1981Masayoshi Son founds Nihon SoftBank in Fukuoka
1990Renamed SoftBank
1994Over-the-counter registration — SoftBank goes public
SoftBank began in September 1981, when a twenty-four-year-old Masayoshi Son set up Nihon SoftBank in the city of Fukuoka. Personal computers were just spreading in Japan, and software of every kind — from games to business applications — was pouring onto the market, yet there was no infrastructure to carry those goods from makers to retailers. To fill that gap, Son signed exclusive distribution contracts with big electronics makers such as Sharp and NEC and built, ahead of anyone, a nationwide wholesale network for PC software. In parallel he published trade magazines and ran industry trade shows, taking hold of the industry’s distribution and its information together; by his second year, monthly sales reached around $1.7M (¥400m).
What made this possible was credit. A founder in his early twenties, with neither a sales network nor a brand, won over makers and retailers by drawing on the backing of Japan’s business establishment — famously, the personal guarantee of a Sharp vice-president. From that base the company scaled quickly: it renamed itself SoftBank in 1990, and in July 1994 completed an over-the-counter registration with the Japan Securities Dealers Association, going public. The capital-raising power that listing conferred would, almost immediately, be turned not inward on the wholesale business but outward, onto investment.
1994Buys Comdex and Ziff-Davis — a window on Silicon Valley
1996Yahoo Japan founded with America’s Yahoo
1996Takes 35% of Trend Micro; a stake in TV Asahi
2000Dot-com crash — sells investee shares to cut debt
The money raised at the 1994 listing went outward. That December, SoftBank bought the American trade-show operator Comdex and the Silicon Valley IT publisher Ziff-Davis. The aim, Son himself later said, was less the earnings of the expositions and publishing than a route of access to information on the mass of venture companies spread across Silicon Valley. Through Ziff-Davis’s reporting network, SoftBank could find promising ventures and back them — the mechanism that became the starting point of its late-1990s internet-investment boom.
In January 1996 it set up Yahoo Japan as a joint venture with America’s Yahoo, taking the earliest-mover position in Japan’s search market. Investments came in quick succession — a stake in TV Asahi for a move into satellite broadcasting, 35% of Trend Micro, the memory maker Kingston. Satellite broadcasting failed when the broadcaster refused, but the initial $2.1M (¥200m) stake in America’s Yahoo turned into a cash cow that would underpin the core of the group’s consolidated asset value for years. From 1995, issuing bonds on a scale larger than its own sales became routine — a financing method the company internalized.
At the peak of the dot-com bubble, the soaring price of America’s Yahoo briefly lifted SoftBank’s market capitalization to some $185.6B (¥20tn), ranking it among the most valuable investment companies in the world. Then the spring-2000 collapse sent prices down and pushed many investees into crisis. Across 2000 and 2001 Son locked in about $1.9B (¥208bn) of gains by selling investee shares, compressing debt and pruning the portfolio. Living through peak and crash in so short a span hardened a conviction — that value is made in the timing gaps of capital flows — that fed directly into the carrier acquisitions and giant funds to come.
2001Yahoo! BB ADSL — a front-loaded broadband push
2004Acquires Nippon Telecom (fiber); buys the Hawks
2006Buys Vodafone Japan by LBO — enters mobile
2013Acquires Sprint for about $18.4B (¥1.8tn)
2016Acquires ARM for about $30.3B (¥3.3tn)
In June 2001 SoftBank launched the ADSL service Yahoo! BB and entered broadband. NTT’s ISDN was then the norm; ADSL promised high-speed access over existing phone lines. Son mounted an unprecedented nationwide campaign of handing out modems free on the street — a front-loaded-investment approach that bought a subscriber base fast. The first year brought a loss of roughly twice sales and outside calls to break the company up, but Son kept investing and turned a profit in the fifth year, proving the model himself. In July 2004 the acquisition of the fixed-line carrier Nippon Telecom, for about $3.1B (¥340bn), carried SoftBank into fiber as well; that November it also bought the Fukuoka Daiei Hawks baseball team — a franchise that made no profit on its own, taken on for the nationwide brand recognition its daily coverage would carry.
In April 2006 SoftBank bought Vodafone’s Japan business for about $15.0B (¥1.75tn) and entered mobile. The money came mainly through a leveraged buyout — a nonrecourse loan secured on the target’s own assets, which limited the impact on SoftBank’s balance sheet. Together with the bond issuance dating back to the founding years, this cemented a structure of continually investing beyond its own size; the sheer capacity to mobilize outside capital had become the core of its competitive advantage. Son led the post-merger integration personally, unified the brand under SoftBank, and from late 2011 poured heavy investment into base stations to close the quality gap with rivals.
In July 2013 SoftBank bought America’s Sprint for about $18.4B (¥1.8tn) and stepped into the North American market, but Sprint’s loss-making structure survived the deal and it ultimately passed out of SoftBank’s direct hands in the 2020 merger with T-Mobile. In September 2016 the roughly $30.3B (¥3.3tn) acquisition of the British chip-design leader ARM accelerated, across every part of the business, the long turn from telecom carrier toward investment holding company.
2017SoftBank Vision Fund formed — about $98.6 billion
2020A $42.1B (¥4.5tn) asset sale; large buyback
2022ARM–Nvidia sale collapses on regulators
2023ARM lists on Nasdaq
2025Stargate and the OpenAI bet
In May 2017 SoftBank formed the SoftBank Vision Fund (SVF1), a roughly $98.6 billion (about ¥10 trillion) vehicle backed by Saudi Arabia’s Public Investment Fund and others. It made single-company stakes of unprecedented size — hundreds of billions of yen apiece — in technology firms around the world such as Uber, WeWork and Didi. The share of outside capital rose from roughly 33% in the era when SoftBank invested mainly its own money to 96%, and the character of a fund manager moved to the front of the whole group: from a solo investor, its centre of gravity shifted to a steward of outside limited partners’ money.
The fund tied the group’s consolidated results to the share prices of its investees: valuation gains in the thousands of billions of yen when they rose, losses on the same scale when markets soured, so that quarterly net profit now swung as a matter of course. The crisis at WeWork, the shared-office operator, and its withdrawn IPO came to symbolize the model’s risk. In March 2020, as COVID-19 hit, SoftBank announced a $42.1B (¥4.5tn) asset sale and a buyback of up to $23.4B (¥2.5tn), putting the recovery of financial stability first; it staged sales of Alibaba shares and drew on ARM in various ways to steady the group’s debt ratio. A 2022 plan to sell ARM to America’s Nvidia collapsed for want of regulatory approval, but ARM itself listed on Nasdaq in September 2023, becoming a fresh source of unrealized gains.
In 2025 SoftBank Group joined with OpenAI and Oracle to launch Stargate, a $400 billion network of AI data centres in the United States, and committed as much as $40 billion to OpenAI — a stake that swelled to $64.6 billion cumulatively — naming the realization of artificial superintelligence its management theme. It also bought Ampere and Graphcore and widened ARM’s design scope from CPUs to GPUs and MPUs, gathering some 8,400 designers across the three. The substance of the company changed: from a holder of its investees’ shares into an owner of AI’s physical reality — power, computing capacity and semiconductors. Net profit for the year ended March 2026 topped ¥5 trillion, the shape of that concentration when it turns well; but the same all-in wager, by its nature, cuts both ways.
A twenty-three-year-old builds his starting point on business-world credit and a gap in distribution
The heart of this founding lies less in the insight itself — that there was a gap in software distribution — than in the fact that the business built to exploit that gap was launched by a twenty-three-year-old founder underwritten by the credit of Japan’s business establishment. During his studies in the United States, Son Masayoshi set himself the discipline of one invention a day and wrung out some 250 ideas; back in Japan he spent a year and a half surveying several industries before narrowing in on PC-software distribution. That an ordinary student pitched a patent to the head of a major corporation, and through that connection drew out the personal guarantee of a Sharp vice-president and the recruitment of seasoned managers, reads less as inspiration than as meticulous design. A new company with neither a sales network nor a brand could make makers and retailers grant it the wholesaler’s position only because it was backed by that credit.
Within a few years of its founding, sales reached the ¥10 billion range, and the wholesaler’s position standing between software houses and retailers was secured. What Son gained here was not merely the immediate margin of distribution. It was a trading network binding retailers and software houses across the country, and the financing craft needed to keep it turning. The company that renamed itself SoftBank in 1990 and, by way of its 1994 over-the-counter registration, went on to buy and invest in American ventures took as its parent body the trading network and credit built in these founding years. The distance is long from the small hypothesis of moving PC software the way bookshops move books to a company staking capital-market money on ventures around the world — yet the founding pattern, of covering youth with business-world credit and seizing a gap by the shortest path, would surface again and again.
A share plucked from an information network that set the company’s value
The heart of this decision lies less in launching a search portal in Japan than in concentrating capital on a small, unproven venture on the strength of nothing but a conviction drawn from an information network. Neither the ¥200 million of 1995 nor the ¥10 billion of 1996 could be justified by working back from the scale of the business at the time. SoftBank could commit those sums because, at the far end of the information network built through the Ziff-Davis acquisition, it could make out the shape of the technology coming next. The logic of betting not on an acquisition target’s earnings but on the information gleaned from it here bore its first fruit.
Just as weighty is that this single episode set the group’s corporate value itself. During the internet bubble, SoftBank’s market capitalization rested not on its own operations but on the unrealized gains tied to Yahoo, and that valuation swayed as the share price turned. That a stake picked up for ¥200 million turned into unrealized gains of trillions of yen shows Son Masayoshi’s nose as an investor, and at the same time carried the danger of entrusting corporate value to another company’s share price. The bet on an unknown company found at the far end of an information network, one can say, shaped what SoftBank would look like for a long time after.
The turn where the fragility of running on unrealized gains came into the open
The question running through the decisions of this period was how to treat a corporate value that a share price had created. A market capitalization of ¥20 trillion was a figure produced not by SoftBank’s own earnings but by unrealized gains on other companies’ shares — and one that would vanish just as fast once prices turned. Selling holdings amid the collapse to cut debt, a selective disposal, was the realistic response of a company acknowledging that fragility and trying to travel light. It was the period in which a balance sheet swollen by aggressive expansion was deliberately shrunk by its own hand.
The decision, in the same period, to buy a failed bank is judged more variously. Suspicions that it would be turned into a captive “institution bank,” and the turmoil of the death of the man at its top, drove home the weight of an investment company holding a business as public in character as a bank. Running a defensive shrinking of assets and a criticism-inviting bank acquisition at the same time, this period can be seen as showing, in its sharpest form, the contradiction carried by an investment company dependent on unrealized gains. The tilt toward ADSL from the following year was one answer to that contradiction.
A company that could tolerate a loss twice its sales
The heart of this decision is not simply that SoftBank entered the ADSL business, but that it tolerated, for several years, a loss exceeding twice its sales. For an ordinary company, faced with its first bottom-line loss since going public and outside voices saying there was “nothing for it but to break the company up,” it would have been natural to rein investment in. Son Masayoshi did not, because he had a calculation that once the single figure of subscriber numbers was piled up, profit and loss would surely reverse, and he concentrated resources on that one point. Whether it could keep investing in the thick of the crisis was what decided this wager.
Just as weighty is that the pattern acquired in the midst of this crisis long defined the SoftBank that followed. Take share first, accepting losses, and recover afterward — the idea established with ADSL was repeated in the fight for mobile subscribers and in PayPay’s lavish giveaways. The experience of enduring a loss twice its sales, one can say, forged the very competitive style of a SoftBank that pours capital in concentration to wrest share away.
The gamble of buying a first-rate company from an unrelated field for ¥3.3 trillion
What makes this acquisition striking is that a company whose core was telecommunications and investment poured a record ¥3.3 trillion into semiconductors, an unrelated field — and into a first-rate, profitable company rather than a loss-making one. What Son was looking at was not the results in front of him but a future thirty years out, in which IoT and AI would change the world. If an age were coming in which every device carried a chip, the company holding the design of those chips would occupy something like a tollgate. Buying ARM was a bet to get ahead and seize that tollgate.
Ironically, SoftBank once tried and failed to let ARM go, and by that very failure reaped its largest fruit. ARM, left in its hands when the sale to Nvidia fell through, roughly tripled in value amid the AI boom that followed ChatGPT and, through its listing, delivered unrealized gains. To the question of whether ¥3.3 trillion was dear or cheap, Son answered seven years after the acquisition with the phrase “a mere ¥3 trillion.” Buying ARM, alongside the forming of the Vision Fund the following year, was the starting point at which SoftBank remade itself into an investment company betting on AI.
From an operating company to an investment company
The heart of this decision lies, more than in betting on the field of AI, in remaking SoftBank into one of the world’s largest investors by levering other people’s capital. A scale of $98.6 billion was one SoftBank’s own strength could never reach. Precisely because it gathered most of that sum from outside — with Saudi oil money as the mainstay — investments of several billion dollars in a single company became possible. This design, drawing in from outside roughly three times its own funds, could, when it turned well, capture enormous returns on a small share of the burden, and could stretch its scale without limit given only the power to find targets and raise money. The investor’s face that Son Masayoshi had sharpened since the ARM acquisition here reached one finished form.
That said, the leverage of other people’s capital bares its fangs the moment outside money stops coming in. When the outside providers pulled back from SVF2, SoftBank had to shoulder over 90% itself and was placed where the price movements of its investees swayed the group’s results directly. WeWork’s valuation loss was the first proof of that fragility. As a third face, following the founding-era software distribution and the 2000s telecom business, SoftBank moved toward being a company that entrusts its performance to the results of an investment fund. From a company that earns through operations to one that rises and falls on the skill of its investing — this single move in 2017 sits at that dividing line.
The heart of this decision lies in fixing OpenAI and AI infrastructure as the true horse of the information revolution and concentrating SoftBank Group’s capital there. The company, turned into an investment company by the ARM acquisition and the Vision Fund, poured capital into a single field through two entrances: direct stakes in the model-development company, and funding for Stargate, which carries its computing base. Its stake in OpenAI alone swelled to $64.6 billion cumulatively in under a year, and consolidated net profit for the year ended March 2026 topping ¥5 trillion mirrors what this concentration looks like when it turns well.
Concentration, though, is on its reverse side a surrender of diversification. As long as the value of OpenAI and AI infrastructure keeps rising, valuation gains enrich the group’s results; but should the AI boom cool, losses rebound by the very same path. And because a considerable part of the investment is financed against collateral such as its own ARM shares, a fall in AI-related stocks could reach as far as its cash flow. Whether this is a bet aiming for winner-takes-all or an excessive risk of wagering too much on a single technology — at the time of writing the answer is not in, and the fate of SoftBank Group hangs on whether the AI investment now underway bears fruit.
Each heading links to the full Japanese analysis — background, decision and outcome, with sources.
This is a condensed English edition. The full, source-by-source history — with the detailed narrative, financial tables, shareholders and executives — is maintained in Japanese: 日本語版（詳細）— SoftBank Group full history in Japanese →
SoftBank Group Corp. — 有価証券報告書 (annual securities reports).
Nikkei Business — 日経ビジネス (Nikkei BP), 7 Oct 2019. Article.