Part IThe Story
On the morning of June 2, 2026, the market value of SoftBank Group passed that of Toyota Motor and became the largest in Japan. A conglomerate built on borrowed money, minority stakes, and one man's private conviction about the future of intelligence had overtaken the company that, more than any other, stood for the Japanese idea of patient industrial virtue: the kanban card, the just-in-time supplier, the lifetime employee. On the Forbes real-time list Masayoshi Son was worth $100.7 billion. That passed Mukesh Ambani, and Son was the richest person in Asia for the first time in more than a decade. The Chinese financial press marked the moment. Two days later, as the enthusiasm in AI stocks cooled, roughly $13.2 billion of his paper wealth disappeared in a single session, and he slipped back behind Ambani and Gautam Adani. The headline writers in Beijing kept the count: Asia's richest man, for three days.
He had been here before, almost exactly. In the first months of 2000, at the top of the dot-com bubble, Son was briefly the richest man in the world. His biographer Lionel Barber, the former editor of the Financial Times, puts the duration at three days. Then the Nasdaq broke, SoftBank's shares collapsed, and Son lost by Barber's reckoning 97 percent of his fortune. For years afterward he held a record nobody wants: more than $59 billion gone in one crash, the most money any person had ever lost. Only Elon Musk took the title from him.
Three days, twice, twenty-six years apart. Most men are given one of these moments. Son has organized his life so that he keeps getting them back, and so that he cannot keep them. Barber, an Englishman who started his career on The Scotsman in Edinburgh and so knows something about being received as a foreigner, has said that when he was trying to get an interview with Son he settled on the one question that would open him up. Why do you hate the Japanese so much?
Hate is probably the wrong word. But the question points at the truth. Son has spent five decades trying to become number one in a country that never quite let his family in, and every time he reaches the summit the ground moves under him. Sometimes he moves it himself.
By the Numbers
The Arithmetic of Masayoshi Son
$20MSoftBank's January 2000 investment in Alibaba, later worth tens of billions
$59B+Personal losses in the dot-com crash, for years a world record
$100BApproximate size of the first Vision Fund, launched in May 2017
10xMultiple on invested capital Son claims for the ¥3.3 trillion Arm acquisition
¥5TSoftBank Group net income, fiscal year ended March 2026 (about $45.6B)
71.36%Son's voting control of SoftBank Group
$500BAnnounced scale of Stargate, the AI infrastructure venture he chairs
A House Built on the Railway's Land
The house where Son grew up did not legally exist. His father, Son Sam-heon, built it on land owned by Japanese National Railways in Tosu, a junction town in Saga Prefecture on the island of Kyushu. The family kept pigs and chickens there. The authorities objected, as they periodically did. Son later told people he still had nightmares about the smell.
The family's history was a history of being tolerated. Son's grandfather, Son Jong-kyung, came from Daegu during the colonial period, when Korea belonged to the Japanese empire, and worked as a miner. After the war, Barber reports, the family tried to go back to Korea and was turned away. They returned to Japan and to the shanty settlements where the Zainichi lived. The Zainichi are ethnic Koreans who stayed in Japan, nominally "temporary" residents even across generations. Many professions were closed to them, so many went into the trades the respectable economy preferred not to see. Son Sam-heon brewed illegal sake. Then he raised livestock. Then, by various accounts, he lent money and ran pachinko parlors, the pinball gambling halls that clatter on every Japanese high street. The bootlegging did well enough that the Sons became the first family in their neighborhood to own a car.
At home they did not use Korean names. Officially the boy was Masayoshi Yasumoto. The name did not protect him. He was bullied for being Korean, and at one point as a child he considered suicide. Decades later, speaking to a hall of Japanese teenagers about to leave for study abroad on the government's Tobitate! scholarship program, which SoftBank backs as its largest corporate supporter, he asked whether any of them had ever looked closely at their koseki, the family register in which Japanese lineage and domicile are recorded. He recalled a single character written in his own: 無. Nothing.
His father did not see nothing. He saw a prodigy.
— Masayoshi Son, recalling his father's words, in a 1987 interview obtained by BloombergYou are a genius, No. 1 in Japan. You'll be a big shot.
Sons raised on that kind of sentence tend to end up as one of two people: the one who spends his life disproving it, or the one who spends it trying to make it literally, numerically true. Masayoshi Son became the second. The family eventually moved out of the railway settlement so he could attend a better school. As an adolescent he found the first in a long line of older men he would chase, charm, and in time collect.
Den Fujita was a Japanese businessman who had studied the West closely and bet that his countrymen could be taught to eat hamburgers. In 1971 he opened the first McDonald's franchise in Japan. He later built it into a chain of some 3,800 restaurants. Son read Fujita's book and wanted to meet him. By Son's account to David Rubenstein, he called Fujita's office long-distance at least sixty times. When the calls failed, he bought a plane ticket to Tokyo and turned up at McDonald's Japan headquarters without an appointment. Fujita gave him fifteen minutes. His advice was to learn English, study computers, and go to America.
Son did all three. He also later put Fujita on the SoftBank board.
The Photograph Under the Pillow
He arrived in California at 16 and treated school as a negotiation. At Serramonte High School in Daly City, just south of San Francisco, he told administrators he already knew what was in the textbooks and should be allowed to sit the college-entrance equivalency exam at once. When they gave in, he said it would be unfair to test him in English. He got permission to use an English-Japanese dictionary, passed, and left high school within weeks. Accounts put it at two weeks or three. He studied English at Holy Names College in Oakland, then transferred to the University of California, Berkeley, to read economics.
In January 1975, at 17, he was in San Francisco paging through a magazine when he came to a photograph he took at first for the plan of a futuristic city or an intricate geometric pattern. It was a microprocessor, a computer small enough to rest on a fingertip.
"In that moment, my finger trembled," he wrote fifty years later, in SoftBank's 2025 annual report, a document otherwise devoted to net asset value and the capital structure of artificial superintelligence. "I was so overwhelmed with emotion that I could not stop crying. It felt as if I were witnessing the instant when humanity began creating something more intelligent than itself." He cut out the photograph, kept it in a folder, carried it with him every day, and slept with it under his pillow.
It is tempting to treat the story as a corporate creation myth, and maybe it partly is. But it is also the most consistent thing about him. Nearly every major SoftBank bet of the next half century can be read as another attempt to stand closer to that photograph. That includes software distribution, internet portals, broadband, mobile phones, a British chip designer, and a $34.6 billion stake in OpenAI.
At Berkeley he made himself into a machine for producing ideas. He set himself a quota of one invention a day and, by his own count, produced nearly 250 in a single year at 19. One of them, a voice-enabled electronic translator built with the help of professors including the physicist Forrest Mozer, he sold to Sharp Corporation for $1.7 million. He made another $1.5 million importing used video-game machines from Japan on credit and putting them in dormitories and restaurants. He started a game company in Oakland called Unison World and sold it to an associate for $2 million. At 19 he also wrote a fifty-year plan: establish himself in his twenties, build a war chest in his thirties, take on a major challenge in his forties, realize his vision in his fifties, hand the business to the next generation in his sixties. He has quoted it so often that it reads like scripture. He would keep to it, more or less, until the last stage.
At Berkeley he also met Masami Ohno, the daughter of a prominent Japanese doctor. They married in 1979. And in California he started using the name his family had hidden. In Daly City and Berkeley he was Son. When he returned to Japan he kept it for business, over the objections of relatives who were afraid of being exposed as Korean. For that decision he became, and remains, a role model for ethnic Korean children in Japan. He naturalized as a Japanese citizen in 1990, as Son. The order matters. He became Japanese only after he had made the Japanese call him by a Korean name.
"If I had stayed all the time in Japan," he told the New York Times in 1999, "I probably would have become much more conservative, just as other Japanese." The sentence is mild on the surface and sharp underneath. Just as other Japanese. He includes himself and excuses himself in the same breath.
Trillions, From an Orange Crate
SoftBank was incorporated on September 3, 1981. Son was 24. It started in Fukuoka with two part-time employees. One morning, as the story goes, the founder climbed onto a makeshift podium, which in the Japanese retellings is a mandarin-orange crate, and told his staff of two that the company would one day count its revenue in trillions of yen. Itchō, nichō: one trillion, two trillion. The two part-timers concluded their boss was out of his mind. Both quit.
The orange crate is the kind of detail that is either true or too good to check. The first business, at least, is documented and much less romantic. Son wanted to be a software distributor, though he had no software to distribute. He talked the Osaka electronics retailer Joshin and the leading game developer Hudson into exclusive arrangements anyway. By 1984, according to The Verge, SoftBank controlled about half of Japan's retail software market. In May 1982 it began publishing Oh! PC and Oh! MZ, magazines for owners of NEC and Sharp computers. By 1989 Oh! PC had a circulation of 140,000. SoftBank became the largest publisher of computer magazines in the country.
That was the first strategy, and Son has repeated it in different forms ever since: don't make the thing, own the roads it travels on. Distribution, publishing, trade shows. When SoftBank went public in 1994 at a $3 billion valuation, he used the currency to buy the American roads too. On April 1, 1995, he paid $800 million for COMDEX, the Las Vegas trade fair where he had first seen Steve Jobs. In November 1995 he bought Ziff-Davis, the largest publisher of computer magazines in America, for $2.1 billion. In twenty-four months he spent $4.6 billion on American technology companies. In 1996 he bought 80 percent of the memory maker Kingston Technology. Its founders, John Tu and David Sun, then announced they would give $100 million of their $1.5 billion windfall to their employees, and the gesture led network newscasts through the Christmas season. Three years later the memory market softened, and SoftBank sold Kingston back to Tu and Sun at a loss, for about a third of what it had paid. It was an early lesson in the cost of buying at the top. He would need it again.
The Americans did not quite know what to make of him. "He's known as Japan's Bill Gates," the Associated Press reported in June 1996. "Son-san—he's walking down the street, you can never see him as somebody who is a billionaire," said Jerry Yang. Yang was the Taiwanese-born Stanford graduate student who had started Yahoo as a list of websites and who would, with SoftBank's money, turn it into the front page of the early internet. Yang described a man who loved late-night pizza parties. CNET reported that Son once committed $100 million over pizza with founders in their twenties who had more passion than business experience. In Japan, meanwhile, the business press gave him a different name: ojin-kiraa, the Old-Man-Killer, for his skill at charming the elders of corporate Japan. "I make a point to respect Japanese culture," he said, "and be courteous to older people."
So he had two personas, one for each country. In 1999 a New York Times correspondent found him in Tokyo in a white shirt and an immaculately pressed black suit, standing stiffly, bowing often to show humility. By then SoftBank owned 28 percent of Yahoo and important stakes in GeoCities, E*Trade and TheStreet.com, about 100 internet companies in the United States and 20 in Japan, worth $15 to $20 billion. "If you want a global player in the Internet, they are the player," said a Lehman Brothers executive whose firm had just published a 195-page report on the company. "There's nobody else that comes close." Son was 41. "For the next 10 years," he said, "we do nothing but Internet."
That June he hosted a reception for 1,400 aspiring entrepreneurs. On the morning of July 5 he met Prime Minister Keizo Obuchi. The bootlegger's son had reached the Kantei. Within a year, he was finished. Then he wasn't.
Six Minutes in Beijing
In 1999 a former English teacher and sometime tour guide from Hangzhou named Jack Ma was running a company out of his apartment. A visitor that summer counted the founders by the toothbrushes jammed into mugs on the bathroom shelf: Ma, his wife Cathy, and sixteen others. They had put everything into it, including the apartment. Ma's ambition was a company that would last eighty years, a human lifetime. He later extended it to 102, so that Alibaba, founded in 1999, would span three centuries. He struggled with math as a student and treated his ignorance of technology as a badge of honor. People called him Crazy Jack.
Ma had a few minutes in a Beijing office to pitch a room of investors. The idea was a business directory, a Yellow Pages put on the internet. Barber says the meeting lasted six minutes. Son recognized something in the underdog. Before the presentation was really over he offered $40 million for 49 percent of the company. Ma turned him down. "Alibaba is just a baby," he said, "and a baby doesn't need this much money." SoftBank colleagues were also unimpressed and advised against it: Ma was not an engineer, not a product man, just an ordinary-looking Chinese teacher with an idea. Son overruled them. On January 18, 2000, the Wall Street Journal reported that SoftBank would invest $20 million in Alibaba.com.
That is roughly what he would have been betting on a Korean boy in a railway shack, had anyone been betting then.
Within weeks the bubble burst, and SoftBank's shares went with it. The paper fortune of the three-day richest man in the world shrank by about 97 percent. Through all of it he held Alibaba. When it listed in New York in September 2014, the $20 million had become something like $60 billion. In 2018 SoftBank's stake, enlarged by later purchases, was worth $132 billion. In September 2020 Alibaba made up 59 percent of SoftBank's net asset value. By May 2023 SoftBank had sold almost all of it, for a profit of roughly $72 billion, and by January 2024 the 23-year-old holding was gone. Many people call it the greatest venture investment in history. It has also been something more useful to Son than a trophy: a bank.
The writer Kevin Xu has pointed out that SoftBank's chart of net asset value over time is made of "home runs" that, unlike in a normal venture fund, never get distributed. Yahoo was nearly half of SoftBank's value before the crash and nearly all of it in 2002–2004. Alibaba filled the same role later. Son keeps his winners as long as he possibly can because they are the collateral that lets him survive his losers. When the Vision Fund was bleeding in 2022, SoftBank raised about $22 billion against its Alibaba shares through prepaid forward contracts. Ma's baby ended up paying for the nursery.
After the crash Son did what he would always do: he went bigger, in a new place. In September 2001, with Yahoo Japan, he launched Yahoo! BB, a broadband service, and started sending modems to Japanese households, with sales reps handing them out on street corners in Tokyo. The incumbent, NTT, owned the lines. SoftBank had to buy its way in on price. Barber says it lost about a billion dollars a year for several years. Son was convinced fast internet in every home was the future, and he was right. Yahoo! BB bought Japan Telecom, then the third-largest fixed-line operator, and became the country's leading broadband provider. The lesson he took was not that he had been reckless. It was that he had been early.
Buying the Worst Carrier in Japan
Masa first met Steve Jobs in the mid-1980s at COMDEX. Their first real conversation, Barber writes, took place in the summer of 1998 under a cherry tree at the Woodside, California, estate of Larry Ellison. Ellison, the Oracle founder and a committed Japanophile, had spent nearly a decade and around $70 million building a twenty-three-acre compound modeled on a Japanese imperial palace, with no nails, mud-plastered walls rated for a 7.3 earthquake, and a waterfall on a switch. The three men were all outsiders who had built cathedrals for themselves. Son and Jobs had a lot in common. Both were autocratic, both had extraordinary self-belief, and both cared about how a device felt in the hand. Son's Tokyo office once had fifty phone handsets mounted on a wall so he could study each one before deciding on SoftBank's own designs. He liked to quote the line from Apple's 1997 campaign: the people who are crazy enough to think they can change the world are the ones who do.
Around 2005, according to Barber's account, Son went to see Jobs in California with a sketch: an iPod with a dial pad drawn on top. A phone. He wanted it for Japan, and he wanted it before he owned a phone company. Jobs reportedly told him to come back when he had one. On March 17, 2006, Son agreed to buy Vodafone's Japanese unit for ¥1.75 trillion, about $15.1 billion, one of the largest acquisitions in Japanese history up to then.
He was buying the weakest player in the market. Vodafone K.K. lost 58,700 customers in January 2005 and 53,200 in February, while NTT Docomo and KDDI's au added hundreds of thousands. Its 3G service had 527,300 subscribers against Docomo's 10 million and KDDI's 17 million. It had imported European-style handsets that Japanese customers found foreign. On the day SoftBank confirmed the talks, March 3, 2006, the company was worth ¥3.49 trillion. Six years later it was worth ¥2.65 trillion. The market had a view.
Son had a different one. That April he told 273 new SoftBank recruits that his goal for Vodafone was "to definitely become No.1 in the industry and have people all over the world use our services and platforms." When people asked why a company with more than a trillion yen in revenue still wanted more, he said that "the Group will undermine its own raison d'être if we lose the spirit of challenge." Starting in May 2006, SoftBank's ads featured the Shirato family, whose patriarch, Otōsan (Father), is a white Hokkaido dog. His son is played by an African-American actor, Dante Carver. A surveying firm ranked the series Japan's most popular commercials every year from 2007 to 2012. It is hard not to read a joke in it about who gets to be the head of a Japanese household.
On June 4, 2008, SoftBank Mobile announced its partnership with Apple. Later that year it launched the iPhone 3G in Japan, ahead of Docomo, and it remained the country's only iPhone carrier until au got the 4S in 2011. That is the stretch in which the worst-run carrier in Japan became, in the words of the value investor Vitaliy Katsenelson, the best one, and the source of most of SoftBank's cash flow. By March 2021 SoftBank's carrier had 45.6 million subscribers. "Every time he does a big bet like this, like with Vodafone, the stock price of SoftBank plummets and everyone says he's gone crazy," said Ulrike Schaede, a scholar of Japanese business at UC San Diego. "And six months later people say, 'Wow, he's got a plan.'"
He also bought a baseball team. On January 28, 2005, SoftBank took over the Fukuoka Hawks, in the Kyushu city where the company had started above a fruit-crate podium. Under SoftBank the Hawks won seven Japan Series between 2011 and 2020, and another in 2025.
In 2012 he made the same kind of bet in America. On October 15, SoftBank announced it would buy 70 percent of Sprint Nextel, the struggling third-place U.S. carrier, for $20 billion. SoftBank's stock fell 17 percent in one day, its biggest drop since listing in 1994. The FCC approved the deal on July 6, 2013, at $22.2 billion for 78 percent, and SoftBank kept buying until it owned about 84 percent. This time he could not fix it. Sprint mostly lost money for years. When it finally merged with T-Mobile US on April 1, 2020, SoftBank came away with 24 percent of the combined company. In 2024, with T-Mobile the world's largest telecom company by market capitalization, Son described the episode as the third stage of SoftBank's evolution. He had got around to calling it a success.
In between came the earthquake. After the Fukushima Daiichi meltdown in March 2011, Son broke with the corporate establishment and attacked the nuclear industry for creating "the problem that worries Japanese the most today." He pledged to build solar farms across the country. The next year SoftBank released a handset with a built-in Geiger counter. His youngest brother, Taizo, filled three shipping containers with emergency supplies for the tsunami zone. It was an unusual stretch in which the outsider spoke for the nation, and for a while the nation listened.
The Retirement That Wasn't
At 57 he picked a successor. Nikesh Arora was an Indian-born former Google executive who had run SoftBank's investment arm, and Son had recruited him, Barber reports, with a $320 million package written on a napkin. In May 2015 Son named him president and representative director. It was the sixties stage of the fifty-year plan: hand over the business to the next generation.
It lasted about a year. In June 2016, under pressure from investors, Arora stepped down. Son, by his own account, had briefly flirted with retirement and found he didn't want to retire. "Son acts on instinct," Satoshi Shima, a former senior lieutenant who later became a professor, told the New York Times. "It's a genius instinct, but it's not logic."
The next month showed the instinct at full strength. On June 23, 2016, Britain voted to leave the European Union and the pound fell. Within weeks Son announced that SoftBank would buy Arm Holdings, the Cambridge chip designer whose architecture runs nearly every smartphone, for £24.3 billion, about $32 billion. It was the largest purchase of a European tech company ever. The price was a 43 percent premium to an already expensive stock, about 48 times earnings, with no obvious synergies. "A bored rich man who needs to spend because he has too much," was how one investor summarized the market reaction to Katsenelson. The deal closed on September 5, 2016. "SoftBank has had a lot of businesses, but they've always ranked, say, third in Japan, or fourth in the U.S.," Son said. "This is the first time we're directly controlling a business that is No. 1 in the world."
The first time. He was 58. All the businesses he had built since the orange crate, and in his own accounting none of them had yet met the standard his father set.
On February 27, 2017, at the Mobile World Congress in Barcelona, Son came on stage in a black turtleneck that everyone in the hall recognized. He told the audience that within thirty years a single chip would have the equivalent of an IQ of 10,000, and that computers would surpass human intelligence. "I really believe this," he said. Ordinary objects would carry more computing power than the people who used them. "We will be less than our shoes." The audience laughed. The growth of machine intelligence, he said, "is why I acquired ARM." Then he promised to double Arm's engineering staff, which cut its profits sharply and worried nearly everyone with a stake in the company.
Arm turned out to be the best thing he ever bought, but not on the schedule he expected. In September 2020 he agreed to sell it to Nvidia for $40 billion, a deal worth about $66 billion by 2022 on Nvidia's rising stock, and announced that the combination would "lead the era" of artificial intelligence. American and European regulators blocked it, and it collapsed in February 2022. What looked like a defeat saved him. Arm listed on Nasdaq on September 14, 2023, raising $4.87 billion at a $54.5 billion valuation, with SoftBank keeping 90.6 percent. As AI demand pushed the shares up, SoftBank's net asset value doubled in a year, from ¥14 trillion to ¥28 trillion. Son now says the ¥3.3 trillion acquisition has returned ¥24.6 trillion, ten times the money.
Project Crystal Ball
Inside SoftBank the plan had a code name. Alok Sama, a banker who took over part of Arora's overseas duties, recalls that it was called Project Crystal Ball. To the world it was the SoftBank Vision Fund. Announced in 2016 and with its first major close in May 2017, it gathered close to $100 billion: $45 billion from Saudi Arabia's Public Investment Fund, $28 billion from SoftBank (of which $8.2 billion came from selling about a quarter of Arm into the fund), $15 billion from Abu Dhabi's Mubadala, plus money from Apple, Qualcomm, Foxconn, Sharp, and Larry Ellison. Son told Rubenstein that the Saudi commitment came out of a forty-five-minute conversation: $45 billion, a billion a minute. He flew to Omaha to pitch Warren Buffett, the investor least like him on earth, and left without Buffett's money.
The fund's structure made it fragile. Just over half its capital was preferred equity, a debt-like instrument paying the Gulf investors a fixed 7 percent whatever the fund's performance. So a fund meant for long, illiquid, uncertain bets had to make regular coupon payments. It had to buy into late-stage companies at peak valuations and generate cash to pay its own investors. Son planned to raise $100 billion every few years and deploy about $50 billion a year.
The money hit Silicon Valley hard. On Sunday, September 17, 2017, Michael Moritz woke in his Pacific Heights mansion. He was the lean Welsh émigré, a former Time journalist, who had become one of Sequoia Capital's billionaire partners, and who spent ordinary weekends on his racing bike or among his German Expressionists. He sat down to write to his partners.
— Michael Moritz, Sequoia Capital, email to colleagues, September 17, 2017There is at least one difference between Kim Jong Un and Masayoshi Son. The former has ICBMs that he lobs in the air while the latter doesn't hesitate to use his new arsenal to obliterate the hard-earned returns of venture and growth equity firms.
He called the Vision Fund "the greatest long-term threat to the health of our business since our formation in 1974," quoted Mike Tyson, and at one minute past noon sent a call to raise $20 billion in response. That started an arms race. Over the next four years Sequoia, Andreessen Horowitz, Benchmark, Silver Lake, and Tiger Global all raised and deployed larger funds to compete with SoftBank. The Korean outsider who had spent his youth trying to get into Japanese rooms had reshaped the economics of the most exclusive club in American capitalism, simply by showing up with more money than anyone thought he should have.
He spent it quickly and on his own instincts. Paytm got $1.4 billion in May 2017, Flipkart $2.5 billion that August, WeWork $4.4 billion that same month, Uber $9 billion in a consortium deal that December that made SoftBank its largest shareholder, DoorDash $535 million in March 2018, Opendoor $400 million that September, Grab, Didi, Oyo, Coupang. The average check in the first fund was $943 million. Son explained his choices with references to the sparkle in a founder's eye and to the Force from Star Wars. He told the founders he would personally connect them to one another so the portfolio would form a family of mutually reinforcing companies. He had a name for this, the "Cluster of No.1 Strategy": stakes of 20 to 30 percent in the leader of every important field, linked together. In the 2018 annual report he called it his real invention, more important than the translator. He said he had drawn on how bacteria evolved over four billion years, through self-propagation and then self-evolution.
The fund also owned a 4.9 percent stake in Nvidia by early 2019, about as good a call on the coming AI boom as anyone made. It had wrapped the position in a collar, a hedge that limits losses and gains alike. When Nvidia's shares fell in late 2018, the fund sold the whole stake in February 2019 for a $3.3 billion return. By June 2024 those shares would have been worth more than $150 billion. "The fish that got away was big," Son said.
The Sparkle in a Founder's Eyes
The founder he loved most was Adam Neumann, the tall, long-haired Israeli who had turned office subleasing into a quasi-religious lifestyle brand and called it WeWork. SoftBank's first investment, $300 million, came in March 2017. More followed: billions from SoftBank and the Vision Fund together, over the objections, according to former officials of both companies, of Son's own lieutenants. By early 2019 WeWork was valued at $47 billion. Son told Neumann to think bigger.
In August 2019 WeWork filed to go public, and the filing exposed the losses and the governance. Investors balked. The IPO was withdrawn on September 30. SoftBank put together a $9.5 billion rescue that included a $1.7 billion exit package for the founder it was removing. On November 6, 2019, SoftBank reported its first quarterly loss in fourteen years, ¥704 billion. Son stood in front of investors in Tokyo and apologized, which a Japanese chief executive does in public only when things are serious.
— Masayoshi Son, SoftBank earnings presentation, November 6, 2019I overestimated Neumann's good side. I turned a blind eye to [his] bad side on things like corporate governance. I have learned a harsh lesson.
"My investment judgment was poor in many ways," he said. "I am reflecting deeply on that." He also said he still expected a "hockey stick" recovery.
There was more bad news. In December 2019 Tadashi Yanai, the founder of the Uniqlo parent Fast Retailing and then Japan's richest man, left SoftBank's board after eighteen years. He was one of the few directors who could challenge Son. In February 2020 the activist hedge fund Elliott Management disclosed a $2.5 billion stake and demanded restructuring and more transparency about the Vision Fund. Then came the pandemic. In March 2020, as SoftBank's shares fell, Son announced a ¥4.5 trillion asset sale, about $41 billion, to fund buybacks of up to 45 percent of the company's stock and to pay down debt. The shares rose almost 19 percent that day. In May the Vision Fund reported an $18 billion loss, and 15 percent of its roughly 500 staff were let go. Jack Ma left SoftBank's board that same month. In June, Son left Alibaba's. The second Vision Fund, announced in July 2019 with a $108 billion target, failed to attract outside investors and ended up funded by SoftBank alone.
Then the liquidity of 2020 and 2021 briefly made him look right again. Coupang, the Korean e-commerce company SoftBank had backed with $1 billion in 2015, went public in March 2021. For the fiscal year ending that month SoftBank reported net profit of ¥4.99 trillion, $45.88 billion, the largest annual profit any Japanese company had ever recorded. Son took it as permission. Within nine months of that summer he quintupled the number of companies in the second fund, and the average check fell from $943 million to $192 million. In the 2021 annual report he redefined SoftBank: not an investment company but a "capital provider driving the Information Revolution," the Rothschilds to the James Watts of artificial intelligence. He admitted to a criticism that clearly bothered him: I like Masayoshi Son as an entrepreneur, but not so much as an investor.
The reversal was severe. For the year ending March 2022 the Vision Fund lost a record ¥3.5 trillion, $27.4 billion. Klarna's valuation fell 85 percent to $6.7 billion. Wirecard was found to have forged client details to win a €900 million investment. Katerra went bankrupt. Zymergen collapsed four months after its IPO. In the April–June 2022 quarter alone the fund lost $23.1 billion. In August 2022, asked about the way he had run it, Son said he was "embarrassed" and "ashamed" of the pride he had once taken in its profits. Barron's called the fund a failed experiment. Bloomberg wrote about his "broken business model." Rajeev Misra, the former Deutsche Bank derivatives trader who had built the fund for him, stepped back that July. CFO Yoshimitsu Goto said, "We need to go full-on defense. SoftBank is pessimistic on the outlook. We do not yet see the light."
It reached Son personally. To raise his own compensation he had structured side deals: 17.25 percent of a Vision Fund 2 vehicle, a similar stake in the Latin America fund, 33 percent of SB Northstar, the trading unit that in September 2020 had been identified as the "Nasdaq whale" buying billions in options on American tech stocks. He paid no cash up front and owed 3 percent interest on the unpaid amount. By November 2022 the losses meant he owed SoftBank $4.7 billion. By February 2023 the figure was $5.1 billion. In March, as Silicon Valley Bank failed, more than a third of his SoftBank shares were reported to be pledged as collateral for margin loans. He insisted there was no conflict of interest. In April 2023 SoftBank sold its Asian early-stage venture arm to a firm run by his youngest brother, and the Financial Times raised governance concerns.
That brother, Taizo Son, was born in Tosu in 1972, fifteen years after Masa. Unlike him, Taizo went to the University of Tokyo, the gateway to the Japanese establishment. While still a student he was recruited into his older brother's Yahoo Japan project, pitched Jerry Yang on how to launch it, managed a hundred staff, and got the site live in three months. He later founded GungHo, whose Puzzle & Dragons was the top-grossing mobile app in the world in 2013, and he became a billionaire himself. In 2018 he moved to Singapore, where, he said, "there's no legacy to break." The two brothers took the same inheritance in different directions. One went into the center of Japanese institutions in order to leave. The other has spent his life trying to buy his way in from outside.
WeWork filed for bankruptcy on November 7, 2023. SoftBank's equity losses were estimated at $11.5 billion, with $2.2 billion of debt still at risk. TIME put the total he had put in at $18.5 billion. "You can recover from mistakes," said NYU's Aswath Damodaran, "but how do you recover from the perception that you don't know what you're doing?" By then Son had said what he had to say. In June 2023, to shareholders, he said: "I fell in love with WeWork." And: "I may be more at fault than Adam, for telling him to be more aggressive."
Six Hundred and Thirty Inventions
For about eighteen months he largely disappeared from public view. In February 2023, for the first time, he skipped SoftBank's earnings call and handed quarterly reporting to Goto. Later he described what had happened that autumn in unusually personal language for a document addressed to bondholders.
"In autumn 2022, there was a moment when I started to wonder how many years I had left in my entrepreneurial journey," he wrote in the 2023 annual report. "At times, these thoughts left me feeling somewhat empty." He said he had spent the forty-two years since founding SoftBank using the left side of his brain, the logical, managerial side, and that the emptiness reawakened the right side, the one that produced 250 inventions in a year at Berkeley. In the eight months after that autumn, he wrote, he came up with about 630 inventions. Ideas came to him in his sleep. He would get out of bed, write them in an inventions notebook, photograph the page, and send it to his team. "Sure, most of them will turn out to be duds," he wrote. "However, I firmly believe some of them will be real gems." Around the same time Business Insider noted that he had compared himself, at various points, to Jesus and to Yoda.
The man who came out of that period had narrowed his focus to one idea, artificial superintelligence, which he defined as intelligence ten thousand times greater than the human brain, arriving within about a decade. That is the same number he had attached to a single chip in Barcelona in 2017. He told shareholders he was "born to realize" it. He started selling assets to pay for it. The remaining Alibaba shares went. T-Mobile was largely sold. In 2025 SoftBank sold its entire stake in Nvidia, about $5.8 billion, its second exit from the company that most clearly fit his long-held thesis, to put the money into OpenAI. By the end of 2025 SoftBank had invested about $34.6 billion in OpenAI for roughly 11 percent, making it the third-largest shareholder after the nonprofit foundation and Microsoft. It bought Ampere Computing for $6.5 billion, agreed to buy ABB's robotics division for $5.375 billion as part of what Son called "Physical AI," and agreed to buy the data-center investor DigitalBridge for about $4 billion. In February 2025 it formed a joint venture with OpenAI to sell enterprise AI to Japanese companies under the name "Crystal Intelligence." TIME reported that he was proposing a $1 trillion AI and robotics complex in Arizona, alongside TSMC, called Project Crystal Land. From Crystal Ball to Crystal Intelligence to Crystal Land: the same image across nine years, with the money behind it rising from billions to trillions.
He also returned to the Americans. On December 6, 2016, after meeting President-elect Donald Trump at Trump Tower, he had promised $50 billion and 50,000 American jobs, and CNN later could not establish how many of those jobs actually appeared. On December 16, 2024, at Mar-a-Lago, he promised $100 billion and 100,000 jobs. On January 21, 2025, the second day of Trump's second term, Son stood at a White House lectern next to Sam Altman and Larry Ellison and announced Stargate, a $500 billion AI infrastructure venture with OpenAI, Oracle, and Abu Dhabi's MGX, with Son as chairman. A staffer set out a step stool so the 5-foot-5 Japanese billionaire could reach the microphone. He looked at it and joked about it, and the room laughed. He had been small in big rooms all his life and had learned to make the smallness work for him.
In May 2026 SoftBank committed up to €75 billion to five gigawatts of AI data centers in France, its largest such commitment in Europe. For the fiscal year ending March 2026 the group reported net income of ¥5 trillion, about $45.58 billion, matching the 2021 record. Its total assets were ¥60.75 trillion. The orange-crate promise of itchō, nichō had turned out to be modest.
A Village on the Thirty-Ninth Floor
In 2026, at 68, Son revised the fifty-year plan. He had been at the sixties stage, handing over the business, and he had attempted it with Arora and abandoned it in a year. Now he wrote, in the annual report: "I find myself wanting to do more." He would keep going into his seventies. The goal would be to make SoftBank "the undisputed No. 1 ASI Platform Provider in Japan, and in the world." In Japan came first. It still matters to him which country gets mentioned first.
As of June 1, 2026, he held 1,865,511,164 SoftBank shares, about 29 percent of the economic interest and, through the structure of those holdings, 71 percent of the vote. He lives in Tokyo in a three-story house valued at $85 million, with a golf range that recreates the weather of the world's great courses. He bought a house in Woodside, near Ellison's palace, for $117 million. He drinks La Tâche from Domaine de la Romanée-Conti. SoftBank's logo, two horizontal bars, comes from the flag of the Kaientai, the shipping company Sakamoto Ryōma founded in 1865. Sakamoto was a low-ranking samurai who broke with the feudal order of the Tokugawa shogunate and helped push Japan into the modern world before he was assassinated at 31. The company still describes itself as "a fleet of similar visionaries" sailing under a 21st-century version of that banner.
Every pivot has been followed by the same criticism, that SoftBank has no idea what it is. "Some people say that SBG changes its core business too often," Son wrote, "but in reality, our core business has remained unchanged all these years. All that has changed is the means." The business is the Information Revolution, and the slogan underneath it, Happiness for everyone, is printed on the company's documents.
A TIME correspondent visited him that year at SoftBank's headquarters in Tokyo's Minato ward, overlooking the water. The top floor is not the usual glass-walled executive suite. It is paneled floor to ceiling in pale pine and built to suggest a traditional Japanese village, with exposed timber joinery and the sound of running water. Below the windows lie gardens laid out for the Edo-period shoguns. "I designed it personally," Son said, looking down on them.
Part IIThe Playbook
Son is easy to admire and dangerous to copy. What follows is an attempt to separate the method from the man. Each principle is drawn from the record above, which includes both the right calls (Alibaba, Vodafone, Arm) and the ruinous ones (Kingston, Sprint, WeWork, the second Nvidia exit). It is not a list of virtues. Several of these principles are also descriptions of how he failed. The point is to understand which parts of the machine produce which results.
The principles
- 1.Bet on the inflection, not the trend.
- 2.Own the roads before you build the cars.
- 3.Buy the broken incumbent your rivals have written off.
- 4.Keep a home run on the balance sheet as a lifeboat.
- 5.Underwrite the founder, then underwrite your own blind spot.
- 6.Manufacture the exclusive before the product exists.
- 7.Use size as a weapon, and price in the recoil.
- 8.Change the means, never the mission.
- 9.Write the fifty-year plan, then revise it in public.
- 10.In a drawdown, survival outranks pride.
- 11.Keep someone in the room who can say no.
- 12.Turn outsiderhood into access.
Principle 1
Bet on the inflection, not the trend
Katsenelson's defense of the Arm deal in 2016 is still the best account of Son's thinking: he looks for industries moving from high linear growth into accelerating, exponential growth, and he pays whatever it takes to be positioned before the curve bends. PCs in the early 1980s, the internet in the mid-1990s, mobile data in the mid-2000s, mobile games in 2013 (Supercell, bought at a $3 billion valuation and sold three years later for more than three times that), chips for AI in 2016. On a trailing basis all of these looked overpriced. At 48 times earnings, Arm was mocked as the impulse of a bored rich man. It returned ten times the capital.
The danger is that this kind of thinking works on industries, not individual companies. Son has been right about the curve far more often than about which company would ride it. He saw AI clearly enough to own 4.9 percent of Nvidia in 2019 and still sold it, because the vehicle holding it could not tolerate the volatility along the way.
Tactic: Before any large bet, write down what has to be true about the rate of change, not the level, and then make sure the structure holding the position can survive the period before the curve bends.
Principle 2
Own the roads before you build the cars
SoftBank started with no product. Son locked up distribution (Joshin, Hudson), then the information channels (Oh! PC, Ziff-Davis), then the meeting place (COMDEX), and by 1984 had half of Japanese software retail. Every software maker in Japan needed him, and every American computer company that wanted Japanese customers had to deal with him. That is how a 24-year-old from a railway settlement came to be Microsoft's main distributor in Japan and a major figure in Cisco's entry there.
The same logic explains his later moves. Broadband and mobile were roads. Arm, which licenses the architecture other companies build chips on, is a toll road. Stargate and the French data centers are roads for AI. When Son loses money, it is usually when he has bet on one vehicle, a WeWork, instead of on the infrastructure every vehicle needs.
Tactic: Identify the chokepoint every player in your target industry must pass through (distribution, standards, physical capacity) and try to own a piece of that before you try to win at the application layer.
Principle 3
Buy the broken incumbent your rivals have written off
Vodafone Japan in 2006 was losing tens of thousands of subscribers a month and held 4.8 percent of the country's 3G market. That made it cheap, and it came with spectrum, towers, licenses, and fifteen million customers. Those are things no startup could build in a decade at any price. Son paid ¥1.75 trillion, swapped European handsets for Japanese ones, put a dog in the role of the father, and got the iPhone. Within a few years the asset everyone had written off was paying for the rest of the group.
Sprint is the warning. It was the same playbook applied in a market where Son had no exclusive product, no regulatory goodwill, and less operating knowledge. The turnaround thesis only works when you hold an advantage the previous owner lacked.
Tactic: When you evaluate a distressed asset, name the single advantage you bring that the seller did not have. If you cannot name it, you are buying a problem, not a turnaround.
⇄
Two Turnarounds, One Playbook
Why the same move produced opposite results.
| Factor | Vodafone Japan (2006) | Sprint (2013) |
|---|---|---|
| Price | ¥1.75T (~$15.1B) | $22.2B for 78% |
| Home-market knowledge | Deep | Limited |
| Exclusive edge | iPhone, exclusive until 2011 | None comparable |
| Market reaction on announcement | SoftBank shares fell | SoftBank shares fell 17% in a day |
| Outcome | Group cash engine | Rescued by T-Mobile merger |
Principle 4
Keep a home run on the balance sheet as a lifeboat
A conventional venture fund returns its big winners to its investors. Son holds his. Yahoo was close to all of SoftBank's net asset value from 2002 to 2004, and Alibaba was 59 percent of it in September 2020. The shares were not just trophies. They were collateral. When the Vision Fund was losing money in 2022, Alibaba stock raised about $22 billion through prepaid forwards, and the gradual sale of the stake produced $34 billion in gains in 2022 and $8.5 billion more by January 2024. Arm now plays the same role.
This is the least discussed part of his success and the most transferable. His appetite for risk depends on one enormous, liquid, appreciating asset behind it. Without that asset, it would be a gambler's habit.
Tactic: Before you increase risk, identify the one asset that will fund your survival if everything else fails, and do not sell it to finance the next bet.
Principle 5
Underwrite the founder, then underwrite your own blind spot
Six minutes with Jack Ma produced the best investment of his life. The same method, deciding quickly on the founder, overruling cautious colleagues, offering more money than was asked for, produced WeWork. Son said it himself in 2019: "I overestimated Neumann's good side. I turned a blind eye to [his] bad side." The quality that let him see Ma, a sympathy for the underdog who could not explain himself on paper, is the same one that kept him from seeing Neumann.
Instinct is not the problem. The problem is when nothing in the process tests it. Ma, notably, turned down half of Son's first offer, saying "a baby doesn't need this much money." That should be a signal: a founder who takes less than you offer has a better sense of his own needs than one who takes everything.
Tactic: When conviction about a founder drives a deal, write down the founder's worst trait before you sign, and give one colleague the explicit job of watching for it.
Principle 6
Manufacture the exclusive before the product exists
Son's sketch of an iPod with a dial pad predated both the iPhone and his ownership of a phone company. The relationship with Jobs went back twenty years, to COMDEX and Ellison's cherry tree, before it turned into exclusive Japanese iPhone rights from 2008 to 2011. It was not luck. He had built the relationship and imagined the product before either could pay off. The pattern repeats in his joint venture with OpenAI for Japanese enterprise customers, set up before most Japanese companies had decided what they wanted from the technology.
Tactic: Pitch the product you need to the person who can build it before you have the platform to sell it, then use the commitment to justify acquiring the platform.
Principle 7
Use size as a weapon, and price in the recoil
The Vision Fund's $100 billion changed venture capital. Moritz's email of September 17, 2017 shows how threatened the incumbents felt. Large checks let Son pick winners in competitive markets by funding one side of a price war (Didi, Uber, Grab, Ola) and set valuations other investors had to follow. Size, however, has costs. The 7 percent preferred coupon forced a long-horizon fund to produce near-term cash. Big checks inflated valuations, which inflated the losses when markets turned. The arms race also brought rivals with more capital to compete for the same deals.
◎
Conventional Venture Capital vs. the Vision Fund Model
| Dimension | Conventional VC | Vision Fund 1 |
|---|---|---|
| Stage | Early, small checks | Late-stage, average ~$943M |
| LP economics | Paid on realized returns | ~Half preferred equity at 7% fixed |
| Role of capital | Fuel | Weapon to pick market winners |
| Peer response | Co-invest | Arms race (Sequoia, a16z, Tiger) |
| Result | Power-law returns | Record losses, FY2022–23 |
Tactic: If you plan to win by outspending everyone, first model what happens to your financing structure if valuations fall by half, because overwhelming capital raises the stakes on mistakes as much as it improves the odds.
Principle 8
Change the means, never the mission
SoftBank has been a software distributor, publisher, trade-show owner, internet holding company, broadband carrier, mobile operator, U.S. telecom owner, venture fund, chip company, and AI infrastructure builder. Each change produced the same criticism: no strategy. Son's answer is that the mission ("Information Revolution—Happiness for everyone") has stayed the same and only the means have changed. In the 2024 report he divides the company's history into six stages, each a new way to stay near the microprocessor he first saw in a magazine in 1975.
This is useful in practice. A fixed mission lets a company drop businesses without an identity crisis. Selling Supercell, SB Energy, Boston Dynamics, Fortress, and finally Alibaba never meant giving up SoftBank's purpose. It meant moving money to the next means.
▲
Six Stages, One Mission
Son's own periodization of SoftBank, from the 2024 Group Report.
1981
Packaged software distribution; Yahoo Japan founded 1996.
2001
Yahoo! BB broadband; Vodafone K.K. acquired 2006; iPhone 2008.
2013
Sprint acquired; later merged into T-Mobile US.
2000–2014
Alibaba partnership carries internet services into China.
2017
SoftBank Vision Funds anticipate the rise of AI.
2023
Arm IPO; NAV doubles; pivot to ASI, OpenAI, and Stargate.
Tactic: Write a mission specific enough to exclude things and general enough to outlast any one product, then judge every divestiture by whether it serves the mission, not whether it preserves the past.
Principle 9
Write the fifty-year plan, then revise it in public
At 19 Son set out a plan by decade: establish, accumulate, challenge, realize, hand over. It gave him a time scale most competitors don't use. The 300-year vision and the strategic framework he drew up in his mid-twenties, twenty-five characters arranged in five tiers and partly based on Sun Tzu, are the same habit applied to the company. These plans let him stay calm through losses that would have made other executives quit, because each loss was a stage in a longer story.
Note, though, where the plan broke. The handover stage failed with Arora. At 68 Son rewrote it openly and pushed the final stage into his seventies. A long plan is worth having for its direction, not its detail, and being willing to revise it publicly is part of what makes it credible.
Tactic: Write a plan with decade-long stages, review it each year, and when reality forces a change, announce the revision instead of quietly dropping the plan.
Principle 10
In a drawdown, survival outranks pride
Son's reputation is for boldness, but his best moves under pressure have been defensive and quick. In March 2020 he announced a ¥4.5 trillion asset sale and buybacks of up to 45 percent of the stock, and the shares rose almost 19 percent. In 2022 he let his CFO say "full-on defense," cut new startup investment by more than 90 percent within a quarter, sold down Alibaba, and stopped presenting earnings himself. The public apologies ("embarrassed," "ashamed," "I may be more at fault than Adam") were part of the same strategy. A founder who admits error early keeps the credibility he needs for the next round.
Tactic: Decide now, while things are going well, which assets you would sell and which statements you would make if your portfolio fell by half, so that in a crisis you are carrying out a plan rather than improvising.
Principle 11
Keep someone in the room who can say no
The pattern in Son's worst decisions is the absence of resistance. He overrode his lieutenants on WeWork. Tadashi Yanai, one of the few directors who could stand up to him, left in December 2019. The side deals that left Son owing SoftBank $5.1 billion, and the sale of the Asian venture arm to his brother, drew governance criticism because nobody inside had the standing to object. Seventy-one percent voting control gives a founder enormous freedom and very little protection from himself.
Tactic: If your control is effectively total, deliberately appoint people with independent wealth or reputation who can veto you on specific categories of decisions, and treat their departure as a warning sign.
Principle 12
Turn outsiderhood into access
Son was shut out of the paths available to a Japanese executive, so he built paths no Japanese executive used. He called Fujita sixty times. He took a high school exam with a dictionary. He became the Old-Man-Killer in Tokyo and a pizza-party regular in Silicon Valley. He had a cherry-tree conversation with Jobs, a forty-five-minute conversation in Riyadh, and repeated appearances alongside Trump. He is comfortable in places where he does not naturally belong, and he does not expect the doors to open on their own, so he works harder at opening them than the insiders do. The step stool at the White House was a small version of the same thing.
Tactic: List the rooms your background was supposed to keep you out of, find the one person in each who can get you in, and be more persistent than seems polite in getting to them.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Masayoshi Son, to The New York Times, 1999If I had stayed all the time in Japan, I probably would have become much more conservative, just as other Japanese.
— Masayoshi Son, to The New York Times, July 1999For the next 10 years, we do nothing but Internet. Because that's where the whole society is going to have a big revolution.
— Masayoshi Son, SoftBank Group Report 2025In that moment, my finger trembled. I was so overwhelmed with emotion that I could not stop crying. It felt as if I were witnessing the instant when humanity began creating something more intelligent than itself.
— Masayoshi Son, Message from Chairman and CEO, SoftBank GroupSome people say that SBG changes its core business too often, but in reality, our core business has remained unchanged all these years. Everything we do is inspired by our corporate philosophy, "Information Revolution — Happiness for everyone." All that has changed is the means to achieve this end.
— Masayoshi Son, to SoftBank shareholders, June 2023I may be more at fault than Adam, for telling him to be more aggressive.
Maxims
- Bet on acceleration. A high trailing multiple is a poor guide to value in an industry that is about to grow exponentially.
- Own the chokepoint. Distribution, standards, and capacity outlast whichever application is popular this year.
- Buy what others have abandoned, if you bring what they lacked. A turnaround needs an advantage, not just a low price.
- Never sell the lifeboat. One large, liquid winner is what lets you afford every other risk.
- Write down the founder's worst trait. Conviction that has never been tested is just affection.
- Sell the product before it exists. The exclusive goes to whoever imagined it first and kept in touch.
- Size cuts both ways. Overwhelming capital raises the cost of mistakes as much as the odds of winning.
- Keep the mission fixed and the means flexible. A permanent purpose makes it easy to let go of any single business.
- Apologize early and specifically. An admission made in time keeps the credibility you will need for the next bet.
- Make sure someone can stop you. The more control you hold, the more you need someone with the standing to say no.
For the fullest account of the man behind these lessons, see Lionel Barber's Gambling Man. For an earlier, more admiring view from inside Japan, see Atsuo Inoue's Aiming High.
