We saw the shape of this coming.
Not the dates. Not the exact valuations. Not the specific sequence of confidential filings and carefully managed leaks. But the inevitability of it — the moment when the AI investment thesis outgrew private capital and had to go looking for a bigger wallet.
That moment is now. And it arrived not as a trickle but as a flood.
On 1 June 2026, Anthropic quietly submitted a draft S-1 to the US Securities and Exchange Commission. No press release. No leak. Just a confidential filing, clean and disciplined, targeting a valuation of $965 billion — nearly a trillion dollars — on the back of a $65 billion Series H funding round it had closed in the preceding days without making a sound. The largest labs in the world had just raised the largest private funding round in history, and almost nobody noticed until it was done.
One week later, OpenAI followed. A confidential filing of its own, targeting a $1 trillion valuation, with Sam Altman offering a characteristically disarming explanation for the timing: "We expect it to leak so we're just announcing it." The rival that Anthropic's founders left to build something safer had arrived at the same destination by a different road.
This Friday, Elon Musk's SpaceX lists directly on the Nasdaq at a fixed price of $135 per share, chasing a $1.77 trillion valuation in what would be the largest traditional IPO raise in history — $75 billion in a single transaction.
Three companies. Three weeks. Combined target valuation: $3.7 trillion.
To put that in context: the entire GDP of Germany is roughly $4.5 trillion. The public markets are being asked to absorb, price, and digest nearly that much in AI and space infrastructure in the time it takes most quarterly reports to clear legal review.
SpaceX is not a distraction from the AI story. It is the same story in a different atmosphere — frontier manifest destiny sold as infrastructure, designed to route around terrestrial economic accountability. The float class of June 2026 shares a philosophy as much as a calendar.
None of these companies are profitable. OpenAI has burned through capital at a rate that would have ended most businesses in their first year. Anthropic, for all its safety-first positioning and enterprise traction, has never turned a profit either. SpaceX is the most operationally mature of the three, but its valuation is built substantially on the promise of Starlink and Mars, not on current cash flows.
What they are selling — what they have always been selling — is a specific version of the future. A future in which AI transforms every sector of the global economy, in which the companies that control the frontier models control the infrastructure of civilisation, in which the capital deployed now will look absurdly cheap in retrospect.
It is a compelling story. It might even be true.
But "might even be true" is doing an enormous amount of work at $3.7 trillion combined.
The dot-com bubble had the same structure. The underlying technology was real. The transformation was real — eventually. But the valuations assumed the future would arrive on a specific schedule, and it didn't, and the carnage in between was severe. The companies that survived and eventually dominated were rarely the ones that attracted the peak capital. Many of the peak-capital companies don't exist anymore.
The South Sea Bubble is the other obvious reference point, and it's less flattering. Pure financial engineering dressed as transformational opportunity, collapsing under the weight of its own narrative. The difference here is the technology genuinely works. But "the technology works" and "this valuation is justified" are different claims, and conflating them has historically been expensive.
There is a detail in the Anthropic filing that deserves more attention than it is likely to receive amid the headline valuation numbers.
As part of its investor pitch, Anthropic has disclosed a $100 billion compute commitment with Amazon Web Services over the next decade.
Read that again.
A company positioning itself as the enterprise-safe, sovereignty-conscious alternative to OpenAI — the lab founded explicitly on the premise that frontier AI development needed to be done more carefully — has just anchored itself to a single hyperscaler for a hundred billion dollars over ten years.
This is not a partnership. This is a structural dependency. It means that Anthropic's ability to train models, serve customers, and fulfil its infrastructure roadmap runs through AWS for the foreseeable future. The sovereignty narrative and the $100 billion AWS commitment exist in the same filing, and the tension between them is not addressed.
For enterprise customers who chose Anthropic specifically because of its governance posture, that is a question worth asking before the roadshow.
When the S-1s become public — approximately 15 days before each company's investor roadshow, with September or October 2026 the current target window for both AI labs — retail investors will finally see the books.
What they will find is companies that have spent years consuming capital at extraordinary rates on the promise that the transformation thesis will monetise before the runway ends. The Series H closes, the S-1 files, the roadshow launches, and then the question passes from sophisticated institutional investors who understood the risk to pension funds, retail portfolios, and index trackers who may not have read the footnotes.
This is how bubbles distribute their eventual cost. Not evenly. Not to the people who made the decisions. Downward, and outward, to the people who bought in late because the story was compelling and everyone else was doing it.
That may not be what happens. The transformation thesis may land on schedule. The monetisation may follow the model projections. These companies may look cheap at these valuations in five years.
But the asymmetry is worth naming: the people taking the risk in September are not the people who will make the decisions between now and then.
Retail investors who read Telegraph headlines on the bus will be holding paper in companies whose burn rates they have never seen and whose S-1 footnotes they will never read. That is not an accusation. It is a structural description of how public markets work. And it is worth being clear-eyed about before the roadshow season opens.
Which brings us, as it always does, to the question of what this means for small, open economies watching from the edge of the frame.
The Isle of Man has spent the last several years positioning itself as an AI-friendly jurisdiction. Innovation challenges, regulatory sandboxes, careful noises about being a good place to do business in the new economy. The pitch has been consistent: we are nimble, we are well-regulated, we are adjacent to the UK but not subject to it, and we understand emerging technology.
That pitch was constructed against a backdrop of private AI investment. Venture capital, sovereign wealth funds, strategic corporate investment — the kind of capital that moves carefully, asks governance questions, and takes positions over years.
Public market capital is different. It moves faster, asks fewer questions, and when it corrects, it corrects hard. The AI economy that the Isle of Man is trying to attract investment from and position itself within is, as of this week, being repriced by retail investors who read Telegraph headlines on the bus.
The fiscal cascade argument we have been making in these pages — that AI displacement of junior professional roles will hollow out the tax base before the productivity gains accrue to anyone on the island — does not get softer when the frontier labs go public. It gets harder. The displacement pressure accelerates regardless of whether the IPO valuations hold. The jobs go either way. The tax revenue question becomes acute either way.
What a market correction would add is the removal of the growth premium. The assumption that IoM-based financial services, tech-adjacent businesses, and professional services firms would ride an AI growth wave rather than absorb an AI displacement shock would become harder to sustain.
Treasury models that have quietly factored in AI-adjacent growth should be reviewed. Not as pessimism. As an urgent intervention against wilful blindness. The Island is hunting for a growth premium that public retail markets are about to violently reprice. That is not a governance question. It is a fire drill.
Three companies. $3.7 trillion. Three weeks.
The public markets are about to find out whether the story is worth what it has been priced at. The answer will arrive slowly, then all at once, as answers always do.
We have been here before. Not with this technology, not at this scale, not with this speed. But the structure is familiar: transformational promise, private capital exhausted, public markets summoned to hold the bag or reap the reward depending on which version of the future turns out to be correct.
The difference this time — the thing that makes this cycle genuinely distinct — is that the technology is not theoretical. It is already displacing workers. It is already concentrating capability. It is already reshaping the economics of professional services in ways that are measurable and accelerating.
The bubble question is whether the valuations are justified. The displacement question is whether the transformation is already happening.
Those are two different questions. Both have the same answer.
Yes.
The Sovereign Auditor covers digital sovereignty, cybersecurity governance, and data protection policy—with particular focus on Isle of Man jurisdiction and Crown Dependency issues.
Payments via PayPal. Credentials delivered by email. No Substack. No Stripe. No middlemen.
