There are ambitious corporate investments, and then there is Masayoshi Son.
Whenever I think the artificial-intelligence spending boom has finally reached a number large enough to make everyone involved reconsider their life choices, SoftBank shows up with another zero.
This time, SoftBank Group is heading into the high-yield bond market looking to raise more than $11 billion, with the money helping finance its enormous push into artificial intelligence and, more specifically, its increasingly concentrated bet on OpenAI.
And when I say “high-yield,” I am using the polite Wall Street terminology.
We are talking about junk bonds.
Lots of them.
SoftBank is marketing approximately $10 billion of dollar-denominated debt along with €1 billion, or roughly $1.1 billion, of euro-denominated bonds. If completed at that size, the transaction would rank among the largest corporate high-yield bond offerings ever.
That's impressive.
It's also slightly terrifying.
Because beneath all the excitement surrounding artificial intelligence, data centers, chips, robots, autonomous agents and whatever digital miracle we're promised next Tuesday, somebody eventually has to pay for all this stuff.
Increasingly, that somebody is the bond market.
And SoftBank appears more than happy to hand it the bill.
Masayoshi Son Has Never Been Particularly Interested in Thinking Small
If you've followed SoftBank for any length of time, none of this should be surprising.
Masayoshi Son doesn't seem genetically equipped for incrementalism.
He doesn't wake up and wonder whether SoftBank should increase its investment somewhere by 4%.
He thinks in terms of technological revolutions.
Entire industries.
Decades.
Trillions of dollars.
And artificial intelligence has become the defining bet of this phase of his career.
SoftBank agreed in February to make an additional $30 billion investment in OpenAI, which would bring its cumulative investment in the company to approximately $64.6 billion and give it an expected ownership stake of roughly 13% once completed.
Just stop there for a moment.
Nearly $65 billion.
Into one private company.
I don't care how excited someone is about artificial intelligence. That number should at least cause a brief pause before clicking the imaginary “confirm investment” button.
SoftBank structured the latest $30 billion commitment in three $10 billion installments during 2026. The first was completed in April, the second in July, and the final $10 billion payment is scheduled for October 1.
Now the bond market is becoming part of the machinery supporting that investment.
And that's where this story gets much more interesting to me.
Because SoftBank isn't merely betting on AI.
It's borrowing heavily to increase the size of that bet.
There is an enormous difference between those two things.
The Price of Believing in the Future Is Approaching 10%
The bond terms tell us something important about how investors currently view SoftBank's strategy.
SoftBank's dollar offering is divided among approximately 3.5-year, 5.5-year and 7.5-year maturities, while the euro bonds are being offered across four- and six-year maturities.
And investors aren't exactly lending SoftBank money at Treasury-like rates.
Early price discussions reportedly placed the shortest dollar tranche around 8.75% to 8.875%, with longer maturities climbing as high as roughly 9.75% to 9.875%.
Basically, we're flirting with 10%.
That's real money.
At 2% interest, corporations can convince themselves that almost anything is financially brilliant.
At 9% or 10%, the investment eventually needs to do something besides appear in a PowerPoint presentation under a heading called “Transforming the Future.”
Those interest payments arrive whether the AI revolution is ahead of schedule, behind schedule or currently rebooting.
This is the part of the AI investment boom I think equity investors sometimes underestimate.
Technology headlines naturally focus on upside.
Artificial general intelligence.
Autonomous software agents.
AI-powered drug discovery.
Robotics.
Self-driving vehicles.
Data centers.
Semiconductors.
Productivity miracles.
But debt doesn't care about narratives.
Debt has a coupon.
The interest payment shows up on schedule.
That's what makes SoftBank's bond offering so fascinating.
Son may eventually be completely right about AI.
But being right about the technological direction of the world is not automatically the same thing as being right about the price you paid to participate in it.
Investors Apparently Want In Anyway
Here's where the story gets even stranger.
Despite those enormous yields and the considerable risks surrounding the AI spending boom, investors appear interested.
Very interested.
SoftBank reportedly attracted more than $20 billion in preliminary demand for the offering, nearly twice the roughly $11 billion initially targeted. Those indications were preliminary and nonbinding ahead of final pricing, but the appetite is still notable.
I can understand why.
Imagine you're managing a massive fixed-income portfolio.
You have money that must be invested somewhere.
Then SoftBank arrives offering debt yielding somewhere around 9%.
It's backed not by some tiny company hoping its third attempt at a business model works, but by an investment conglomerate with holdings that include enormously valuable technology assets.
That gets attention.
SoftBank's bonds are still speculative grade—Fitch assigned the proposed notes a BB+ rating, just below investment grade—but the company owns assets that potentially give it considerable financing flexibility.
So investors are looking at the yield and thinking:
Almost 10% from SoftBank?
Maybe I'll hear them out.
I probably would too.
But I would be reading every page of the prospectus.
Twice.
The AI Boom Is Becoming a Debt Boom
What bothers me isn't necessarily this one SoftBank offering.
It's the broader pattern.
According to reporting citing Goldman Sachs credit strategists, global AI-related debt issuance has already exceeded $575 billion in 2026.
Think about how quickly the AI story has evolved.
Not long ago, the investment thesis centered largely on semiconductor companies.
Then it became cloud computing.
Then data centers.
Then electricity.
Then nuclear power.
Then networking.
Then cooling systems.
Then land.
Then transmission infrastructure.
Now we're increasingly talking about the financing architecture necessary to fund all of it.
AI isn't just a technology story anymore.
It's becoming a capital-markets story.
That changes the risk profile.
There's a huge difference between Microsoft funding AI development from mountains of operating cash flow and companies increasingly borrowing to finance infrastructure based on future demand assumptions.
The more leverage enters the ecosystem, the less room there is for those assumptions to be wrong.
That doesn't mean an AI collapse is coming.
It means mistakes become more expensive.
Imagine building a data center assuming AI compute demand will grow 50% annually and discovering it grows 20%.
Twenty percent growth sounds fantastic.
Unless your financing model required 50%.
That's how investment booms become financial problems.
Not because the technology failed.
Because expectations outran economics.
SoftBank Has Been Preparing for This
To SoftBank's credit, none of this appears accidental.
The company arranged a $40 billion bridge loan facility in March, primarily to support its additional OpenAI investment and other corporate needs.
Bridge financing is basically what the name suggests.
You borrow money temporarily until you arrange longer-term funding.
SoftBank later said it would repay and refinance that facility through a combination of asset-backed financing, bond issuance, potential asset sales and other sources.
Its CFO has made clear that the company prefers asset-backed financing where practical because selling valuable holdings means surrendering potential future upside. At the same time, SoftBank views global bond markets as an important source of longer-term financing.
This current bond offering fits directly into that strategy.
SoftBank also announced earlier this month that it would repay $25.9 billion of the outstanding bridge facility after having drawn $30 billion against the $40 billion line.
So this isn't simply:
“SoftBank suddenly needs $11 billion.”
It's more complicated.
SoftBank has been moving money across short-term and long-term financing structures while funding enormous investments in OpenAI and other AI-related businesses.
Financially, that distinction matters.
But economically, the big picture doesn't change much.
SoftBank is using its balance sheet aggressively because Son believes the value created by artificial intelligence will dwarf today's financing costs.
That's the bet.
Son Thinks AI Could Become a $46 Trillion Piece of the Global Economy
To understand why SoftBank is comfortable doing this, you have to understand the scale of Masayoshi Son's expectations.
Son has argued that AI-related industries could eventually represent approximately 20% of global economic output by 2040, equivalent to something like $46 trillion.
If he's remotely correct, $65 billion invested in OpenAI might someday look surprisingly reasonable.
That's the strange thing about transformational technologies.
The early numbers always sound ridiculous.
Amazon spending billions building warehouses sounded excessive until e-commerce became normal.
Cloud-computing investment looked enormous until companies moved their infrastructure online.
Semiconductor fabrication plants cost absurd amounts of money until the global economy became dependent on advanced chips.
Every major technological shift creates a period where investors have trouble distinguishing visionary spending from reckless spending.
Sometimes the answer is both.
That's probably the most intellectually honest way I can describe SoftBank's current strategy.
It may be visionary.
It may also be reckless.
Those aren't mutually exclusive.
OpenAI Doesn't Need to Fail for SoftBank's Bet to Become Difficult
This distinction matters enormously.
People often discuss investment risk as though there are only two outcomes.
Success.
Failure.
Reality is messier.
OpenAI could become one of the most important companies in the world and SoftBank could still earn disappointing returns if it paid too much.
That is the uncomfortable truth of investing.
The greatest company on Earth can still be a terrible investment at the wrong valuation.
Likewise, an extraordinary technological revolution can produce mediocre financial returns for companies that spend too aggressively trying to capture it.
OpenAI's computing requirements are enormous.
The AI ecosystem surrounding it requires chips, servers, networking equipment, power plants, data centers and financing on a scale that would have sounded absurd several years ago.
SoftBank clearly believes it is positioning itself near the center of that ecosystem.
But the capital intensity means the bar for financial success keeps rising.
If you invest $1 billion and turn it into $3 billion, everyone throws a party.
If you invest $65 billion and eventually own something worth $80 billion, congratulations: you've created a gigantic company and possibly a mediocre investment.
Scale changes the math.
There's Another Number I Keep Coming Back To: 25%
SoftBank says its financial policy is to maintain its loan-to-value ratio below 25% under normal market conditions, with an upper threshold of 35% during extraordinary circumstances. It also aims to maintain enough liquidity to cover at least two years of bond redemptions.
I actually like that discipline.
It gives investors a framework for judging whether management is stretching the balance sheet too aggressively.
SoftBank isn't pretending debt doesn't matter.
Its executives clearly understand the risks.
The problem is that loan-to-value ratios depend on both sides of the equation.
Debt matters.
But so does asset value.
And technology asset values can move quickly.
SoftBank knows this better than almost anyone.
It has lived through massive swings in the value of Alibaba, Arm and its Vision Fund investments.
When your assets rise dramatically, leverage suddenly looks conservative.
When those same assets collapse, identical debt can look uncomfortable almost overnight.
That's one reason the AI boom's valuation levels matter so much.
SoftBank's financing capacity is partly supported by valuable technology holdings.
If AI valuations remain elevated or continue rising, everything looks manageable.
If valuations suddenly reset downward, the math gets less pleasant.
Arm Is an Important Piece of the Puzzle
SoftBank's enormous stake in Arm gives it a particularly valuable asset to lean against.
That matters because Son isn't financing his AI ambitions from an empty cupboard.
SoftBank owns substantial stakes in businesses that could potentially be sold, pledged or otherwise monetized.
Its CFO has openly discussed using portfolio holdings through asset-backed financing while trying to avoid unnecessary sales that would sacrifice future upside.
That's intelligent capital management.
It also creates an interesting circularity.
AI enthusiasm helps support semiconductor valuations.
Higher semiconductor valuations increase SoftBank's asset value.
Higher asset value increases financing capacity.
That additional financing helps SoftBank invest more money into AI.
As long as the cycle continues upward, everybody looks brilliant.
The real test comes when markets stop cooperating.
That isn't a prediction.
It's simply how leverage works.
Bond Investors Are Starting to Ask Harder Questions About AI
SoftBank's timing is interesting because credit investors are becoming more selective toward AI-related borrowing.
Recent market reporting suggests AI-linked borrowers are increasingly paying wider credit spreads than the broader corporate market as investors worry about the sheer volume of borrowing needed for chips, data centers and other infrastructure.
That's healthy.
I would be more worried if investors simply threw money at anything containing the letters “AI.”
Credit markets are supposed to be skeptical.
Equity investors get paid to imagine what might go right.
Bond investors spend much of their lives asking what could go wrong.
If SoftBank wants more than $11 billion to help fund its AI ambitions, somebody should be asking uncomfortable questions.
What happens if OpenAI's growth slows?
What happens if AI model costs collapse and infrastructure becomes less valuable?
What happens if cheaper competitors emerge?
What happens if regulatory requirements increase?
What happens if today's enormous compute requirements prove temporary?
What happens if enterprises decide they're unwilling to pay enough for AI services to justify the capital being deployed?
None of those outcomes require AI to disappear.
They simply require the financial return on AI investment to be lower than expected.
The Comparison With the Dot-Com Era Is Too Easy
I hesitate to call this another dot-com bubble because that comparison gets abused.
Any time technology stocks rise quickly, someone dusts off a photograph of Pets.com.
Today's AI boom is fundamentally different in several important ways.
Many of the largest spenders are extraordinarily profitable companies.
Nvidia sells enormous quantities of actual hardware.
Microsoft, Amazon, Alphabet and Meta generate tens of billions of dollars in cash.
OpenAI has a huge global user base.
There are real products.
Real customers.
Real infrastructure.
Real revenue.
This isn't a collection of companies going public because someone registered a domain name.
But financial bubbles rarely form around things that are completely fake.
They usually form around something genuinely transformative.
Railroads changed civilization.
Telecommunications changed civilization.
The internet changed civilization.
All three produced periods of spectacular overinvestment.
A technology can change the world while investors simultaneously lose fortunes financing it.
That's the paradox I'm watching with AI.
SoftBank Could End Up Looking Brilliant Again
It's easy to mock aggressive investors when they're doing something uncomfortable.
It's harder to remember all the times they were mocked before being proven right.
SoftBank invested early in Alibaba.
That became one of history's great venture investments.
Son acquired Arm when many investors questioned the deal.
Arm later became an enormously valuable piece of the AI semiconductor ecosystem.
Those experiences matter.
Son's willingness to place enormous concentrated bets has produced both spectacular successes and painful failures.
That's what makes SoftBank different from a conventional conglomerate.
It doesn't really behave like one.
It behaves more like a giant public venture-capital vehicle operated by someone who genuinely believes he can see the next technological era arriving before everyone else.
Sometimes that worldview looks prophetic.
Sometimes it looks expensive.
Usually it looks boring only in hindsight.
But 10% Money Changes My Attitude
This is the part I can't ignore.
If SoftBank were financing everything internally, I would view the strategy differently.
Once the company starts offering nearly 10% yields to fund portions of the expansion, my standards go up.
The return on that capital needs to justify the financing cost.
Not immediately.
OpenAI obviously isn't being evaluated like a traditional mature company.
But eventually.
Debt converts an open-ended technological vision into a series of very specific financial obligations.
That's why I think this bond sale deserves more attention than another headline announcing another AI partnership.
It's evidence that the AI boom is entering a new stage.
The first stage was excitement.
The second stage was capital expenditure.
The third stage may be financing.
And once financing becomes central to a boom, I start watching the credit market almost as closely as the stock market.
The Most Important AI Indicator Might Eventually Be Bond Spreads
Investors obsess over Nvidia earnings.
They watch Microsoft Azure growth.
They count ChatGPT users.
They track data-center construction.
I think another indicator deserves attention:
the price companies must pay to borrow money for AI investment.
If AI-related bond spreads continue widening, that's information.
It means lenders are demanding more compensation for uncertainty.
If offerings struggle to attract buyers, that's information.
If projects require increasingly expensive financing to generate acceptable returns, that's information.
On the other hand, if AI investments begin generating enormous, predictable cash flows, credit markets will eventually recognize that too.
Debt markets have a wonderful habit of stripping excitement out of a story.
A bond investor doesn't care whether your keynote presentation had lasers.
They want their money back.
With interest.
So Is SoftBank Crazy?
I don't think that's the right question.
The better question is whether the eventual economic value of SoftBank's AI assets will grow fast enough to justify today's combination of valuation risk, concentration risk and financing cost.
Nobody knows.
Certainly not me.
And probably not anyone confidently posting a 17-part social-media thread claiming the answer is obvious.
What I do know is that SoftBank is making one of the largest concentrated bets on artificial intelligence anywhere in global finance.
Its expected cumulative OpenAI investment is approaching $64.6 billion.
It has used a $40 billion bridge financing facility as part of the funding strategy.
Now it is marketing more than $11 billion of high-yield bonds, with portions of the dollar debt potentially yielding close to 10%.
Meanwhile, the broader AI ecosystem has already generated hundreds of billions of dollars in debt issuance this year.
These aren't small numbers anymore.
We're watching an entire financial architecture being built around the assumption that artificial intelligence will generate extraordinary economic returns.
Maybe it will.
I actually think AI is going to be enormously important.
But believing AI will transform the economy doesn't require believing every dollar invested in AI will earn a great return.
Those are completely different propositions.
What I'm Watching Next
I'm going to be watching three things.
First, the final pricing of SoftBank's bonds.
Strong investor demand combined with tighter-than-expected final yields would tell me credit markets remain comfortable funding Son's AI strategy.
Second, OpenAI's economic trajectory.
User growth is fascinating, but eventually I want to see how revenue, margins and infrastructure spending evolve relative to the astonishing amount of capital flowing into the company.
Third, SoftBank's leverage.
The company has laid out clear financial guidelines. If it can continue making enormous investments while keeping its loan-to-value ratio disciplined and maintaining substantial liquidity, that will strengthen the argument that management understands exactly how aggressively it can push the balance sheet.
If those guardrails start weakening, I'll become considerably more cautious.
My Take
I don't view SoftBank's $11 billion bond sale as evidence that the AI boom is about to collapse.
I view it as evidence that the AI boom is growing up.
We've moved beyond enthusiastic venture capitalists writing checks to startups.
We're now talking about giant corporate bond offerings, multibillion-dollar bridge facilities, project financing, asset-backed loans and infrastructure investments measured in numbers normally associated with governments.
That makes AI more consequential.
It also makes it more financially fragile if expected returns disappoint.
Masayoshi Son appears willing to accept that risk because he believes the opportunity is historic.
Maybe he'll be right.
If OpenAI becomes one of the dominant economic platforms of the next several decades, people may someday look at SoftBank's $64.6 billion investment and wonder why everyone thought it was so outrageous.
But if AI economics prove less attractive than today's valuations imply, historians may look at nearly 10% junk bonds funding enormous private AI investments and wonder how the warning signs could have been any clearer.
That's what makes this fascinating.
SoftBank isn't simply predicting the future anymore.
It's borrowing billions against its conviction that the future will arrive.
And now the clock is running.
Because optimism can wait.
Debt payments can't.
Comments
Post a Comment