥5 trillion in net profit

First, the facts. SBG's FY2025 (year ended March 2026) net profit was ¥5,002.3 billion. News outlets reported it as "the largest ever for a Japanese company." Two engines drove it: the Vision Fund investment business and the telecom operations under the group.

In particular, the Vision Fund segment's pre-tax income reached ¥6,444.6 billion, a dramatic swing from a ¥115.0 billion loss the prior year. Net asset value (NAV) also rose to ¥40.1 trillion at the end of March 2026, up sharply from ¥25.7 trillion a year earlier. By scale alone, SBG's "comeback" is vivid (all as of the FY2025 results announced May 13, 2026).

The profit is really "OpenAI's paper gain"

So where did this enormous profit come from? The answer is clear. Much of the roughly $46 billion in annual income the Vision Fund booked stems from the surge in the value of its OpenAI investment.

The value of an OpenAI stake bought for $34.6 billion swelled to $79.6 billion, and that gap lifted the year's profit. This is not "cash from selling OpenAI high"; it is a valuation gain—"what it would fetch if sold now." If OpenAI's enterprise value falls later, the same logic can turn it into a valuation loss. The record ¥5 trillion is a number tightly linked to the valuation of a single company.

The shift to an "investment company" and an all-in bet

SBG is accelerating this trend further. In February 2026 it signed a definitive agreement to invest an additional $30 billion in OpenAI through its second Vision Fund. The pre-money valuation was set at $730 billion, with the money to be deployed in tranches of $10 billion each in April, July and October 2026. This lifts SBG's total committed investment in OpenAI to about $64.6 billion.

This is an unprecedented scale of concentration in a single company. SBG once leaned on the paper gains of its Alibaba stake, and later booked enormous losses amid the Vision Fund's slump. This time the lead role has been swapped for OpenAI. A strategy of betting on the company seizing AI dominance can generate outsized returns if it pays off, but it also carries the fragility of having results dictated by one company's fortunes. Large swings in profit are the inevitable consequence of that concentration.

How to read results where paper gains create the profit

The lesson for business readers is clear. When you see a "record profit" headline, always separate whether that profit comes from actual cash flow or from a revaluation of held assets. The two are entirely different in nature. Valuation gains lift accounting profit, but they do not increase the cash available for dividends or reinvestment by the same amount.

In SBG's case, the core telecom business generates steady cash, while most of the swing in profit derives from the valuation of its investees. That is precisely why investors look at the trajectory of NAV and at the contents and concentration of the holdings, rather than the absolute net profit. Judge results not only by "how much was earned" but by "where that profit came from and how certain it is"—the more record-breaking the result, the more this basic discipline matters.

Arm, the next move, and the visible risks

SBG's other pillar is the chip-design firm Arm. Arm has set a goal of growing revenue and EPS fivefold by 2030 and taking 15%-plus share of a market worth over $100 billion. With tailwinds from rising demand for AI chips, SBG is assembling a lineup that covers both ends of AI—OpenAI (models and services) and Arm (semiconductors).

The risk, meanwhile, is plain. As long as profit depends heavily on OpenAI's valuation, a broad cooling of AI investment could flip valuation gains into losses. The large additional investment commitments require cash outlays, adding to financing pressure. SBG's record results reflect the current position of "the company that has bet most boldly on the AI era," and at the same time show the magnitude of swing that bet contains. Both the praise and the caution arise from the same single structure.

Key takeaways

• SoftBank Group's FY2025 (year ended March 2026) net profit was about ¥5,002.3 billion—a record for any Japanese company (announced May 13, 2026). NAV expanded to ¥40.1 trillion (from ¥25.7 trillion a year earlier).
• Most of the profit is valuation gains on OpenAI and other holdings. The OpenAI stake's year-end fair value was about $79.6 billion, a roughly $45.0 billion gain against a cost of about $34.6 billion. Note this is not realized cash.
• In February 2026 SBG agreed to invest an additional $30 billion in OpenAI (pre-money $730 billion; $10 billion each in April, July and October), raising total commitment to about $64.6 billion. The single-company bet pairs large returns with large swings.
• "Record profit" is a prime case for separating cash-derived from valuation-derived gains. While SBG covers both ends of AI with Arm and OpenAI, a cooling of AI investment leaves the risk of paper gains turning into paper losses.