Anthropic’s revenue forecasts are difficult to underwrite

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 2:47

   the new filing we do know that the compute bill over the next few years is going to be very big we already knew it was going to be quite big and being able to put this new number on it 900 billion dollars i mean sounds enormous that may also change as we go forward it may get bigger but the revenue is also growing really quickly as you said it slowed this quarter to about you know 1100 percent growth year on year which slowed from like 1400 i mean these numbers are completely unhelpful, basically, because these growth rates are so high. And extrapolating exactly how much growth we're going to see next year is really, I mean, I think it's a fool's errand. What we're looking at, though, broadly is that Anthropic has been guiding, we've heard, to revenue of about $200 billion in the year 2028. So even if it's spending, you know, like $100 billion a year on

2. 3:36

   compute, the revenue still exceeds that. The question, of course, is whether you believe it can hit any of these numbers or whether these numbers will start to look small the closer we get to those deadlines if it does hit 200 billion dollars it seems like this might be a business that that is feasible the trouble is it seems like it seems like a seriously big if i mean we had revenue of $17 billion. We also know that growth is obviously slowing, but also that prices appear to be coming down. And they said this, they had these price reductions from Q2 to Q3. They're now coming out with these discounts, largely due to the open-weight models that we've been seeing that have been kind of flooding the market coming from China. I mean, is $200 billion, is that reasonable?

3. 4:21

   Should we take that seriously? What do you think? I mean, as you say, there are plenty of risks weighing on this, and the price war is a big one. The discounts that they put on their products, it seems, were quite small, according to this new amendment, maybe like 3% of revenue in discounts. But really, and if they do make $200 billion, by the way, then a valuation of something like $1.4 trillion, which is kind of roughly in the ballpark of what we're talking about, wouldn't be ridiculous compared with companies like SpaceX, if you believe.

4. 4:48

   If this SpaceX's valuation is, same too. But my issue with this IPO from the beginning has been that it is almost impossible to predict with any certainty where these revenue numbers are going to go. And we also have in this latest amendment some new tweaks to the risk factors section of the prospectus. Not huge ones, but things that remind you that there are some sort of, you know, unpricable factors that may weigh on this.

5. 5:11

   the one that they've talked about in the latest edition is this idea that they may be getting very close to or have passed what they describe as the moment of dramatically accelerated progress attributable to AI, which is like when AI starts basically improving itself more broadly. And when that happens, then they are supposed to step on the brakes in certain ways to make sure that AI is being developed safely. What does that do to the revenue? It's like, it really is impossible to say at this point. One of the more interesting developments in AI over the past week was something that came from Elon Musk. He tweeted, someone had pointed out how GrokBot, which is