THE MAN WHO SAW THE STACK

Masayoshi Son, SoftBank, and the 40-year thesis that is finally coming true

On the morning of January 21, 2025, a Japanese businessman stood at a podium in the East Room of the White House alongside the President of the United States, the CEO of OpenAI, and the founder of Oracle. He announced the largest private investment in AI infrastructure in history — $500 billion, committed to building the computing backbone of the next civilisational era. Camera flashes. Handshakes. History being made.

His name appeared briefly in the coverage. Most Western readers scrolled past it.

This is the central paradox of Masayoshi Son（孫正義）. He is, by any serious measure, the most consequential Japanese business figure of the AI era — a man whose capital decisions of the past four decades are now structurally embedded in the technology that every analyst, investor, and executive is scrambling to understand. And yet he remains, outside the circles of finance and geopolitics, largely invisible. No mythology like Jobs. No cultural saturation like Musk. No boardroom reverence like Buffett.

The invisibility is not accidental. It is, in a way, the point. Son has never built for recognition. He has built for the moment when the thesis converges — and that moment, by any reasonable measure, is now.

I. The Outsider

To understand why Son thinks the way he does, you need to understand what Japan said to him before he said anything back.

He was born in 1957 in Tosu, a small city in Saga Prefecture, into a Korean-Japanese family living in poverty at the margins of postwar Japanese society. His father ran a small business that included illegal alcohol distillation and pig farming. The family lived, by Son's own account, in a shanty near the railway tracks. As a Zainichi Korean — one of the ethnic Korean community born and raised in Japan but historically denied full citizenship — he carried a social designation that closed doors before he could knock on them. Japan's postwar economic miracle was being built around him, but not for him.

He left for the United States at sixteen, finished high school in California in three weeks by passing a proficiency exam, and enrolled at the University of California, Berkeley. What happened next tells you everything about how his mind works.

At Berkeley, he gave himself a discipline: spend five minutes each day on an invention. Not a product. An invention. He made a list of criteria — it had to be something he could work on for the next fifty years, something that would matter, something at the intersection of computers and human need. Within a year, he had designed a pocket electronic translator, persuaded university professors to help him build a prototype, and sold the patent to Sharp Corporation for $1.7 million. He was nineteen.

He then imported Japanese arcade game machines to the United States, placed them in campus restaurants, and hired UC Berkeley students to manage them — generating nearly $4 million in revenue before graduation. He arrived in Japan in 1980 not as a supplicant seeking institutional permission, but as a young man with capital he had built with his own hands and a conviction that was already fully formed.

SoftBank was founded in 1981. On the first day, Son stood on an apple crate in front of two part-time employees and told them the company would, within five years, be handling billions in revenue and would one day become one of the largest companies in Japan. Both employees quit that day. He continued anyway.

What the conventional biography misses is what the origin story actually demonstrates. Son didn't beat the system. He circumvented it entirely. The Zainichi Korean who couldn't access Japan's elite university pipeline went to Berkeley instead. The man who couldn't inherit a keiretsu network invented his own entry capital. The outsider who was denied institutional legitimacy built a company so large that legitimacy came to him. Every constraint became a bypass route.

This is not a story about resilience. It is a story about a mind that never accepted the architecture of the room it was supposed to stay in.

II. One Thesis, Not Many Bets

The conventional narrative of Son is a man who makes enormous, concentrated bets — sometimes brilliantly, sometimes catastrophically. This framing is wrong, or at least incomplete. The bets are not separate decisions. They are instalments on a single conviction he has held, publicly and consistently, since the early 1980s: that intelligence — artificial, networked, distributed — would become the organising infrastructure of human civilisation, and that the companies controlling its architecture would be the defining enterprises of the next century.

Every major SoftBank investment is legible through this lens.

In the mid-1990s, Son moved aggressively into internet infrastructure in Japan and the United States, acquiring stakes in hundreds of online companies. His paper net worth briefly exceeded $75 billion in early 2000 — making him, for a few weeks, the wealthiest person on earth. When the dotcom bubble burst, SoftBank's portfolio collapsed by more than 90%. Son lost more money, on paper, than any individual in history. He did not pivot. He did not diversify into something safer. He rebuilt on the same thesis.

In 1999, at a "speed-dating" session for Chinese entrepreneurs in Beijing, he met a former English teacher named Jack Ma who was building an online marketplace for Chinese manufacturers. The meeting lasted six minutes. Ma had no business plan and zero revenue. Son committed $20 million on the spot — reportedly pushing to invest $40 million, with Ma negotiating him down. He later said he invested because of Ma's eyes. "Strong, shining eyes," he told Bloomberg. "I could feel his passion."

That $20 million, invested in Alibaba in 2000, eventually returned over $150 billion. It is, by most measures, the greatest single venture capital investment in history.

In 2016, Son agreed to acquire ARM Holdings — the British chip architecture company whose instruction set is embedded in virtually every smartphone on earth — for $32 billion. At the time, it looked expensive and strategically odd. ARM was profitable but slow-growing. The logic escaped most analysts. Son's argument was straightforward: as artificial intelligence moved from cloud servers into devices, into vehicles, into physical infrastructure, ARM's architecture would be inside every node of that network. He was buying the instruction set of the intelligence layer.

He was right. ARM listed on Nasdaq in September 2023, and SoftBank — which retained approximately 90% of the company — watched its value climb past $120 billion. The acquisition that looked overpriced in 2016 now looks like the most important single purchase of the AI era.

III. The Cost of Vision

There is a structural tension at the heart of SoftBank that is essential to understanding both its failures and why those failures do not invalidate the thesis.

A visionary investment philosophy requires time. A fund structure requires returns. These two imperatives are not compatible at the same horizon, and the collisions between them have produced the most expensive decisions in SoftBank's history.

In 2017, Son launched the Vision Fund — a $100 billion vehicle backed primarily by Saudi Arabia's Public Investment Fund and Abu Dhabi's Mubadala. It was the largest technology investment fund ever assembled. The thesis was the same: back companies building the intelligence infrastructure of the future, at scale, before the market understood what it was looking at. The execution was different: the fund had a defined term, quarterly reporting obligations, and LP relationships to manage. Vision required patience. The fund required performance.

The result was a series of decisions that made sense as fund management and destroyed value as investing. Son's team poured capital into companies — WeWork most visibly — that had compelling growth narratives but no structural moat, no path to profitability, and valuations that only made sense if the narrative never broke. When it broke, the Vision Fund recorded losses of tens of billions. Son appeared at a press conference and said, with characteristic directness, that his own judgment had been poor.

Nvidia tells the same story more cleanly. SoftBank's Vision Fund acquired approximately $4 billion worth of Nvidia shares in 2017, capturing the early phase of the GPU-driven AI compute boom. By 2019, with the Vision Fund needing liquidity after WeWork and other write-downs, they sold. The position they exited at roughly $6 billion would have been worth over $150 billion had they held it to 2024. Then, in October 2025, SoftBank sold its remaining Nvidia position — 32.1 million shares — for $5.83 billion, days before Nvidia's market capitalisation crossed $5 trillion.

The thesis was right. Twice. The fund structure forced the exit. Twice.

This is not a story about bad judgment. It is a story about the permanent incompatibility between visionary capital and institutional capital management. Son understood what Nvidia would become before almost anyone else on earth. He could not hold it because the vehicle he built to deploy capital had obligations the holding required him to breach.

The WeWork losses were genuine errors — investments in narrative rather than architecture. But the Nvidia exits were structural, not intellectual. The distinction matters if you are trying to assess what SoftBank actually is.

IV. The Convergence

In December 2016, three weeks after Donald Trump won his first presidential election, Masayoshi Son flew to New York, rode the elevator to the 26th floor of Trump Tower, and pledged $50 billion in US investment and 50,000 jobs. Cameras captured the two men holding up a piece of paper with the numbers written on it, both grinning. Most observers read it as a photo opportunity.

It was the opening of a decade-long strategic relationship.

In December 2024, Son was back — this time at Mar-a-Lago — pledging $100 billion and 100,000 AI-focused jobs. On January 21, 2025, the second day of Trump's second term, Son stood at the White House podium alongside Sam Altman and Larry Ellison and announced Stargate: up to $500 billion over four years, dedicated to building the physical infrastructure of American AI — data centres, power systems, semiconductor fabrication capacity. The largest private infrastructure commitment in history.

Son is not close to Trump because they share a worldview. He is close to Trump because Son goes directly to whoever holds executive authority over the markets and regulatory environments he needs. Before Trump in 2016, he had met India's Modi and South Korea's Park within the same period. He has maintained a thirty-year relationship with Larry Ellison, who introduced him to Steve Jobs in the 1990s. Son built his Silicon Valley credibility across decades — he is not a Japanese outsider seeking Western access. He is a peer who has been in the room since the technology was young.

What Stargate represents, beyond the headline number, is Son's final move in a forty-year game. He has backed the distribution layer (Yahoo Japan, early internet infrastructure). He has backed the application layer (Alibaba, the Chinese internet economy). He has backed the chip architecture layer (ARM). Now he is building the physical layer — the land, the power, the cooling, the compute density that the intelligence era requires at scale.

The Project Crystal Land proposal — a reported $1 trillion AI and robotics complex in Arizona, incorporating a free-trade zone alongside TSMC — extends the logic further. Son is not positioning SoftBank as an investor in the AI era. He is positioning it as the contractor building the room the era happens in.

SoftBank Group's stock tells its own story. From a long period trading around ¥10,000 pre-split — equivalent to roughly ¥2,500 post the 1-for-4 split executed in 2024 — the stock reached an all-time high of ¥9,074 in June 2026. From trough to all-time high, the trajectory mirrors the convergence Son has been predicting for forty years.

On June 1, 2026 — as this issue was being written — Son stood beside French President Macron in Paris and announced €75 billion in AI data centre investment across France. Three gigawatts of compute capacity by 2031. The same week, he told CNBC that the AI revolution is "probably fifty times bigger than the dot-com boom." The man who lost more money than anyone in history when the dot-com bubble burst is now saying the dot-com boom was a footnote. He has earned the right to that comparison.

DATA CARD — THE TRAJECTORY

V. The Diagnosis

Japan is one of the great engineering civilisations. Its manufacturing precision, its materials science, its logistics infrastructure, its quality systems — these are genuine and deep structural advantages that have shaped global industry for seventy years. The country produces craftsmen of the highest order, in every field it touches.

What Japan has not produced, with any consistency, is founders who think at civilisational scale.

This is not a failure of intelligence or capital. Japan has both in abundance. It is a failure of cultural permission. The consensus-formation systems that make Japanese organisations exceptionally stable and capable of long-term execution also make them exceptionally hostile to the kind of unilateral, contrarian, decade-horizon thinking that produces platform companies. The salaryman loyalty structure, the nemawashi consultation process, the deep social cost of visible failure — these are not bugs in the Japanese business system. They are features. They produce Toyota and Fanuc and Keyence: extraordinary companies built on refinement, precision, and the disciplined execution of proven models.

They do not produce someone who meets an entrepreneur for six minutes with no business plan and commits $20 million because of his eyes.

Masayoshi Son could only think the way he thinks because he was never fully inside the system. His Korean-Japanese identity placed him outside the social architecture that distributes Japan's institutional permission. His American education gave him a reference frame that had no ceiling on ambition. His early capital was self-generated, not inherited — which meant he owed no deference to the networks that would have constrained it. He built everything from the outside in, and in doing so demonstrated what becomes possible when the system's permission structures do not apply.

The question Japan now faces is whether Son's example is legible inside the system — whether the founders who grew up watching SoftBank can extract the transferable principle (think in platforms, not products; bet on architecture, not applications; hold the thesis through the disconfirming evidence) without the outsider status that made the principle available to him. Or whether the cultural conditions that produced him are precisely the conditions that cannot be replicated from inside.

Son himself seems to believe the former. He talks about Japan's AI future with the same conviction he brought to his two Berkeley employees in 1981. He sees Stargate not as an American project that SoftBank is financing, but as the beginning of a capital and technology corridor that runs through Japan — through ARM's architecture, through Japan's materials and semiconductor equipment companies, through the AI robotics industry he has been building through SoftBank's portfolio.

The man who saw the stack, forty years ago, is now building it. Whether Japan can see what he sees — and whether it matters if it can't — is the question that will define the country's position in the next era of global capital.

For the investors and executives reading this: the answer to that question will be visible in where the capital flows. Son has spent his career making his bets legible to anyone willing to read capital behaviour rather than corporate narrative. He is still betting. The direction has not changed.

Of all the figures shaping the AI era, Son may be the most underrated — largely invisible to the audiences who follow Jensen Huang and Larry Ellison daily, yet regarded by both as a peer. That gap between recognition and significance is precisely what made this piece worth writing.

The Hub Lens 維點 does not provide investment advice. All figures cited are for analytical and editorial context only.