Five trillion yen is a strange number even by the standards of corporate Japan. The record did not belong to Toyota, the industrial colossus that sells vehicles around the world, or to a megabank harvesting higher interest margins. It belonged to an investment holding company with ¥7.799 trillion in annual sales. SoftBank Group’s results, released May 13, showed net income attributable to owners of the parent at ¥5.002 trillion, up 333.7% from ¥1.153 trillion a year earlier.

The source of the surge was larger than the final profit itself. OpenAI-related investment gains totaled approximately ¥6.730 trillion, representing more than 92% of SoftBank’s ¥7.286 trillion in total investment gains. Financing costs, foreign-exchange effects, taxes, movements elsewhere in the portfolio and income belonging to minority shareholders then reduced the figure that reached the bottom line.

That distinction is the key to reading the headline honestly. SoftBank did become the most profitable Japanese company by annual parent-attributable net income. OpenAI did not, however, transfer ¥5 trillion of cash into SoftBank’s bank account. Most of the gain came from marking a privately held investment to a higher estimated fair value under international accounting rules. The value may be economically real. Unlike cash, it can also move again with funding rounds, an eventual IPO, competition, regulation and investor sentiment.

¥5.002 trillionFY2025 net income attributable to owners of the parent
¥6.730 trillionOpenAI-related investment gains during the year
¥7.799 trillionConsolidated net sales; net income equaled roughly 64% of sales
$34.6 billionCumulative OpenAI investment cost at March 31, 2026
$79.6 billionFair value of that OpenAI position at year-end
$45.0 billionCumulative investment gain on OpenAI at year-end

What the Japanese profit record really measures

SoftBank’s earnings presentation ranks the ¥5.002 trillion result as the highest full-year net profit in Japanese corporate history. The previous peaks were SoftBank’s own ¥4.988 trillion in the year ended March 2021 and Toyota Motor’s ¥4.945 trillion in the year ended March 2024. Son’s company therefore broke its own asset-driven record and again moved ahead of Japan’s manufacturing champion.

The comparison is legitimate, but the businesses are not alike. Toyota’s profit is the accumulated result of designing, manufacturing, financing, selling and servicing millions of vehicles. MUFG earns through lending, payments, securities and asset management. SoftBank Group controls operating businesses, including a telecom company and Arm, yet economically it increasingly resembles a portfolio of enormous public and private technology stakes. The quality, repeatability and cash conversion of the profits differ.

The ¥5 trillion crown is real. But it is made less from solid gold than from a mirror reflecting OpenAI’s expected future value.

The year $34.6 billion became $79.6 billion

At March 31, 2026, SoftBank Vision Fund 2’s cumulative investment cost in OpenAI was $34.6 billion. SoftBank assigned the position a fair value of $79.6 billion. The difference was a cumulative gain of $45 billion. For the fiscal year alone, OpenAI-related investment gains were $43.9 billion, or roughly ¥6.730 trillion using the company’s reporting translations.

The acceleration began with one of the largest private funding rounds ever attempted. SoftBank committed in March 2025 to invest up to $40 billion. It invested $7.5 billion at the first closing in April and another $22.5 billion in December through Vision Fund 2, while outside co-investors supplied $11 billion. After OpenAI’s recapitalization, Vision Fund 2 held preferred shares in OpenAI Group PBC.

The wager continued after the fiscal year closed. In February 2026, SoftBank agreed to invest another $30 billion at a $730 billion pre-money valuation. The first $10 billion tranche closed April 1, the second closed July 1, and a final $10 billion is scheduled for October 1. Completion would bring cumulative investment to $64.6 billion and the expected ownership interest to about 13%.

The distance between profit and cash

This is the central accounting fact. OpenAI is not publicly traded. There is no continuously quoted price on Nasdaq or the Tokyo Stock Exchange. Fair value must be estimated using recent transactions, contractual terms, comparable companies and independent valuation work. Under the fair-value-through-profit-or-loss treatment used by SoftBank, changes in that estimate enter the income statement. Rising valuations create profit; falling valuations can create losses.

It would be equally misleading to dismiss the gain as imaginary. If another investor will buy the stake at that valuation, if OpenAI lists publicly, or if it conducts a repurchase, the value can become cash. Yet SoftBank cannot assume it could immediately sell the entire position for $79.6 billion. A block that large faces contractual restrictions, taxes, market capacity and the price impact of the sale itself.

SoftBank’s own risk disclosures are unusually useful here. They warn that OpenAI must retain talent and users, reach expected earnings, secure vast amounts of computing capacity and semiconductors, and navigate litigation and regulation. SoftBank is a minority investor with limited influence over OpenAI’s decisions. OpenAI’s public-benefit mission may sometimes take priority over maximizing shareholder value, and an IPO could be delayed, repriced or never completed.

A software “bank” founded in 1981

To treat the result as a one-year AI fluke is to misunderstand SoftBank. Masayoshi Son founded Nihon SoftBank in 1981 at age 24. Its first business distributed packaged personal-computer software. The name described an ambition: a bank of software that would become infrastructure for the information society.

The company expanded into computer magazines, trade shows and American publishing assets. It jointly established Yahoo Japan in 1996. Around 2000, Son put roughly $20 million into a small Chinese e-commerce company called Alibaba. That investment defined his legend and SoftBank’s culture: identify the next information network early, then place a large bet on the company most likely to dominate it.

SoftBank acquired Vodafone Japan in 2006, giving it the network that helped bring the iPhone to the Japanese mass market. It bought Sprint in the United States in 2013 and acquired British chip designer Arm for about £24 billion in 2016. Arm’s architecture would later become one of the group’s most important assets in the AI era. Son’s bets are often early, oversized and dangerous. When the timing is right, the result can exceed the normal scale of corporate investment.

Alibaba’s success and the industrialization of venture capital

The nearly $100 billion Vision Fund, launched in 2017, attempted to industrialize the Alibaba method. It would supply promising technology companies with more capital than competitors could match, accelerate their expansion and manufacture category winners. Capital itself became a competitive strategy.

The model looked overwhelming in a rising market. Then WeWork’s failed IPO, governance failures, the pandemic, higher interest rates and collapsing technology valuations exposed its weaknesses. Vision Fund recorded an $18 billion loss in the year ended March 2020. Its loss reached roughly $26.2 billion in the year ended March 2022, followed by further valuation damage. SoftBank’s results had acquired the ability to swing from the largest profit in Japanese history to historic loss.

SoftBank’s commitments to WeWork ultimately totaled roughly $16 billion. Its collapse was not merely one bad investment. It became a judgment on founder mythology, private-market valuations and the assumption that enough capital could guarantee growth. Son shifted into “defense,” sharply limiting new investment. His return to attack would center on generative AI.

OpenAI is not simply “the next Alibaba”

The analogy is attractive but structurally wrong. SoftBank’s early Alibaba investment was small, with years to wait for growth and listing. OpenAI is already one of the world’s most valuable private companies, and SoftBank’s exposure runs into tens of billions of dollars. The upside may be vast, but so are the capital requirements for chips, electricity and data centers.

SoftBank is attempting to own or influence multiple layers of the AI stack. Arm supplies computing architecture; Ampere and Graphcore deepen chip exposure; DigitalBridge and SB Energy connect the group to data centers and power; OpenAI supplies frontier models; robotics investments offer a route into the physical world. Son’s ambition to become a leading “ASI platformer” is less a single stock pick than an attempt to assemble an integrated system.

The OpenAI partnership called Cristal intelligence is intended to deliver advanced enterprise AI in Japan, with SB OpenAI Japan marketing it to major Japanese companies. SoftBank Group has said it will spend $3 billion annually deploying OpenAI products across its own companies. In the United States, SoftBank and OpenAI are lead partners in Stargate: SoftBank carries financial responsibility, while OpenAI carries operational responsibility. The strategy is to turn investment gains into software sales, computing demand and infrastructure revenue.

The other giant number: borrowing

Building that system requires capital on a historic scale. In March 2026, SoftBank arranged a $40 billion bridge facility, primarily to fund its additional OpenAI investment and other commitments. On July 1, it drew $10 billion under the facility to finance the second OpenAI tranche. Another $10 billion payment is scheduled for October.

Management emphasizes that SoftBank’s NAV reached ¥40.1 trillion at March 31, its loan-to-value ratio was 17%, and its standalone liquidity position was ¥3.5 trillion. Its standing financial policy is to keep normal LTV below 25% and hold at least two years of bond redemptions. In management’s view, the borrowing remains manageable relative to asset value.

The risk sits in the denominator. If much of the asset value depends on technology holdings such as Arm and OpenAI, a market reversal can shrink the equity cushion quickly. Debt retains its face value while collateral is repriced. That asymmetry has appeared before in SoftBank’s history.

Four numbers investors should not confuse
  • Net income: the accounting bottom line after taxes, minority interests and other items.
  • Investment gain: includes realized gains and unrealized changes in estimated fair value.
  • Cash flow: money that actually entered or left the business; it does not equal net income.
  • NAV: the holding company’s estimated asset value after adjusted net interest-bearing debt.

Has the meaning of Japanese corporate power changed?

SoftBank’s record symbolizes a change in Japan’s corporate map. Twentieth-century Japan expanded through automobiles, steel, appliances, precision machinery, trading houses and banks. SoftBank’s 2026 profit reflects a twenty-first-century capitalism in which intellectual property, AI models, chip architecture, private shares and data-center contracts can move value faster than factories can be built.

That does not mean the Japanese economy as a whole has converted to AI profit. OpenAI is American, and much of the computing infrastructure is being built overseas. Whether SoftBank’s success leaves durable capacity in Japan depends on domestic productivity, talent, power systems, data centers, robotics and startups.

Calling ¥5 trillion a national victory requires looking beyond the valuation gain. Can SB OpenAI Japan spread useful technology beyond a small group of corporate giants? Can Arm’s value help cultivate semiconductor capability? Can massive AI electricity demand be reconciled with cost and environmental pressure? The largest corporate profit in Japanese history is not automatically an industrial-policy success.

Three conditions for keeping the crown

First, OpenAI must build earnings worthy of its valuation. User growth and technical leadership are not enough; inference costs, enterprise contracts, intellectual property, litigation, regulation and competing models will decide the economics. Second, SoftBank must convert portfolio value into cash or durable operating revenue. Third, it must preserve liquidity and collateral capacity through the next technology downturn.

An OpenAI IPO could create a major path to liquidity, but listing does not guarantee value. Public investors must support the private-market price. Lockups eventually expire. Selling reduces exposure to future appreciation; holding increases concentration. The exit itself creates a new strategic problem.

If the plan succeeds, SoftBank will become more than a fortunate shareholder. OpenAI, Arm, power, data centers, enterprise AI and robotics could create real demand for one another. Son is not merely trying to find the gold mine. He is trying to own the picks, the railway, the power plant and the market.

Forty-five years compressed into one income statement

SoftBank’s history follows a recognizable rhythm: identify a new information platform, deploy extraordinary capital—often including debt—secure a central company, and create world-scale value when the cycle rises. Then concentration and leverage reverse direction and produce a crisis. Personal computers, the internet, telecoms, smartphones, semiconductors and now generative AI: the target changes, but the method remains remarkably consistent.

The ¥5 trillion result does not erase WeWork. Vision Fund’s losses do not nullify the OpenAI gain. Both emerged from the same operating philosophy. Son’s great strength is his willingness to read the future at enormous scale. His weakness is the possibility that conviction outruns discipline on price and position size.

The decisive question is not only whether Masayoshi Son correctly identified the future of AI. It is whether SoftBank bought that future at the right price, with debt it can carry through the next storm.

In 2026, Japan’s profit champion was created not on an assembly line but inside a balance sheet reflecting the value of a private AI company. It is a genuine corporate turning point—and a story about the uncertainty of accounting and markets. The ¥5 trillion is not an ending. It is a vast advance payment on SoftBank’s ability to convert OpenAI’s valuation into cash, products, infrastructure and productivity.

Sources and further reading

This report relies primarily on SoftBank Group’s FY2025 financial statements, investment disclosures and risk factors, supplemented by official historical material and independent reporting. Fair value for a private investment is an estimate and does not guarantee a future sale price.