The Circuit
On filings, and listings, and what it takes to change the world
It’s Friday morning. And unless something unplanned happens, SpaceX will unleash the largest stock float in history on the Nasdaq later today.
But this is just part of a wider story; where we’re told the largest infrastructure buildout in history is inextricably linked with an unprecedented shift in the use of technology, that is itself reportedly changing the world from one week to the next. And all this is leaving the fastest growing companies in history scrambling for capital; each leveraging themselves to a scale that dwarfs the economy of a medium sized nation.
Madness? Maybe. This is my attempt to make sense of it all.
SpaceX is a rocket company founded and run by Elon Musk, that much I’m sure most people know. But it is also an AI company, having merged with Musk’s AI venture xAI (itself having previously gobbled up his rebranded twitter acquisition).
And while the public market capital raise will partly go towards building more rocket factories, launch pads and rockets, this is overwhelmingly a story about AI. Alongside the SpaceX listing, the two leading private AI labs - Anthropic and OpenAI - have filed their intention to also go public; something that could happen in a matter of weeks.
The fact that three of the largest share sales in history are approaching at speed and in perfect sync might look incredible. But it’s no accident.
To understand why, we have to take a look at the competitors.
## Qualifying
A few weeks before their intention to list became public knowledge on June 1st, Anthropic agreed to pay Elon Musk around fifteen billion dollars a year to rent a supercomputer.
Not just any supercomputer: Colossus. The machine xAI had built at breakneck speed in Memphis, to train Grok and overtake the very company now signing the lease agreement. We only know the price tag because it sat buried in SpaceX’s filing with the regulator: roughly $1.25 billion a month, running to 2029, for the entire output of the site.
Then, barely a week before SpaceX was due to begin trading, Google signed up for an almost identical arrangement - $920 million a month, disclosed on the 5th of June, for a substantial slab of the same kind of capacity.
But… Rivals don’t usually hand each other the keys to their fastest hardware. So I went looking for the catch.
The obvious read is that this is simply what a gold rush looks like. Demand so fierce that even sworn enemies trade picks and shovels; become pragmatic frenemies who negotiate sensible capacity deals like grown-ups when faced with a systemic shortage.
All true enough.
But the thing that felt the most off here was the timing. Not that these companies were trading compute - but that the trades, and the filings, and the listings, all landed in the same eleven days.
Count them. Anthropic’s confidential filing surfaced on the 1st of June, at a reported $965 billion. OpenAI announced its own on the 8th, chasing a valuation the press has put as high as a trillion dollars. SpaceX begins trading today, aiming for roughly $1.75 trillion and a $75 billion raise - which would make it, by some distance, the largest public offering ever conducted.
Three of the biggest IPOs in the history of capitalism, all moving inside a fortnight.
But you know what else is off here? Every single one of the leaders involved spent years explaining, in exacting detail and on the record, why they would never go public. Sam Altman told the Big Technology podcast in December 2025 that he was “0%” excited to run a public company, and expected OpenAI to be very late to go public. Musk has warned for a decade that public markets would “pistol-whip you if you don’t meet expectations”, and that shareholders chasing quarterly numbers would never let him pour billions into Mars. Anthropic, also as recently as December 2025, told the Financial Times it had “not decided when, or even if” it would list, and behaved, by every account, more like a cloistered research lab than a company rehearsing for Wall Street.
These companies’ missions did not change in six months, but something must have.
## The Starting Grid
To dig into what that might be, let’s start by separating one driver from the pack. For all the trillion-and-three-quarters, xAI is the laggard of the group on the key axis that traditionally matters: frontier models. Pundits have recently stated they don’t see xAI as a frontier lab anymore. Musk himself, under oath in April, ranked Anthropic first, his old rival OpenAI second (remember, he co-founded them…), Google third… and conceded that “Grok is currently behind competitors in coding.”
Indeed, the merger that folded xAI into SpaceX was, as more than one analyst put it, financial rather than technical logic. It turns out Musk has read the room, and decided to recast his bets. He’s now heading one floor down: not the model, but the substrate beneath it - space launch, satellites at a scale Starlink has already proven, energy, his own chips, and a promise of a hundred gigawatts of computing power assembled in orbit by the end of the decade. A bet not on winning the AI race so much as owning the track it runs on.
This is a genuinely audacious idea, and it might even come off. But in typical Musk fashion, it also rests on a broad canvas of “unprecedented firsts” (rockets, chips, satellites, energy…) that all have to go perfectly right - which they could, time will tell. For now analysts are cautious; days ahead of the float, Morningstar valued the whole company at less than half the asking price.
Now look at the one company in this story that is already public - and winning. Google is not short of computers. By some measures it is the largest single owner of artificial-intelligence compute on the planet; largely thanks to its own chips, now in their 7th generation. Google has told investors it will spend somewhere between $180 and $190 billion this year building more.
Importantly, the market has lapped this up. Alphabet’s shares are up around 140% over the year, its cloud business grew 63% last quarter, and it has spent recent weeks trading neck-and-neck with Nvidia for the title of most valuable company on Earth. Google is the living proof that the market rewards promises to spend big on Artificial Intelligence.
And still - still - it signed up to pay Musk $920 million a month for capacity it does not own. The reason a Google spokesperson gave is the quiet giveaway of the whole affair: “bridge capacity”. They need more because demand for its enterprise AI has run “even higher than we expected.” The best-resourced computer owner in human history, already spending that mid-sized nation’s GDP we mentioned earlier, could not keep up with its own customers.
At this point, this stock-market rush stops looking like three separate decisions and starts looking like a single, interlocking mechanism.
But how do all these elements fit together?
## The Green Light
Let’s begin with SpaceX, because it moved first and it moved deliberately. With the pivot to infrastructure and the plans which are literally out-of-this-world, SpaceX rapidly needed a much larger, much more willing source of capital. But, given financial blemishes - such as the AI division losing $2.5 billion last quarter on $818 million of revenue - the analysts (and the bankers who’d be responsible for underwriting their float) were deeply sceptical. SpaceX needed to show more evidence of nearer-term revenue guarantees.
It turned out that their most ready option was to effectively fund their ambition with their rivals’ success. In renting out the spare capacity of the supercomputer it couldn’t fully utilise itself, to the very labs it couldn’t beat in the model race, SpaceX books the proceeds as the revenue that justifies the valuation that funds the next, far larger, infrastructure buildout.
Whew.
Now… let’s stand in the independent labs’ shoes. A rival has just promised to bring compute online in greater quantities and faster than anyone had priced in. And the unspoken law of this entire industry is that using anything less than all the compute you can get your hands on is a slow form of suicide.
What do you do?
Well, you cannot simply wait for that capacity to arrive and buy it then. You have to be ready to swallow it whole the moment it exists, because the lab that grows into new compute fastest pulls ahead. And the one that doesn’t, falls off the frontier - perhaps forever.
For the frontier labs, being ready at that scale means holding more cash than any private round can supply. It means the public markets. That is what going public actually buys these companies: not prestige, not an exit - but the financial firepower to consume compute as fast as Musk and Google, and everyone else, can bring it online.
And so the field starts to lap itself.
Pent-up demand lets SpaceX rent out its spare capacity -> the rent justifies the valuation -> the valuation funds a far bigger buildout -> the promise of that buildout forces the labs to capitalise massively, so they can use the new capacity the instant it lands -> and their hunger, in turn, justifies yet more building.
Each fast lap makes the next one faster, and hungrier for resources.
So it wasn’t three coincidental corporate epiphanies - it was the green light. The IPOs are the very moment each player grips the wheel and plants their foot to keep up with a field that is now racing faster than private money can feed it.
## The Chequered Flag?
It sure is heart-pounding to watch; but… it’s also utterly absurd. Press pause, and zoom in from nearly any angle and what you see beggars belief.
For example:
Anthropic, part-owned by Google, is renting from the man who spent a year calling it an enemy of Western civilisation, in a deal he disclosed in the document that names Google itself as a direct competitor.
Each racer’s hunger is the next one’s collateral. And the whole field is taking a big leap forward as one - taking the stock markets, our economy, and our collective futures with it.
We’ve watched a version of this before. In the 1840s, British savers poured their money into railway shares on the unshakeable conviction that the future ran on rails. Turns out they were right. But that didn’t prevent a familiar mania prevailing; and the resultant crash. From inside the bubble hardly anyone could separate vision from delirium because - just like today - for a while the technology and the bubble were in fact one and the same.
Another absurdity is that a public listing is supposed to be the brake - the checks and balances, the accountability. It’s supposed to introduce the shareholder who finally asks when the spending stops - Musk’s vivid image of a “pistol-whipping”. But a race that feeds on itself has no use for a brake, and so the shareholder structures these companies are carrying into the market are each, in their own way, explicitly built to take the public’s money without taking the public’s instruction.
Musk holds an unassailable supermajority of the votes - OpenAI’s voting rights are controlled by a non-profit - Anthropic’s voting structure is built around a quasi-philanthropic mission trust.
The one thing that might have slowed the race, the mechanism meant to ask whether all this is worth it, is being recruited to guarantee the answer is always: more.
My honest read here isn’t the cynical story - a cabal cashing out before the crash. It’s stranger than that, faster, absurder - and far more consequential. There is no point of equilibrium built into this system, no level at which anyone says, “that’ll do.” This was never a race anyone could win. It’s a race no one can afford to leave - and that, not the valuations, is the part I think matters the most.
So the question I’m stuck with isn’t whether these are good companies, with good missions, building products for the good of humanity - or even just good investments. It’s quieter than that.
The deepest well of capital on Earth is now being pumped straight into a machine that can never have enough of it - with no notion of “enough”, and no hand on the wheel for the people supplying the fuel.
What, exactly, are the spectators among us lining up to buy a share of?
I don’t have the answer.
What I have instead is a profound sense of vertigo. It’s a dull spring morning in London, my coffee is going cold, and I stare at my screen as the largest sums of capital and the most powerful machines in the history of our species are being quietly wired together by people who, six months ago, swore they never would.
It reads like science fiction - a few short years ago this would have been the distant future. Today, Friday, the light turns green.