Last Updated on July 7, 2026 by Eytan Bijaoui
Forty billion dollars. Unsecured. Twelve-month term.
That’s the loan SoftBank just closed with JPMorgan, Goldman Sachs, and four Japanese megabanks to fund its $30 billion stake in OpenAI. Not equity. Not convertible notes. A straight-up loan from the most risk-averse institutions on the planet.
And I think most people are reading this story completely wrong.
The Number That Should Change How You Think About AI
The hot take is obvious. SoftBank, the company that turned WeWork into a $47 billion cautionary tale, is borrowing money it doesn’t have to bet on AI. Cue the “here we go again” crowd.
But zoom out for a second.
JPMorgan Chase doesn’t lend $40 billion unsecured because Masayoshi Son asked nicely. Goldman Sachs didn’t join this syndicate because they’re feeling generous. These are institutions that survive by being right about risk. They’ve seen every bubble, every crash, every “this time it’s different” pitch in financial history.
They said yes.
That’s not a signal about SoftBank. That’s a signal about AI.
What the Banks Actually Know
Here’s what JPMorgan’s risk committee would have evaluated before approving this: OpenAI is on pace for $20 billion in annualized revenue this year. That’s up from $3.7 billion last year. A 5x increase in twelve months. The company is preparing for an IPO that could value it at $1 trillion. And the loan term is twelve months, which means the banks expect an IPO or refinancing event within that window.
So the banks aren’t betting on AI hype. They’re betting on OpenAI’s cash flows, its IPO timeline, and its revenue trajectory. They did the math. The math worked.
SoftBank’s total OpenAI exposure is now over $60 billion. That sounds insane until you realize the potential upside on a $1 trillion IPO. If OpenAI goes public at even $800 billion (conservative), SoftBank’s stake could be worth $120 billion or more. The loan pays for itself many times over.
This is not WeWork math. WeWork had $1.8 billion in revenue and $1.9 billion in losses when SoftBank doubled down. OpenAI has $20 billion in revenue and the most dominant product in consumer AI.
But Here’s What’s Actually Scary
The loan was signed on March 27. Three days earlier, US startup funding data for March showed the sharpest slowdown of the year. American startups raised about $13 billion total in March, down dramatically from February’s $189 billion (though most of that was OpenAI’s own round).
And all of this is happening against the backdrop of the Iran War, which started February 28 and has sent stock markets tumbling. The S&P 500 is down. Asian private equity fundraising hit its lowest level in a decade. Investors everywhere are pulling back.
Everywhere except AI.
That’s the K-shaped venture market we’ve been watching all month, playing out in real time. Capital is flowing upward to a tiny group of AI mega-companies while everyone else watches their funding dry up. When banks lend $40 billion to one AI company during a war, while thousands of startups can’t close a $2 million seed round, something structural has shifted.
The OpenAI IPO Is the Real Story
I think most founders are sleeping on what happens next.
If OpenAI goes public in Q4 2026 at anything close to $1 trillion, it creates the largest tech liquidity event since Facebook’s IPO. SoftBank, Amazon (which put in $50 billion), Nvidia ($30 billion), and hundreds of early employees all get liquid.
That money goes somewhere. A chunk of it funds new startups. It happened after Google’s IPO. It happened after Facebook’s IPO. It happened after Alibaba’s IPO. Every mega-liquidity event creates a cascade of angel investments, new funds, and first-time founders who suddenly have the capital to try something.
And this one is bigger than any of them.
The downstream effects could be enormous. New AI-focused funds. Ex-OpenAI founders launching competitors. An entirely new generation of startups funded by people who actually understand AI, not just VCs who read about it.
Why This Makes Your Life Harder (and Maybe Better)
If you’re a founder right now, here’s the honest truth about what this means.
The harder part: The competitive landscape is about to get brutal. When the biggest banks in the world are backing AI companies, when entire categories of enterprise software are being rebuilt by AI agents, and when a single company can absorb $110 billion in a single round, you are not competing on a level playing field. You never were, but now the gap is visible.
The maybe-better part: Institutional validation of AI means the market isn’t going away. If JPMorgan is lending $40 billion for AI, your enterprise customers are going to keep buying AI products. The budgets are real. The demand is real. The question isn’t whether AI companies will make money. It’s which ones.
And here’s the thing I keep coming back to: the companies that raised the most in the last bubble (2021) were mostly horizontal platforms trying to be everything to everyone. The companies winning now are the ones picking a specific industry, a specific workflow, and owning it completely. Harvey at $11 billion for legal AI. Figure at $39 billion for humanoid robotics. The market is rewarding depth, not breadth.
What Founders Should Actually Do With This Information
I’m going to resist giving you a numbered list because the situation is more nuanced than that.
The core insight is this: we’re entering a period where AI is no longer a startup bet. It’s an institutional bet. That changes the game in ways that aren’t immediately obvious.
If you’re building an AI company, your competition just got better funded. But your customers also just got more committed. The enterprises that were “evaluating AI” in 2025 are now “deploying AI” in 2026, partly because their banks, their boards, and their competitors all decided AI is real.
If you’re building a non-AI company, you need to figure out how AI makes your thing better or cheaper. Not because AI is magic, but because your competitors will. And they might be funded by someone who just made a fortune on the OpenAI IPO.
If you’re pre-revenue and looking for funding, understand that the bar just went up. VCs aren’t going to fund your AI wrapper when they can put money into companies with actual revenue and actual differentiation. The era of “we’re like X but with AI” is genuinely over. Maybe I’m wrong about the timing, but the direction is clear.
The $40 Billion Question
SoftBank has been wrong before. Spectacularly wrong. The Vision Fund lost $32 billion on WeWork, and the scars are still visible on every pitch deck that gets rejected for “not having unit economics.”
But SoftBank has also been right before. Their early bet on Alibaba turned $20 million into $60 billion. Sometimes the crazy bet is just early.
The difference this time isn’t SoftBank’s judgment. It’s that five of the world’s largest banks ran their own analysis and came to the same conclusion. When the entire financial system agrees on something, it doesn’t mean they’re right. But it does mean the world is about to move in that direction whether you agree with it or not.
The $40 billion has already been lent. The chips are on the table. The only question that matters for founders is: what are you building while the biggest players are all looking in the same direction?
Because in my experience, the best opportunities usually show up in the places nobody’s watching.
