Masayoshi Son is the premium GenAI clown

Masayoshi Son has spent the past year transforming himself from an eccentric billionaire investor into GenAI’s premium clown, complete with trillion-dollar prophecies, accusations of blasphemy and a financial strategy apparently inspired by asking a chatbot whether borrowing tens of billions to buy more OpenAI shares sounds visionary.

The performance properly began with Stargate in January 2025. OpenAI announced a new company that would supposedly invest as much as $500 billion into American AI infrastructure over four years, with Son appointed chairman. The division of labour told you most of what you needed to know: OpenAI had “operational responsibility”, while SoftBank had “financial responsibility”. Sam Altman would run the technology. Oracle, Nvidia and various infrastructure partners would supply the machinery. Son would provide the money and gaze into the middle distance as though the Omnissiah had selected him as its authorised finance broker. OpenAI’s own announcement could hardly have described his role more clearly.

Stargate may still produce plenty of real infrastructure, assuming its remaining colossal headlines survives contact with financing, power availability, planning permission and reality. Son has also bought into robotics, energy and semiconductor businesses around the wider AI frenzy. None of that changes where the overwhelming financial weight sits: SoftBank is selling assets, raising debt and pouring more than $60 billion into OpenAI shares.

OpenAI builds the models. Nvidia supplies the chips. Oracle leases much of the computing infrastructure. Son buys a minority stake and delivers speeches suggesting that this makes him the chosen prophet of artificial superintelligence.

The machine god speaks through poultry:

Son’s recent public comments have gone from exuberant to genuinely fucking hilarious.

When investors raised the possibility that AI valuations and infrastructure spending might be forming a bubble, Son called the suggestion an “insult to AI” and “blasphemy against AI”. We have apparently reached the stage where a technology has developed religious sensitivities before achieving reliable factual accuracy, and asking whether the numbers add up risks angering the silicon gods.

At the same shareholder meeting, Son attempted to reassure investors by comparing SoftBank to a goose laying golden eggs.

“Eggs do not lay eggs, the goose lays the eggs,” he explained, before describing SoftBank as “the factory that lays the eggs” and asking how long he would have to fight to convince investors that “the goose did a good job”.

Imagine being a SoftBank shareholder watching tens of billions flow into one privately valued company, then receiving a lecture on reproductive poultry from the man arranging the loans. The goose has sold some of its possessions, borrowed against the farm and placed a vast proportion of its eggs inside Sam Altman’s industrial furnace. Anyone asking about the temperature is apparently committing heresy. Reuters covered the full sermon.

Son managed to become even more detached a few weeks later. He predicted that global AI investment would reach $5 trillion every year by 2040, AI-related industries would account for 20 per cent of global GDP, data centres would require three terawatts of electricity, and the world would contain 100 trillion autonomous AI agents.

Humanity would lose its place as Earth’s highest life form. Artificial superintelligence would become 10,000 times smarter than humans. Nuclear fusion would eventually power the whole glorious machine civilisation. Presumably the goose will oversee treasury operations.

Reuters noted that Son gave no explanation for how he calculated the $5 trillion figure or the 20 per cent of GDP. He simply declared confidence, called the bubble question “absurd” and suggested that anyone asking it did not understand AI. No workings, no methodology, just prophecy.

Perhaps Altman showed him a private GPT-6 demo and Son heard the machine god whisper directly through the speakers. Perhaps he spent too long talking to a chatbot that enthusiastically validated every civilisation-scale prediction he typed into it. We obviously do not know what happened behind closed doors, but the public result resembles a man who saw a spectacular demo, extrapolated it into digital divinity and then rearranged SoftBank’s balance sheet around never having to admit that the extrapolation might be bollocks.

AI can transform entire industries while investors simultaneously build a grotesque bubble around it. Son refuses to entertain that elementary distinction because he has already placed an enormous financial bet on treating technological potential and investment value as the same thing.

The supposed man from the future:

Before accepting Son’s role as the turtleneck-wearing prophet of GenAI, it helps to look at the actual history of this supposed future man.

His legend rests almost entirely on one investment. SoftBank put roughly $20 million into Alibaba in 2000, obtaining an early stake that eventually became worth tens of billions. It was one of the greatest investment returns ever recorded and has spent the following quarter-century covering a spectacular amount of financial stupidity.

Alibaba is repeatedly offered as proof that Son possesses supernatural foresight. His wider record suggests he made one phenomenal decision, became richer than God from it and then spent decades mistaking the result for a permanent ability to see around corners.

Amazon provides the funniest comparison.

Before Amazon floated, Son reportedly offered Jeff Bezos $100 million for 30 per cent of the company. Bezos wanted a valuation around $350 million rather than Son’s $300 million, leaving them roughly $30 million apart. The deal collapsed, and Son later described himself as stupid for missing it.

This is the same future-seeing financial wizard who somehow could not close a $30 million gap for almost a third of Amazon, yet has now located more than $60 billion for roughly 13 per cent of OpenAI after ChatGPT became one of the most famous products on Earth.

He failed to complete the cheap deal before the future became obvious. Now he is hurling borrowed billions at the expensive deal after every investor, consultancy, government department and idiot with a LinkedIn account has spent years shouting that AI will change everything.

Perhaps SoftBank genuinely could not find the extra money at the time. That hardly rescues the mythology. Son had direct access to Jeff Bezos, correctly identified Amazon as a major opportunity and still failed to bridge a comparatively tiny gap. A quarter of a century later, he appears so traumatised by missed opportunities that he intends to prevent a repeat by paying almost any price for OpenAI.

Selling Nvidia before the Nvidia boom:

Son’s Nvidia record is equally impressive, provided you enjoy watching a supposed futurist repeatedly sell the future before it arrives.

SoftBank’s Vision Fund built a 4.9 per cent position in Nvidia and sold it in 2019 for a return of around $3.3 billion. That looked respectable until Nvidia became the central hardware supplier of the GenAI boom and the discarded stake would have been worth well over $100 billion. Son later called it the fish that got away.

SoftBank then managed to acquire Nvidia shares again and sold those too, raising about $5.8 billion partly to support its OpenAI financing. Son reportedly said he was almost crying because he did not want to sell.

The grand strategic genius therefore sold the dominant AI chip company twice, including once to finance a minority stake in one of the largest buyers of those chips. Nvidia designs the scarce hardware everyone desperately needs. OpenAI consumes staggering amounts of it while burning cash. Son chose the customer.

That may work brilliantly if OpenAI captures enough of the future value. It may also become one of the funniest reallocations of capital in modern technology history.

Arm survived Son’s foresight:

Arm adds another layer to the legend.

SoftBank bought the British chip designer in 2016 for approximately $31 billion, with Son describing it as a foundational investment in the connected future. Arm’s engineers continued developing the architecture, expanding into servers, automotive systems and cloud computing while increasing research spending.

Four years later, Son agreed to sell the entire company to Nvidia for up to $40 billion.

That was his grand vision for one of the most strategically important semiconductor companies on Earth: buy it, hold it for a few years, then hand the whole thing to Nvidia for only a few billion dollars more than SoftBank originally paid.

The deal collapsed because regulators blocked it. Son did not experience a last-minute revelation about Arm’s importance. Competition authorities prevented him from cashing out. SoftBank then floated Arm and retained most of the shares, just in time for the AI market to attach an enormous premium to anything vaguely connected to chips.

Arm now forms the centrepiece of SoftBank’s renewed AI mythology. Son talks about it as though its survival inside the group demonstrates his foresight, when his preferred strategy was to sell the entire company shortly before its value exploded.

Arm’s engineers built the technology. Regulators preserved SoftBank’s ownership. Son now receives the credit.

Alibaba paid for the cemetery:

Once Amazon, Nvidia and Arm are placed beside Alibaba, Son’s reputation starts looking considerably less mystical.

Alibaba gave him enough money, status and credibility to survive a cemetery of failures. WeWork alone cost SoftBank more than $8 billion, after Son backed Adam Neumann’s office-rental cult at increasingly absurd valuations. Vision Fund investments produced tens of billions in write-downs across companies that had been sprayed with capital on the assumption that scale would eventually produce a business model.

His defenders treat each disaster as the unavoidable cost of thinking further ahead than ordinary people. Every failure becomes evidence of boldness. Every surviving asset proves genius. Every missed opportunity is quietly folded into the mythology, while Alibaba remains permanently displayed beneath museum lighting as proof that the next deranged bet could also work.

Perhaps Alibaba involved extraordinary judgement. It still represents one successful early bet rather than a lifelong pattern of technological foresight. Son has spent decades chasing trends, overpaying, missing opportunities, selling winners too early and confusing access to enormous amounts of capital with an ability to deploy it intelligently.

Alibaba did not prove Masayoshi Son could repeatedly predict the future. It gave him enough financial armour to remain celebrated as a visionary while repeatedly failing to do so.

Sixty-four billion dollars for someone else’s company:

SoftBank’s expected cumulative OpenAI investment is around $64.6 billion, giving it roughly 13 per cent ownership.

The scale deserves to be stated plainly. Son could not bridge a $30 million gap for 30 per cent of Amazon before its IPO. He can apparently find more than $60 billion for 13 per cent of OpenAI at a valuation already inflated by years of global AI hysteria.

SoftBank is receiving preferred shares, rather than building and controlling $64.6 billion of productive infrastructure. The money enters OpenAI and helps pay for staff, research, chips, data-centre contracts and the company’s enormous operating losses. OpenAI decides how it is spent. SoftBank receives a minority financial interest and hopes the eventual valuation rises far enough to justify the whole thing.

Son likes presenting himself as one of the architects of the coming machine civilisation. His primary contribution consists of selling assets, arranging loans and purchasing equity in Sam Altman’s company.

SoftBank reported a stand-alone cash position of about $22 billion at the end of March 2026. Even that generous figure included cash equivalents, short-term investments, bonds and borrowing capacity. The OpenAI commitment is approaching three times that total.

The difference came from financing. SoftBank arranged a $40 billion bridge facility largely to fund the OpenAI investment, with repayment expected through existing assets and other financing measures. Translated from corporate language, SoftBank borrowed now and intends to sell assets, issue longer-term debt, refinance the loans or borrow against other holdings later.

Son is converting liquid or borrowable assets into an enormous minority position in one volatile private company. He has no operational control over OpenAI, no control over its cash burn, no guarantee of an IPO and no control over whether public investors eventually accept the private valuation.

He does, however, have poultry metaphors.

The trillion-dollar escape hatch:

OpenAI’s rumoured $1 trillion IPO is currently treated as the event that vindicates Son and turns SoftBank’s enormous private investment into real wealth. That description skips several inconvenient stages between a valuation appearing on CNBC and money arriving in SoftBank’s bank account.

OpenAI was already valued at $852 billion after its March 2026 funding round. Reaching $1 trillion would add roughly 17 per cent, which sounds considerably less miraculous once you remember that Son is putting another $30 billion into the company at almost exactly this altitude. He has borrowed tens of billions to purchase near the top of the private market, then presented the next 17 per cent as confirmation that he saw the future.

At a $1 trillion valuation, SoftBank’s expected 13 per cent holding could theoretically be worth around $130 billion. The word theoretically is doing enough work to qualify for overtime. SoftBank would not immediately receive $130 billion. It would own a very large pile of newly public shares quoted at whatever price the relatively small IPO float happened to support.

The IPO proceeds would go to OpenAI unless SoftBank was allowed to sell some of its own shares in the offering. The remainder would probably be locked up for months, during which Son could stare lovingly at the number while being unable to realise it. When the lock-up ended, trying to sell tens of billions of dollars in stock would create exactly the sort of enormous insider overhang capable of murdering the price.

SoftBank’s preferred shares would convert into ordinary shares during the listing, giving Son liquidity at the cost of exchanging private contractual protection for exposure to the public market. New shares issued by OpenAI could dilute SoftBank’s ownership before it sells anything. Taxes, financing costs and the difficulty of unloading such a gigantic position would further reduce whatever amount eventually escaped the machine furnace.

A flotation could still inflate SoftBank’s net asset value and allow Son to borrow against the shares, which would presumably inspire another poultry-themed celebration. Borrowing against a volatile AI stock to repay debt used to purchase that same AI stock sounds less like successful monetisation and more like discovering a second storey in the casino.

The valuation itself remains demented. OpenAI says it is generating around $2 billion in monthly revenue, giving it an annualised run rate of approximately $24 billion. A $1 trillion flotation would price it at roughly 42 times revenue while it continues burning investor money by the billion. Reported shareholder figures put the first-quarter 2026 cash burn at $3.7 billion against $5.7 billion in revenue.

Son therefore needs more than an IPO. He needs public investors to accept the private market’s fantasy valuation, keep accepting it after the lock-up expires, absorb any SoftBank sales without collapsing the price and ignore the industrial quantities of cash disappearing into compute.

A $1 trillion OpenAI IPO would give the goose a magnificent golden egg to display in the investor presentation. Turning that egg into $130 billion without cracking it, crushing the market or revealing that it was mostly gold paint would be another matter entirely.