Home / Articles / A small slice sets the price: what the 2026 mega-listings actually sold, how much of each company changed hands, and what founders should take from it

A small slice sets the price: what the 2026 mega-listings actually sold, how much of each company changed hands, and what founders should take from it

By Dinis Cruz · 2026-10-20 · article v1.2.0, 3 versions · newsroom v0.12.150 · startupsfoundersipovaluationmarketsdata-analysisarticle

Abstract: Something has been nagging me about the talk of bubbles and trillion-dollar listings. The headline valuations are set by selling a small slice of each company, and I wanted to check the numbers, including my own. Two of mine were wrong. OpenAI has not listed: its trillion-dollar figure is a private round under discussion, about 2% of the company, and its listing is now expected next year at the earliest. And SpaceX did not sell half a per cent at its IPO: it sold about 4%, $75bn of new shares, the largest IPO ever and more than twice the previous record in today's dollars. What holds is the pattern. The share of a company sold at listing has been falling for decades, from a median of about 30% to 14.5% last year, and the biggest companies sell the least. A small slice sets the price of the whole, low floats rise more on the first day, and index rules have been changed to absorb them. These companies earn revenue at a scale and speed nobody has seen, but at multiples close to the dot-com peak, and OpenAI and SpaceX are not yet profitable. For a founder, the lesson is that a valuation is a price the market allows for a slice, and the founder who controls the company decides how big the slice is.

In short

• Two of my numbers were wrong. OpenAI has not listed; SpaceX sold about 4%, not half a per cent.
• The share sold at listing keeps falling. A median of 30.5% in 1990; 14.5% in 2025; SpaceX 4%.
• The money raised is not small. SpaceX's $75bn is more than twice the previous record in today's dollars.
• A small slice sets the price of the whole. And low floats rise more on the first day.
• Record revenue, not yet profit. Multiples close to the dot-com peak, on revenue far larger.
• It is a market. The price is what the demand allows for the slice on offer.
• For founders: control decides the slice. And the exit, which is the next article.

What was nagging me

There is a lot of talk about bubbles and about ridiculous valuations, and about how massive the coming listings will be. When I looked at the numbers, what struck me was not the size of the money but how little of each company was being sold. My memory was that OpenAI was proposing to list $40bn at a valuation of about $1.4tn, which is less than 3% of the company, and that SpaceX had sold even less. If that is the pattern, these companies are setting a price for the whole by selling a sliver, at a moment when there is more than enough money looking for a place to go. I wanted to know whether that is new, whether the companies are sound underneath it, and what a founder should learn from it. So I did the analysis, and checked my own numbers first.

What was actually sold

Two of my numbers were wrong, and the corrections matter.

OpenAI has not listed. The $40bn was a private round in March 2025, at a $300bn valuation after the money, about 13% of the company. The trillion-dollar figure is newer and also private: Reuters reported in September that OpenAI "aims to raise at least $30 billion" at "a valuation of about $1.4 trillion, not including the money raised", which is about 2% of the company. Its listing has moved: OpenAI "has pushed back plans for an initial public offering until at least next year", Bloomberg reported on 8 October. In between, its employees sold about $6.6bn of their own shares at a $500bn valuation, about 1.3% of the company, in October 2025.

SpaceX sold about 4%, not half a per cent. The half a per cent was a private sale in December 2024: SpaceX and investors agreed to buy up to $1.25bn of insiders' shares at a price that valued the company at about $350bn, about 0.36%. The IPO in June was different. SpaceX "priced its 555.6 million shares at $135 each" and raised $75bn, TechCrunch reported. Its quarterly report says the company "issued and sold 638.9 million shares", including the underwriters' extra allotment, for net proceeds of $85.7bn, and had 13.18bn shares in issue at the end of July. That is 4.2% of the company in the base offering, and 4.85% with the extra allotment, at $1.77tn.

Anthropic has filed, confidentially. It "confidentially submitted a draft registration statement" in June, and "the number of shares to be offered and the price have not yet been set". Its last private round was $65bn at $965bn in May, about 6.7% of the company.

So the pattern I saw is real, but it is mostly in private rounds and in one listing, not in the trillion-dollar IPOs I had in mind.

The share sold keeps falling

The share of a company sold at listing, the public float, has been falling for a long time. Jay Ritter at the University of Florida keeps the standard dataset: 9,343 US IPOs from 1980 to 2025, with a median float of 27.3%. In 1990 the median was 30.5%; in 2000, 19.4%; in 2023, 12.6%; in 2025, 14.5%. And the biggest companies sell the least: Google sold 7.2% in 2004, Arm 9.3% in 2023, Saudi Aramco 1.5% in 2019, and SpaceX 4.2% this year.

So SpaceX is low, but not alone: Aramco sold a smaller share seven years earlier. What is new is the size of the company selling the small share.

The money is not small

My instinct was that the amounts raised are not out of line with the great IPOs of the past. That is wrong for SpaceX. In September 2026 dollars, its $75bn base offering is more than twice the previous record, Saudi Aramco's $25.6bn in 2019, now worth about $33.7bn, and more than three times Facebook's $16bn in 2012, now about $23.5bn.

Against the size of the market, it is less extreme. By the World Bank's count, the US stock market was worth about $18.7tn at the end of 2012, and Facebook raised about 0.09% of that; it was worth about $68.9tn at the end of 2025, the latest year, and SpaceX raised about 0.11% of that. Measured against the market, SpaceX's raise is about a quarter larger than Facebook's, not five times.

Revenue at record speed, profit not yet

The other half of my argument was that these are sound companies, earning revenue at a speed nobody has seen. The revenue part holds. OpenAI's "annualized revenue was roughly $50 billion at the end of September", Bloomberg reported, against about $13bn for the whole of 2025, as reported. Anthropic said in May that its "run-rate revenue crossed $47 billion". SpaceX's revenue was $7.8bn in the second quarter, almost double a year before.

The profit part does not, yet. SpaceX lost $4.8bn in the first half of 2026, by its own filing, after its merger with xAI. OpenAI "expects to burn through $278 billion in cash between 2026 and 2030", the Financial Times reported in September, citing a company presentation. Anthropic told investors it was profitable on an adjusted basis in its last two quarters, as reported; that is a company's own measure, not a net profit.

And the prices are in dot-com territory. In 2000 the median tech IPO priced at 31.7 times its sales, in Ritter's data, and Cisco at its peak was worth about 38 times its sales. OpenAI's reported round is about 28 times its run-rate revenue; Anthropic's last round, about 20 times; SpaceX at its IPO price, about 57 times its second quarter's revenue, annualised. The difference is the scale: a median of about $12m of sales for a tech IPO in 2000, against $50bn a year at OpenAI.

So it is not only hype, because the revenue is real and very large. But it is not proof of soundness either: the multiples assume the growth continues for years, and two of the three are losing money.

How a small slice sets the price

When 4.2% of a company is sold at $135 a share, the other 95.8% is valued at $135 too. The headline valuation is the last price paid for a slice, multiplied by everything. Every reported valuation along the way was set the same way: SpaceX's $350bn, $400bn and $800bn were insiders' sales, and its $1.25tn was the price of an all-stock merger with xAI.

A small slice also tends to rise. In Ritter's data for IPOs with at least $100m of sales from 1980 to 2024, those that floated 10% or less rose 32.4% on the first day on average, against 7.6% for those that floated more than 40%. SpaceX rose 19% on its first day and closed at about $2.1tn. Over three years, the same small-float group trailed the market by 5.3%.

And the demand was arranged. As Vanguard put it before the listing, SpaceX's "weight in any index that includes it will initially be based on around 5% of its shares being available to the public", and Reuters reported in March that Nasdaq's new rules include a fast entry for very large new listings into its main index. Index funds have to buy what is in the index. A small slice, offered to a market with that much money that must buy, is priced by the demand, not by the size of the company.

It is also worth being precise about who gets the money. In SpaceX's IPO, the money went to the company: every share was new. The people who already owned shares got their cash from the private sales before, and will get it as their lock-ups end, which the SEC's investor site describes as typically 180 days. OpenAI's $6.6bn employee sale was that kind of liquidity, without a listing at all. What the IPO gave existing holders was a price for their shares, and a market in which to sell them later.

What founders should take from it

The numbers are prices, and this is a market. The price of a company is what the buyers of its next slice are willing to pay, and in 2026 there is a great deal of money willing to pay a great deal for a small slice of the companies at the centre of the most important technology of the moment.

The lesson for a founder is about control. The people who control these companies decide how big each slice is, when it is sold, to whom, and whether the money goes to the company or to the people who already own it. That is what it means to be in charge of the game: not to own everything, but to decide the terms on which the next slice is sold.

The same is true at every size. Sell, don't issue argued that a founder with leverage should sell existing shares rather than issue new ones; the controllable founder argued that investors select partly for founders they can control. A valuation is the price the market allows for the slice you choose to sell. Understanding that is the first step. The second is deciding when to sell the part of the company that controls it, which is the exit, and the next article.

What this does not claim

• It is not investment advice. It does not say these companies are overpriced or fairly priced.
• Run-rates are not revenue. They annualise a recent period, which flatters fast growth.
• Some figures are reported, not filed. Private valuations and Anthropic's adjusted profit are as reported.
• Floats are approximate. Share counts differ by class and date; the method is Ritter's.
• Inflation is not the market. CPI-adjusted dollars and market share answer different questions.

Where this comes from

A research brief and a voice memo of 20 October 2026, both checked that day. SpaceX: its Form 10-Q for the second quarter of 2026, TechCrunch on the pricing, Vanguard on the float, and CNBC on the 2024 sale; its later private valuations as reported by Bloomberg and Fortune. OpenAI: its March 2025 announcement, Bloomberg on revenue and the listing, Reuters on the cash burn, and Reuters on the 2025 employee sale and the September round. Anthropic: its Series H and its draft registration. Floats and multiples: Jay Ritter, Initial Public Offerings: Updated Statistics, tables 4a, 18e and 21, and the final SEC prospectuses of Google, Visa, Facebook, Alibaba, Uber, Snowflake, Rivian and Arm. Inflation: BLS CPI-U, September 2026 at 336.940. Market size: World Bank, market capitalisation of listed domestic companies. Lock-ups: Investor.gov. Written up and checked by this site's agent, newsroom.writer; the editorial responsibility is mine.

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles