Anthropic's 355-page prospectus shows $16.2bn in revenue for the first half of this year, but losses doubled to $80.7bn over the same period.
Seven founders, including CEO Dario Amodei, hold a single share giving them a 50.1% voting majority that outside investors cannot match, a structure the prospectus warns may make it harder to replace directors and management.
The hosts frame the expected $1.75tn to $2tn IPO valuation as the central question for investors, weighing that against governance concerns Danny Fortson likens to asking for trust.
Anthropic's IPO filing asks investors to pay trillions on trust
Anthropic's IPO filing shows a company growing at extraordinary speed and losing money almost as fast, while asking public investors to pay up to $2tn without a vote. The founders keep majority voting control through a single share, and the trust above them can remove directors but not set strategy. The case for buying rests on improving economics and on trust in the founders.
Takeaways
Revenue is exploding, but losses are larger. Revenue rose 1300% in a year to $16.2bn in the first six months of this year, or about $6.6m per employee. Losses for the same half were $80.7bn, double last year's.
Compute efficiency is improving, but commitments are enormous. Computing spend fell from $2.50 to $0.50 per dollar of revenue, and about 1,900 customers spend $1m or more. Commitments still total close to $800bn, and Danny Fortson expects spending to explode.
Investors must price a $2tn listing. The IPO aims for a value between $1.75tn and $2tn, which the hosts treat as the central question. With Anthropic raising around $100bn, Meta up to $95bn and Musk $40bn of GPU debt, about $250bn could be raised in 24 hours.
Founders keep control through one super-voting share. The seven founders, including Dario and Daniela Amodei, hold a single share with a 50.1% voting majority while at least three remain. Outside investors get no vote. Fortson compares it to Zuckerberg at Meta and calls it asking for trust.
The Long-Term Benefit Trust has limited power. The trust has up to seven trustees who can remove board members but cannot directly shape strategy. Neil Shah leads it, and others include Richard Fontaine, Ben Bernanke and Christopher Stubbs. Fortson calls OpenAI more of a viper's nest.
Timing and missing data add risk. The filing omits the last three months through September. Fortson suggests cheap open-source models have taken market share. The hosts also expect Democrats to sweep next week's US elections, and a backlash against data centres could slow the build-out.
The prospectus pairs heavy risk language with big promises. The risk section runs 90 pages and uses the word risk 213 times, including a warning that the structure may make it harder to replace directors. Yet it also says AI could accelerate science and cure diseases.
Where the hosts differed
Katie Prescott says Anthropic would be the first frontier AI company to go public. Danny Fortson questions that, citing SpaceX and Grok.
Prescott wonders whether earlier leaks dampened expectations for the filing.
Fortson notes Anthropic once talked about curing cancer and Alzheimer's, and says the good side of AI has been drowned out by scary stories because no breakthrough has yet appeared.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Who supplies the computing
Amazon is described as a huge investor in Anthropic and is also one of its computing suppliers.
Other suppliers are NVIDIA, Broadcom, AMD, SpaceX AI and EnScale.
Spending with the British company EnScale is reported at $44.6bn.