Anthropic – A Shot in the Foot

AI doomsday chatter dents IPO prospects

• The AI Doomsday debate (see here) that was largely triggered by Dario Amodei has led to a delay in the IPO and questions around its ability to keep up its breakneck pace of growth in 2027, which together raise doubts about the success of the listing.
• Anthropic was supposed to go public in October, but this looks to have been pushed to November or later, as Anthropic has triggered the mother of all product liability issues, meaning that the IPO documents need to be revised.
• Prior to the recent concerns, many commentators were predicting that Anthropic would manage something in the region of $300bn – $400bn of revenues in 2027, which was instrumental in supporting the idea that the company could go public with a valuation of $2tn.
• At these sorts of revenues, the company would be very nicely profitable, leading to investors having little difficulty in seeing further upside in the shares from $2tn, which is what one needs for the shares to go up following a listing.
• Instead, an employee resigned from the company and raised a series of evidence-free concerns, termed by the media as “whistleblowing”, which Anthropic then doubled down on by calling for an industry slowdown in the development of advanced AI algorithms.
• The fear is that these algorithms, when let loose, would be so advanced that they would be able to improve themselves, decide that humans are a waste of resources and exterminate the lot of us.
• When a company goes public, it has to disclose risks it thinks that may cause its business to fail to meet expectations, and by its own admission, Anthropic has hinted at the mother of all product liabilities.
• To be fair, if Anthropic wipes out the human race, there will be no one left to file a lawsuit, but I have a feeling that the risk section of the S1 now needs to be revised, which is why the IPO has slipped by one month.
• This topic does not fill me with dread, as there is no hard evidence whatsoever that we are anywhere close to kind of superintelligence that would imperil humanity (see here), and the industry has cried wolf on this topic several times already.
• Unfortunately, one side effect of this is that investors have been shaken awake by this debate and are now starting to wonder if Anthropic can actually deliver the kind of numbers that were thought to be gospel just one month ago.
• OpenAI appears to have retaken leadership in terms of revenue growth, and RFM Research has concluded that it may have a big cost advantage over its rivals in 2027.
• Suddenly, investors have gone from swallowing huge numbers with barely a gulp to asking rational and nuanced questions, which has been triggered by the AI Doomsday debate that Anthropic itself has nurtured and amplified.
• I continue to think that if Anthropic really believed in the potential of AI to destroy humanity, it would disclose exactly what its employees have seen and delay its IPO until such time that these concerns have been definitively addressed.
• It is not doing anything like this, leading me to think that at best, the aim is regulatory capture and at worst, a cynical attempt to flatter short-term profitability.
• I have to reserve final judgement until I have had an opportunity to examine the S1, but at the moment, I think that if I had to choose one, OpenAI is probably the better bet.