Three. That is how many companies Jensen Huang named on June 1 as the first big buyers of Nvidia's new Vera CPU: Anthropic, OpenAI, and SpaceX. It is also how many companies make up the 2026 IPO cohort that Wall Street has spent five months waiting for. The two lists are the same list.
Huang delivered the names as a flex. The Vera chip, he told Bloomberg, runs AI workloads 1.8x faster than x86, and its earliest large customers are the most demanding compute buyers on earth. The framing was about Nvidia's product. The information was about something else: who Nvidia's revenue depends on, stated by name, on the record.
The same morning, the first of those three names filed confidentially for an IPO. Anthropic submitted a draft S-1 to the SEC, joining a queue it had been circling since winter. OpenAI's CFO had already said the company would "for sure" reserve shares for retail investors. SpaceX, the third name, spent June 1 disclosing the mechanics of its own listing — reserving up to 5% of Class A shares for employees and friends and family, with extended lock-ups on 60%+ of the float.
The three companies Nvidia named as its flagship Vera customers are the same three companies preparing to sell shares to the public in 2026. The buyers of the compute are the sellers of the shares.
Compared to what
A coincidence of three names is not, on its own, a finding. The question is always: compared to what? Compared to the structure of the last two years, this is a reversal in the direction of money.
For most of the AI buildout, capital flowed one way and privately. Nvidia sold chips. Private investors — and, increasingly, Nvidia itself — funded the companies that bought them. By May, Nvidia had made $40 billion-plus in equity commitments in 2026 alone, including $30 billion into OpenAI. Reporting has documented the wave of circular deals between Nvidia and OpenAI, and traced how Nvidia became the AI industry's most powerful financier. The chipmaker was funding its own demand. The loop ran inside private markets, where no one had to print a balance sheet.
On June 1, the loop reached for a different source of money. The three biggest Vera buyers are not raising their next round from Nvidia or SoftBank or a Gulf sovereign fund. They are raising it from the public — from index funds, pensions, and retail brokerage accounts. The private circular machine is being plumbed into the public markets, and the same day Huang read out the customer list, the customers started filing the paperwork to fund the purchases.
A chipmaker that funded its own demand has run out of private money large enough to keep funding it. The next round comes from the public.
The fourth name
If the three names were the only signal, this would be a tidy pattern and not much more. But the same day produced a fourth data point that tells you the door is wider than three startups.
Alphabet announced it would raise $80 billion in equity to fund its AI spending plans — including a $10 billion investment from Berkshire Hathaway. The most disciplined value investor in the world, the firm whose entire reputation rests on not overpaying for growth narratives, put ten billion dollars into a tech giant's data-center buildout on the same day three of that buildout's biggest beneficiaries lined up to go public.
Alphabet is not short of cash. It is one of the most profitable companies in history, and it is selling equity anyway — diluting its own shareholders to pay for capacity. When a company that prints money chooses to raise more by selling stock, the message is not that it needs the money. The message is that the buildout is large enough that even Alphabet would rather not fund it entirely from the income statement. The capex has outgrown the cash flow of the richest companies on earth, and the gap is being closed in the public markets.
What the customer list is
Return to Huang's three names. A vendor naming its biggest customers is ordinary. The overlap with the IPO calendar is not, because each list reveals what the other conceals.
The customer list tells you where Nvidia's revenue is concentrated. The IPO list tells you that the same revenue base is about to be repriced by people who can see the books for the first time. SpaceX stayed private for twenty-three years precisely so it would never have to publish quarterly numbers; its first S-1 turned that silence into a public ledger, and the ledger was red. Anthropic's draft S-1 will do the same for a company that missed its own growth forecast by 8x and responded by signing compute contracts with every supplier that had capacity. The companies buying the most Nvidia silicon are about to become the companies whose every assumption is subject to a prospectus.
This is the structural fact underneath the coincidence. For two years, the AI buildout was financed by people who chose to believe — private investors pricing a narrative, a chipmaker investing in its own customers, a secondary market trading slices of companies that disclosed nothing. The public markets work differently. They price the disclosure, not the narrative. And the disclosure, in three separate prospectuses, will all point back to the same vendor that just read their names aloud.
The counter
The obvious objection: an IPO is not a sign of stress. It is a sign of strength. Companies go public because they can, because demand is overwhelming, because the bankers expect Anthropic alone to raise more than $60 billion. Going public is what winners do. The three names are on Huang's list because they are the best customers, and on the IPO calendar because they are the best companies. Two strong lists overlapping is not a reversal. It is just what dominance looks like.
The objection has weight, and the timeline supports it — this cohort was flagged as the "year of the mega IPO" back in January, long before anyone called it stress. But strength and structural dependence are not mutually exclusive, and the overlap is the point regardless of which reading you prefer. Whether these companies are going public from strength or from the need to fund a buildout they cannot finance privately, the result is the same: the public is being invited to underwrite Nvidia's customer base directly. The retail shares OpenAI's CFO promised, the friends-and-family allocation in SpaceX's filing, the $10 billion of Berkshire's capital — all of it ends up, by a short chain of contracts, paying for chips. The customer list and the shareholder list are converging.
Three, recontextualized
Three names is a small number. On June 1 it did a large amount of work. It was the count of Vera's flagship customers and the count of the 2026 IPO cohort, and those two counts were the same count because they were the same companies. Huang named his biggest buyers in the morning; by the afternoon, those buyers were filing to sell shares that would, in the end, pay him back.
For two years the question about AI capex was who would fund it. The private answer was: Nvidia, the Gulf, the venture funds, the chipmaker recycling its own revenue. On June 1 a different answer started filing paperwork. The public will fund it — through index funds that hold Alphabet, through retail accounts that will hold OpenAI, through Berkshire's ten billion dollars. The loop that ran quietly inside private markets has found the door to the largest pool of capital there is. Three names walked through it on the same day. The rest of the cohort is in the hallway.