They Found $4 Trillion in Eighteen Months. They Just Never Looked for You.

For forty years the answer was “how will you pay for it.” Then $4 trillion appeared for three companies that lose money—no vote, no debate. Here’s the receipt.

Jun 05, 2026

by Lawrence Winnerman

2026 is a hell of a year.

It’s the 250th birthday of the American experiment. It’s the midterm election that decides whether the experiment gets a 251st. It’s the year Grand Theft Auto VI—a video game about stealing everything that isn’t nailed down—is projected to clear $3 billion, possibly the biggest entertainment launch in human history.

And it’s the year roughly $4 trillion in private capital converged on three companies as they lined up to go public.

Anthropic, the makers of Claude. OpenAI, the makers of ChatGPT. And SpaceX, the maker of rocket ships and satellites and AI and whatever else Elon Musk has welded into his unholy alliance this quarter.

Here’s the first thing you need to know about these three companies: they lose money. Not a little money—massive, historic, oceanic amounts of money.

They are on the spending spree of the century, pouring billions into data centers the size of small towns and, in SpaceX’s case, into pieces of rocket ships that usually don’t explode.

And here’s the second, which is the one I want you to sit with for this entire essay. The same political class that has spent forty years insisting there was no money for housing, or health care, or child care, found $4 trillion for this—instantly, and without anyone having to ask how we’d pay for it.

By the end of this piece, I’m going to show you that the scarcity they’ve sold us since the Reagan years was never a fact about the world. It was a choice about where the money was allowed to go.

Let me walk you through how I know that.

The Filings

Here are the confirmed facts, in filings three weeks apart.

On June 1, 2026, Anthropic confidentially filed for an IPO, with a potential October debut. Days earlier it had closed a Series H round, raising $65 billion and valuing the company at $965 billion—nearly the first $1 trillion private company ever. To feel the velocity: in March of 2025, all of Anthropic was valued at $61.5 billion. About fourteen months later, that number is sixteen times higher.

On May 22, 2026, OpenAI confidentially filed an S-1—the SEC paperwork that begins an IPO, the document where a company finally has to open its books—targeting a September listing above $1 trillion. OpenAI still projects roughly a $14 billion loss in 2026 alone, and has said out loud, in public, that it does not anticipate being profitable until 2030.

SpaceX filed in May 2026, reportedly seeking up to $75 billion in new funding at a valuation of $1.8 trillion—more than twice the largest IPO in history. SpaceX posted a $4.28 billion net loss in the first quarter of 2026, with an accumulated deficit of $41.3 billion, largely driven by its AI operations.

Three companies, within weeks of each other, each valued at or above $1 trillion. Nearly $4 trillion in combined valuation.

Now hold that against the size of the thing it’s happening inside. The entire U.S. economy is about $30 trillion. These three companies, which mostly lose money, are claiming a combined value equal to roughly one-eighth of everything the United States produces in a year—before a single share has traded publicly.

So here’s the big question: where do $4 trillion new dollars come from?

How does that much money get manifested out of thin air?

Follow the Money

You might be interested to learn that some of the players are names you’ve been hearing all over the news lately. Sovereign wealth funds, for example—and not as passive index money. As direct round participants.

• GIC and Temasek, out of Singapore, anchored Anthropic’s Series H.

• The Qatar Investment Authority backed an earlier round.

• Abu Dhabi’s MGX—a $100 billion state AI fund—is invested in OpenAI, Anthropic, xAI, and the $500 billion Stargate project.

• Saudi Arabia’s Public Investment Fund and Abu Dhabi’s Mubadala have both scaled their AI allocations hard.

Read that list again, because it changes the nature of the bet. The floor under these valuations is no longer just institutional demand. It is, in part, nation-state strategic interest. When the savings of Singapore and Qatar and Saudi Arabia are direct backers, the question stops being “will this company succeed” and becomes “will these governments let it fail.” That’s a different kind of risk, and it’s one ordinary public investors have no way to price.

But here’s what should make the hair on your neck stand up. These aren’t unfamiliar names. These are the same funds—in several cases the same people—who have spent the last two years buying their way into the orbit of the American president and his family. The money pouring into the AI buildout and the money showing up in the corruption headlines are coming out of the same accounts.

Let’s walk through it.

Saudi Arabia’s Public Investment Fund—a backer in the frontier-AI rounds—is the same PIF that put $2 billion into Jared Kushner’s investment firm, Affinity Partners, six months after he left the White House. The fund’s own screening committee recommended rejecting the deal, citing the “inexperience” of Kushner’s management and fees it called “excessive.” Crown Prince Mohammed bin Salman, who chairs the board, overruled them. By 2024, the firm had returned no profit to its investors but had collected roughly $157 million in management fees from foreign governments—$87 million from Saudi Arabia alone. The same PIF funds LIV Golf, which has paid to hold tournaments at Trump’s properties, and is now scrutinized by ethics experts who can’t even measure how much the Trump family is pocketing because there’s so little transparency.

Abu Dhabi’s MGX—the $100 billion state AI fund in OpenAI, Anthropic, and the Stargate project—is the same MGX that made a $2 billion investment in the crypto exchange Binance using USD1, the stablecoin from the Trump-family-backed World Liberty Financial. The deal was announced by Zach Witkoff, son of Trump’s Middle East envoy. It is expected to generate hundreds of millions of dollars for the Trump family. Senators Warren and Merkley have demanded the records. MGX is steered by Sheikh Tahnoon bin Zayed—the UAE’s national-security chief, the so-called “Spy Sheikh”—whose interests separately bought a $500 million stake in that same Trump crypto firm, $187 million of which flowed to Trump family entities, in the same window the administration agreed to ship the UAE up to 500,000 of the most advanced AI chips a year.

Qatar—whose Investment Authority backed an earlier AI round—is the same Qatar that offered the Trump administration a $400 million Boeing 747, a “flying palace” destined, after his term, for his presidential library. Legal scholars across the spectrum called it a likely violation of the Constitution’s emoluments clause. It was waved through anyway.

These are not separate stories.

The sovereign wealth of the Gulf is the connective tissue running through all of it—the jet, the crypto coin, the golf tournaments, the son-in-law’s fund, the AI cap table.

The same predator class, moving the same money, buying the same access, now writing the checks that prop up the most valuable companies on Earth.

I’m not alleging a single conspiracy with one set of instructions. I don’t have to. The pattern indicts itself: this is what it looks like when the floor under American capital is foreign state money, and the door that money walks through runs straight past the people who are supposed to be guarding it.

So when I tell you the bet is “geopolitical, not just financial,” understand what that means in plain English. The people backing these valuations have already demonstrated, in public, on the record, exactly what they expect their money to buy.

Then there’s the circular financing—a situation where suppliers are funding their own customers.

Nvidia, headed by Jensen Huang, makes the GPUs that are the brain and the heartbeat of the massive data centers being built for AI. Nvidia invests in OpenAI. OpenAI commits hundreds of billions of dollars to cloud providers. Those cloud providers then turn around and buy Nvidia chips to fill the data centers running the AI models these frontier labs are building. The money goes out one door and comes back in another.

Analyses in 2026 estimate that $800 billion in these arrangements exist. OpenAI’s infrastructure commitments alone total about $1.15 trillion across seven vendors. And this isn’t a fringe concern—Morgan Stanley projects roughly $3 trillion in global data-center spending between 2025 and 2028, with only about half of it funded by actual cash flow. The rest is debt.

If you’re hearing an echo of the dot-com bust of the early 2000s, you’re not wrong. Back then, companies bought each other’s services to inflate the appearance of growth. The exact same thing is happening now. And this isn’t just a vibe—it’s measurable. The Shiller price-to-earnings ratio, the classic gauge of how stretched the stock market is, topped 40 in May 2026. That is higher than the peak of the dot-com bubble. Higher than almost any reading in history except the final euphoric weeks before the crash of 2000. The market is not merely rhyming with 1999. By the measure economists actually use, it is past it.

And here’s the tell that even the players know how this looks: in late 2025, Nvidia announced it might invest up to $100 billion in OpenAI. By March of 2026, Jensen Huang told a Morgan Stanley conference that a $100 billion investment was “probably not in the cards,” and the deal was quietly settled at $30 billion. When the man writing the checks starts walking back the size of the checks right before the IPO, that’s worth noticing.

Don’t take my word for how fragile this all is. Take it from inside the industry.

“If OpenAI and Anthropic can’t make money,” William de Gale, a portfolio manager at BlueBox Asset Management, told CNBC, “this whole thing falls apart.”

Here’s a fun tidbit. There’s been a lot of noise lately that the average Jane and Joe on the street are missing out on the amazing opportunity to buy into these startups the way the big institutional investors can. So Robinhood opened private-market access to 150,000 retail investors—ordinary people, sold pre-IPO stakes at high-hundred-billion-dollar valuations, right before the lockup-and-dump risk.

Imagine: your hard-earned 401(k) dollars can now turn you into a potential millionaire—if only you’re brave enough to pour them into the next round of the technology bubble, at the top, helping to cash out people who were already rich to begin with.

The Vanishing Public Market

Here’s another set of facts worth holding together:

• The public market is disappearing. Public companies are down to about 4,000—less than half the 1996 peak.

• Billion-dollar private companies have gone from 280 in 2017 to roughly 1,582 in 2025, a 465% increase.

• Companies now stay private for about sixteen years on average.

• The IPO has stopped being how a company raises money and become a liquidity exit—a way for the early money to cash out.

Stack those up and you see what’s actually happened: the entire high-growth phase of a company’s life now happens in private, where ordinary people can’t invest and regulators can’t see. By the time a company “goes public,” the real gains have already been captured, in the dark, by the people who were let in early.

And that darkness is the whole game. Every AI valuation since 2023 has been a private assertion—unaudited, marked up round after round, each round citing the last as “price discovery.” (Translation: they guessed at a price, and then—oops—discovered the price was even higher than the last time they guessed.)

These IPO filings will be the first time audited GAAP financials exist for a frontier AI lab. That document will be one of the most closely watched in financial history, because it does one of two things. It either anchors the entire $4 trillion valuation stack—or it exposes how much of that stack was circular validation all along. We are, for the first time, about to find out in public whether the most valuable companies in history actually make money.

And here’s the part that should keep you on the edge of your seat: we haven’t even seen it yet. Both OpenAI and Anthropic filed confidentially—which companies are allowed to do, letting the SEC review the books in private first. The full, public S-1, the one the entire financial world is waiting to read, drops only weeks before the stock actually lists. So the most revealing financial document of the decade is sitting in a regulator’s inbox right now, and the rest of us don’t get to see it until the people who do have already decided what they think. Won’t that be fun.

There Was Never No Money

So here we are, in the miracle year of 2026, where three companies are worth up to $4 trillion privately, right now, before a public share has ever traded.

You may ask yourself: where does $4 trillion come from?

You may ask yourself: how did we get here?

For forty years, that was the question we were never allowed to win. How do we get universal health care? How will you pay for it. How do we end homelessness? How will you pay for it. How do we make sure every child has child care, that student debt gets relieved, that every kid gets three square meals a day? How will you pay for it.

We even ran the experiment and watched it work. In 2021, the expanded Child Tax Credit drove child poverty to a record low of 5.2%—and then we let it expire, and child poverty shot right back up, and in 2024 the Senate blocked an attempt to bring it back. The money to keep millions of children out of poverty was right there. We decided not to spend it.

And now, in the last eighteen months, $4 trillion has materialized for compute. No vote. No debate. No “how will you pay for it.”

The scarcity we were sold for forty years was never a fact about the world. It was a decision about who deserves the money and who does not. Letting the days go by. Same as it ever was.

To put a number on what we say we can’t afford: the National Alliance to End Homelessness estimates it would cost about $9.6 billion to provide permanent housing for every household that stayed in an emergency shelter in a year. That’s not $4 trillion. That’s roughly two-tenths of one percent of it.

We found four hundred times that much for three companies that lose money, and we found it in a year and a half.

In an earlier essay in this series, I wrote about how the massive spending on data-center infrastructure—$650 billion in 2026 alone—is propping up the rest of the U.S. economy, which without it would be sitting at zero percent growth. That same spending, going into data centers for companies that are currently losing money, is also propping up the trillion-dollar valuations of those same companies. The capex justifies the valuation; the valuation justifies the capex. I’m not saying the predator class is running a game of three-card monte. But man, it sure looks like it, doesn’t it?

The logical question is: why isn’t this money available for other social goods? And the answer, now and always, is that money flows where power points it—at the highest private return, never the highest human one.

I’ve said it before and I’ll say it again: the greatest harm ever done to capitalism in the name of capitalism was Milton Friedman’s. In 1970, in the pages of The New York Times Magazine, he declared that “there is one and only one social responsibility of business—to use its resources and engage in activities designed to increase its profits.”

One. And only one.

Not to employ people well. Not to leave the country better than it found it.

For five decades, corporations have used Friedman’s sentence as a mandate—the highest private return, and nothing else—and that mandate has now metastasized to the scale of the entire global capital market. The money for human goods was always there. Generations of bankers and politicians simply decided never to point it at humans.

What We Can’t Un-Know

If some part of you is screaming we have to stop these companies from listing—let me deflate that balloon. They will list. Whether they’re truly worth a combined $4 trillion is an open question; some of that value is real, and some of it is almost certainly a bubble. We are all about to learn which is which, in what I suspect will be the most expensive lesson in financial history.

But that’s not the thing to take from this.

The thing we cannot un-know is that when global capital wanted the money for this specific purpose, the money simply appeared. Out of thin air. In eighteen months. Without a vote.

Which means every single time, from here forward, that someone tells you we can’t afford to house the homeless, or feed the children, or care for the sick—you now hold the receipt.

The money was always there.

They just told us it wasn’t.

Cliff’s Note: You just read the kind of piece the big outlets won’t run — not because it isn’t true, but because the people it’s about buy the ads, fund the panels, and golf with the publishers. Lawrence followed the money, named the names, and showed you the receipt: $4 trillion for them, “how will you pay for it” for you. That reporting takes time, and time is the one thing nobody at Blue Amp is billing to a sovereign wealth fund.

So here’s my ask, and you know I only make it when it matters.

Subscribe if you haven’t—free gets you in the fight, paid keeps the lights on.

Upgrade or donate if you can; we’re funded by readers, not by anyone with a yacht and an agenda, and that’s exactly why we can publish pieces like this one.

Share it—send it to the friend who keeps saying “there’s no money for that,” because now you’ve got the receipt to hand them.

And get in the comments and tell us: what were YOU told we couldn’t afford? Lawrence reads every one. So do I.

Discussion about this post

Ready for more?