Anthropic’s IPO is all about ignoring the future

Microsoft

Wed 17:17

The author is a Reuters Breakingviews columnist. The opinions expressed are her own.
By Karen Kwok
NEW YORK, Oct 28 (Reuters Breakingviews) - High-tech public-market hopefuls are often short on profitability and long on dreams. Just look at the wave of electric-vehicle startups that listed after the pandemic, or deepening losses at rocket-maker SpaceX SPCX.O prior to launch. Artificial intelligence lab Anthropic hawks perhaps the most futuristic technology going, warning that its research is so incredible that it may threaten humanity. Yet its hunt for a $2 trillion valuation hinges on finances that may look better today than they do a year from now.
The Claude maker’s Monday IPO filing revealed some surprisingly good news for the famously cash-incinerating business of training AI models. Yes, it reported a gigantic net loss of $81 billion in the six months to June, largely due to the changing value of convertible bonds. Yet the company led by Dario Amodei saw spending on data-crunching servers fall to 55% of revenue, from 200% a year earlier. Better yet, operating income turned positive in the second quarter.
But there’s a big catch: some $800 billion of leases and purchase commitments for computing power, many of which have not yet kicked in. A $114 billion contract with Google Cloud, deals with Amazon.com AMZN.O and Microsoft MSFT.O, and an agreement with Broadcom AVGO.O to lease Google-designed chips are non-cancellable. An expensive pact with SpaceX can be turned off, but provides a huge number of vital chips. Add those contracted payments together and split them evenly over their commitment periods, and Anthropic may be on the hook for roughly $97 billion in 2027, Breakingviews calculates.
To sustain its 55% computing-cost-to-revenue ratio achieved in the first half of the year, the company’s top line would need to rise to $177 billion in 2027. In the six months to June, Anthropic generated around $16 billion of sales, according to its prospectus, up ten-fold from the same period in 2025. If they can keep growing at such a phenomenal rate, the math should work. 
Early indications are that this will be very difficult. Quarter-over-quarter growth already slowed to 45% in the three months through September, people familiar with the matter previously told Breakingviews. If that rate holds steady for the rest of the year, sales would have to leap by about 200% in 2027.
The spending burden will only grow. Anthropic’s filing warns of "substantial capital expenditures" that will "change the cash flow profile of our business." While it moves to lock in these massive sums, business customers can – as it admits – cut usage or terminate contracts with little or no notice. Big Tech likes to promise an ever-brighter future. In AI, the next twelve months are rife with danger.
     
Follow Karen Kwok on LinkedIn and X.
     
CONTEXT NEWS
Artificial intelligence lab Anthropic publicly filed the registration statement for its planned initial public offering on October 26. 

(Editing by Jonathan Guilford; Production by Maya Nandhini)
((For previous columns by the author, Reuters customers can click on KWOK/karen.kwok@thomsonreuters.com))