The numbers don't add up. Let me say that plainly.
Anthropic IPO at $2 trillion? SpaceX at $1.75 trillion? These aren't valuation estimates — they're unicorn fanfiction. Based on my years tracking on-chain data and off-chain deal flow, valuations at this magnitude belong to Apple, Microsoft, and Saudi Aramco, not AI startups still burning cash at a rate that would make a crypto project blush.
But here's what the skeptics miss: the numbers might be inflated, but the story is real.
Speed is the currency, but accuracy is the vault.
The core claim — that Anthropic is planning a Nasdaq IPO, potentially as soon as October, setting up a showdown with OpenAI's "we're not going public" stance — that's not hallucinated data. That's a genuine structural split emerging in how AI's two most influential labs handle their relationship with public markets. And THAT story matters far more than any specific valuation headline.
**Context: The AI Capital Divergence
**
Let's reframe. Forget the trillion-dollar fantasy for a moment and look at what's actually happening.
Anthropic, backed by Google and Amazon, is moving toward a public listing. OpenAI, backed by Microsoft, is cementing its private status. This isn't a coincidence. It's a deliberate governance choice that reveals fundamental differences in how these companies view their own futures.
Sam Altman has been explicit: OpenAI won't go public because AI poses "existential risks" that require flexibility private companies possess. That's a compelling narrative — and also a convenient shield against the massive governance complexity making an IPO nearly impossible for OpenAI's current structure. The nonprofit controlling entity, the complex shareholder arrangements, the ongoing tension between safety research and revenue generation — these don't disappear with a ticker symbol.
Anthropic took a different path from inception. They structured as a Public Benefit Corporation (PBC), theoretically designed to balance mission and profit. That's the whole point — a PBC is supposed to be the "responsible" option, the way for a safety-first AI company to access capital markets without abandoning its principles.
Here's the problem nobody's talking about: a PBC structure has never been tested at this scale. When your valuation was $18 billion, the fiduciary tension was manageable. When you're answering to public shareholders demanding quarterly growth, that "benefit" clause becomes a legal and operational nightmare.
**The Core Data Problem
**
Let me be direct about what I see in the numbers. The $2 trillion valuation claim for Anthropic — even being generous with "up to" framing — represents a roughly 30x premium over their last verified funding round of approximately $60 billion. For SpaceX, the $1.75 trillion claim represents roughly a 5x premium over their latest secondary valuations around $350 billion.
Two inflated numbers appearing in the same story. Same "people familiar" sourcing. Same "Nasdaq victory" framing.
That's not a coincidence either.
This looks like the AI equivalent of the 2017 ICO whitepaper cycle — where every new project was "disrupting" industries worth billions, where token allocations were "aligned incentives" rather than wealth transfers, where the founders spoke in sweeping visions while the token economics told a different story.
The valuation inflation tells us something specific: someone wants this narrative to be big. Nasdaq wants to claim the AI IPO crown. Underwriters want to set price anchors. And Anthropic's investors — Google, Amazon, Lightspeed — want maximum exit optionality.
But here's the test that matters: the S-1 filing.
When Anthropic files their S-1 — if they file — it'll contain the numbers that actually matter. Annual Recurring Revenue. Burn rate. Customer concentration. Gross margins. These disclosures will either validate the bull case or expose the gap between narrative and fundamentals.
I've seen this pattern before. The Terra Luna collapse didn't happen because nobody knew the 20% yield was unsustainable — it happened because retail investors were so dazzled by the narrative that they never bothered to look at the underlying cash flows. The AI IPO wave will be different in scale, but identical in structure. The story will always try to outrun the numbers. Our job is to make sure we're watching the tape.
**The Contrarian Angle
**
Here's the angle the mainstream coverage is completely missing: Anthropic going public might actually be bad for the AI safety narrative.
Right now, Anthropic's "safety-first" positioning is their primary brand differentiator. Constitutional AI. Responsible Scaling Policies. These aren't just marketing — they're genuine organizational commitments that set them apart from competitors.
But here's what happens when you go public: you get quarterly earnings calls. You get activist shareholders. You get institutional investors who specifically bought your stock because they think you can "monetize the safety angle" — meaning cut the expensive safety research that's slowing down model releases.
The PBC structure was designed for this, but there's a reason no major tech company has successfully navigated the PBC-to-public-market transition. The fiduciary duty to shareholders and the "benefit" commitment to stakeholders were never designed to coexist under the pressure of a stock price.
OpenAI's "we're staying private because AI is too dangerous" stance might actually be the more honest position — not because existential risk is the reason, but because it's the result. They've built a governance structure that makes IPO practically impossible, and they're using the safety framing to make that constraint feel like a choice rather than a limitation.
Meanwhile, Anthropic is about to find out whether you can actually run a safety-first AI company when your shareholders can vote you out at the next annual meeting.
**The Takeaway
**
Watch the S-1. Not the rumors, not the "people familiar" leaks — the actual S-1 filing will tell us everything we need to know about whether this valuation has any connection to reality.
If Anthropic files and the numbers show ARR in the billions range with a $2 trillion valuation, that's not a misunderstood opportunity — that's a signal to step back from the entire AI IPO narrative until the dust settles. The infrastructure play (NVIDIA, cloud providers, data centers) remains solid regardless, because compute demand is real even if valuations aren't.
But if they file and the numbers make sense — if the growth trajectory and unit economics justify even a fraction of the narrative — then the AI capital floodgates open. Every AI company from here to Singapore will use that S-1 as their pricing anchor.
Either way, we're about to find out whether the AI industry can survive contact with public market discipline. Echoes of 2017 whisper through every new bull run — but this time, the leverage is institutional, the stakes are macroeconomic, and the story is being written in regulatory filings rather than whitepapers.
Don't blink. The ledger doesn't forget.