💥 Zero Bullshit
Chapitre 03

Delusions of Grandeur

The marketing crowd had their “disruptive” phase 15 years ago. Finance had the golden age of blockchain, then NFTs, tokenisation, real estate crowdfunding, equity crowdfunding, structured products. Every time, the market turns, retail tells itself it was just bad luck this time, and the pros know full well they never sold the products the rich, the bankers and the lizard people actually buy. Once the last Stellantis structured product had finished losing the last billion, a new magic formula had to be found to hoover up fees. Say hello to my little friend late stage.

All secondary transactions combined, including portfolio sales between funds, global volumes reached $200bn to $240bn in 2025, up about 50% in a year, and are expected to pass $250bn this year5. How much of that is buying shares in a company rather than stakes in funds? No public breakdown exists.

💡 Three words, three different things

Late stage describes a stage in a company’s life, not a transaction: a firm that is already mature, valued in the billions, raising big rounds. A late-stage investment can be primary or secondary.

Primary: the company issues new shares and takes the money. The saver is funding the business.

Secondary: existing shares change hands. The money goes to the seller. The company gets nothing, and doesn’t always have a say.

Pre-IPO means nothing in law. It is a sales line, which assumes a stock market listing whose date and price nobody guarantees.

One last confusion, the most convenient one: in private equity jargon, a secondary fund buys stakes in funds from other investors. That has nothing to do with buying shares in a company. Yet the 2 are sold under the same word.

The retail promise is simple: invest in a big scale-up about to go public, and cash in once the bell has rung. The sales pitch is even simpler: hugely famous companies that no longer need marketing, a risk close to zero since everyone has already invested, and an IPO just around the corner.

The details are a lot less simple: if it’s the deal of the century, why would someone who invested 2 rounds ago want out now, going through 4, 5, 6 or even 7 middlemen?

It’s all in the nuance that marketing swallows.

These opportunities exist, and they are not heresy in themselves. Early investors in SpaceX or Uber waited more than 10 years before seeing a whiff of an IPO, with the risk that everything would collapse in the meantime. Getting out calmly, on sound terms, because a need for cash comes up, is nothing crazy.

What is a bit crazier is the fees. Because they are, by and large, unreadable.

Back to SpaceX. The deal had been announced in Hakim Benotmane’s club (in French). Surprisingly, it was one of the announcements that earned him the most accusations of making things up... when it was perfectly credible. And, above all, true: he had indeed secured an allocation. What remained was to understand the terms.

The seller, a professional Zero Bullshit was able to speak to but who declined to go on the record, had come in in 2023 in the Series J, $1.9bn raised at a valuation of about $48bn. Not directly on the cap table: through an SPV, which, incidentally, was not the one run by a16z, the round’s lead.

So what exactly is Hakim selling his members? Well, the seller puts his shares into another SPV, belonging to a platform that pools secondary transactions, which itself has a middleman in Europe, which sold the deal to Hakim, who sells it on to his members, via a PSI, a licensed investment firm, in France, which should have been Equisafe. The number of middlemen on this deal is close to the number of real unicorns left in France: 5.

As for who takes how much... let’s just say it is easier to count the founders who actually made money after an exit than to work out the real fees. Partly because of the regulation itself, which defined very precisely what a fee is when MiFID II came in. Roughly, and lawyers are welcome to set me straight in the comments: fees are what the buyer pays the seller. In the SpaceX case, those of Benotmane and of whichever PSI organised the round table. Bearing in mind that access to the club is already paid for by subscription. Another reading would have been the gap between the valuation of the seller’s shares (before the middlemen) and the buyer’s.

According to our information, the seller was letting his shares go at $185, slightly below the secondary market of summer 2025, which was then trading at around $200-210 a share6, a valuation in the region of $400bn. In December 2025, the internal share buyback would value the company at $800bn, at $421 a share7. A 2x in 6 months. Figures, every one of them, completely inconsistent with those in Benotmane’s sales documents, which showed a carry of 50% of profits for the SPV and projected valuations of €300bn within 3 years and €1,000bn within 10 years, whereas his own sale price of $210 a share implied, depending on the number of shares used, a valuation already close to $900bn.

The SpaceX sheet in Hakim Benotmane’s sales documents.

So, did the club members lose out? Nine months later, the maths looks easy. IPO priced at $135 on 12 June 20268. Bought at 210, listed at 135, and even 112 at the end of July. -46%. Cheers, Hakim.

Except that sum is wrong. You have to do it the other way round.

In the meantime, in May 2026, SpaceX split its shares 5 for 1 to prepare for its listing, bringing the fair market value down from $526.59 to $105.329. One share bought for $210 in autumn 2025 became 5 shares. The club members’ real cost price is therefore $42 per post-split share.

Which gives, for anyone who actually bought at that price:

For $42 paid per share after the split: the multiple, by date

Then there’s the lock-up. The shares are locked for 180 days after the IPO: see you on 9 December 2026 to collect anything at all. Yet in 6 weeks, the stock has lost half its value, fell back below its IPO price in mid-July and is now dragging along its lows, around $107-112. Between the ×5.4 of 16 June and what can actually be cashed in December, the gap is already counted in halves. Admittedly, to lose money in absolute terms, the stock would have to break below $42, another -63%. But the final gain depends on a share price in 5 months that nobody controls, on a stock that has just shown it can do -50% in 6 weeks.

And then there’s where the gain came from. The club was paying $210 when the seller was getting out at $185. That is 13.5% above the price the seller accepted, before the club subscription and the layers of SPVs. What saved everyone was the repricing between autumn 2025 and the IPO: $421 a share at the December tender, then the merger with xAI, completed on 2 February 202610, which slipped Grok and the social network X into the same vehicle (plus the Colossus data centres) and took the whole thing to $1,250bn, ahead of a listing targeting up to $1,500bn.

Series J

$1.9bn raised at a valuation of about $48bn. The seller comes in, through an SPV.

The secondary market at around $200-210

A valuation in the region of $400bn.

The seller gets out at $185, the club buys at $210

13.5% above the price the seller accepted.

Internal buyback at $421 a share

The company is valued at $800bn.

Merger with xAI

Grok, X and the Colossus data centres in the same vehicle. The whole lot is worth $1,250bn.

Split by 5

The fair market value goes from $526.59 to $105.32. The club’s cost price drops to $42.

Listing at $135

The peak, $225.64

×5.4 for the club.

$112.55

Back below the IPO price since mid-July.

End of the 180-day lock-up

The first day club members can sell.

All the more so as all of this is gross. For the net performance, you would have to subtract the subscription, the middlemen’s spread, any carry, the PSI’s cut. How much exactly? Nobody knows. That is precisely the problem.

The kind of problem that sometimes ends up in court. In the United States, Late Stage Management, a New Jersey company, sold pre-IPO deals presented as free of charge, explaining that it only got paid on exit, from 20% of the profits. The DOJ11 established that it was in fact taking hidden markups of 10 to 50% on the way in: $528M raised between 2019 and 2022, $88.6M misappropriated, and guilty pleas to show for it in early 2026.

Late Stage Management, 2019 to 2022

And all this is made more complicated by the fact that not all shares, and therefore not all sales, are equal. At Alphabet, listed since 2004, 3 classes coexist: A shares (1 vote), B shares (10 votes, never listed, in the founders’ hands), C shares (0 votes)12. Same company, same profits, same rights on everything else, except the vote. In private companies, it’s worse: liquidation preferences, transfer clauses, everything is negotiated round by round. At SpaceX, the IPO prospectus filed with the SEC8 describes the same mechanism: class B shares with 10 votes each, which leave Elon Musk with about 82% of the voting rights. The free float votes for show.

The same Benotmane had, by the way, offered a little later to invest in Mistral AI. Asked several times, the company denied it, and Benotmane was branded a fantasist of epic proportions, which, let’s be honest, is so unlike him. Well, guess what: the information… was true. But in reality, what was being sold was neither Mistral shares nor even entry into some SPV: it was an investment option held by an Indian family office, palmed off by an obscure middleman.

That is what marketing calls getting access to the same deals as the big funds.