Eliza Ward: Hey — good to be here.
Brian Reed: Yeah, same. You've had the Journal piece open for how long now?
Eliza Ward: Since this morning, honestly. Okay — Anthropic is filing its IPO as early as tomorrow. October 21st. That's the date. Wall Street Journal reviewed the draft prospectus.
Brian Reed: Right — and the number that just stopped me cold was the net loss. Eighty point seven billion dollars.
Eliza Ward: On sixteen point three billion in revenue. In the first half of this year alone.
Brian Reed: So that's — wait, let me just say that plainly. They burned more than five dollars for every dollar they made. And they're asking the public markets to fund more of that.
Eliza Ward: Nasdaq listing. Morgan Stanley, Goldman Sachs, JPMorgan on the underwriting side. This is not a small shop operation. Dario Amodei's company is going fully public.
Brian Reed: The part I don't get — is the eighty point seven billion the thing that should scare people, or is it actually the thing that tells you how expensive frontier AI is to build? Because those are very different reads.
Eliza Ward: Both, actually — but the second one is what makes the first one irrelevant. Because the burn rate doesn't tell you whether to invest. The noncancelable commitments do. The Wall Street Journal pulled language from the prospectus: roughly $450 billion in locked cloud compute contracts over the next decade. You can't raise your way out of noncancelable.
Brian Reed: Okay, so — think of it like a homeowner who signed a 30-year mortgage on five houses before their salary arrived. The payments are locked in whether you have income or not. You have to find money, fast. That's Anthropic's infrastructure position right now.
Eliza Ward: Wait — but didn't they just raise sixty-five billion in May? The Series H.
Brian Reed: Because yes — sixty-five billion, largest private AI fundraise on record. Private capital clearly hasn't dried up for them. So the Axios framing, that this is financial necessity, that they've grown too big for private markets — I mean, is that actually right? Or is it something more specific?
Eliza Ward: The distinction is the size of the obligation. Computing and infrastructure hit nearly nine billion in H1 2026 alone — that already cleared what they spent in all of 2025, which was seven point three billion. And Bloomberg put the total future commitments at potentially seven hundred fifty billion when you include what's still under negotiation.
Brian Reed: Seven hundred fifty — wait, really?
Eliza Ward: The prospectus says over six hundred billion, another hundred fifty under negotiation. No private round — not sixty-five billion, not even three of those back to back — covers a commitment that size. The sixty-five billion is real, but it's not even close to the denominator.
Brian Reed: So the IPO isn't desperation exactly — it's the only instrument that can actually reach the scale of what they've already signed. The SEC got the confidential draft in June. That's already done. The process is moving whether the valuation debate resolves or not.
Eliza Ward: Right — and that's the signal. Not the headline number, not the hype. The filing happened. The commitments are locked. The question now is whether the public market will accept what private capital already did.
Brian Reed: But that's actually where the circulating take falls apart — because the comparison everyone keeps reaching for is SpaceX. SpaceX went public in June at roughly two trillion. So the argument is: Anthropic targets two trillion, SpaceX hit two trillion, therefore the number is real. And I just... that comparison doesn't hold the weight people are putting on it.
Eliza Ward: No, it doesn't. SpaceX's cost and revenue profile is completely different — and the investors reading the Anthropic prospectus apparently agree, because some of them are already revising toward one point five trillion. That's a twenty-five percent haircut from the target.
Brian Reed: Twenty-five percent off before the roadshow even starts.
Eliza Ward: And here's what makes that revision legible — the Series G in February valued Anthropic at three hundred eighty billion. The Series H in May, three months later, came in at nine hundred sixty-five billion post-money. Nearly tripling. In ninety days. That's not new math. That's narrative momentum compounding on itself. The prospectus is the first time the actual numbers are in front of institutional eyes.
Brian Reed: Right — but Morgan Stanley, Goldman, JPMorgan. Those aren't firms that get captured by hype. If the banks signed on, doesn't that push back on the 'narrative only' read?
Eliza Ward: It does, and I don't want to overstate it — wait, actually, that's almost the point. The banks signal institutional seriousness. But the same institutional investors are the ones marking down to one point five. So you can have serious underwriters and still have a valuation that was built on momentum rather than disclosed financials. Those aren't contradictory.
Brian Reed: The losses widened from thirteen billion in H1 2025 to eighty point seven billion in H1 2026 — so the direction of travel is the problem, not just the level.
Eliza Ward: That's what makes Wilson Sonsini interesting to me. They handled Google's 2004 IPO. You bring that firm in when you're treating this as a landmark transaction, not a fast raise. This is being run methodically. And there's still one piece of the prospectus we haven't touched that — I think it's going to complicate everything else we just said about investor confidence.
Brian Reed: The existential risk language.
Eliza Ward: Yeah — and that's the part that doesn't fit anywhere comfortable. The prospectus tells the SEC — under penalty of perjury — that Anthropic's own models could pose existential risk to humanity. That's in the same document asking pension funds to buy in.
Brian Reed: So picture the pension fund analyst — October 21st, she opens the draft, and on one page it's civilizational upside, comparable to electricity and industrialization. Two pages later it's... existential risk to humanity. She has to sign off on both.
Eliza Ward: And she can't dismiss it as boilerplate. That's the thing. Anthropic was founded specifically as an AI safety company — Dario Amodei and others who left OpenAI over safety concerns. So when the safety warning appears in the S-1, it's not a liability hedge. It's... actually, it's load-bearing to the whole identity.
Brian Reed: Which means retail investors can't just file it under 'CEO is being honest and refreshing.' The disclosure creates a real decision tree.
Eliza Ward: Right — and CoreWeave is sitting right there as the live cautionary signal. Publicly traded AI infrastructure company, already under scrutiny for revenue prospects against its debt load. The market has a recent data point on what happens when AI infrastructure bets meet quarterly earnings pressure.
Brian Reed: So the existential risk line — does it actually tank retail demand, or does institutional money just... price it in and move on?
Eliza Ward: I mean — I don't know yet. But here's what's concrete: OpenAI has also announced IPO plans. If Anthropic lists first, that's not just financial, it's reputational. First mover in public markets for frontier AI. That changes what OpenAI can ask for on its own filing.
Brian Reed: And Renaissance Capital is already flagging that missing the October window compresses or kills a 2026 listing entirely — midterm elections, holiday season. The window is narrow even after a public filing goes in.
Eliza Ward: The prospectus itself conditions timing on market conditions. So the concrete thing to watch is whether the October filing date holds — because if it slips, the question isn't 'when do they list,' it's 'do they list this year at all.'
Brian Reed: And that's — I mean, that's the question I can't get past. If the market prices Anthropic at one point five trillion despite the losses and the locked capex, every other AI company just got a floor and a ceiling at the same time. That's the thing. A floor because, okay, this is what frontier costs. A ceiling because... the math is visible now.
Eliza Ward: One point five trillion — even at the revised number — that's still one of the largest public offerings in history. Full stop. We don't need to dress that up.
Brian Reed: Right — but does the market absorb that as 'cost of the frontier,' or does it read the $80.7 billion loss and the $450 billion in noncancelable compute and decide this is a warning about the whole sector? Because those are genuinely different outcomes and I don't think we know yet which way institutional money lands.
Eliza Ward: No, we don't. And that's — actually, that's the honest place to stop. The filing is real. The commitments are locked. Whether public markets call it normal or a ceiling for everyone chasing Claude's position — that answer is ahead of us, not behind.
Brian Reed: Watch October 21st.
