Anthropic’s boosters are rushing to celebrate their ARR.
As they do, be mindful that the acronym ARR can actually mean two rather different things, Annual Recurring Revenue or Annualized Run Rate. (And 65 other things, less immediately relevant.) You might think they are the same, but they are not.
What the boosters want you to think of is Annual Recurring Revenue. Annual Recurring Revenue is great; that’s things like subscriptions, which means next year likely to be same as (or better than) this year. Annual Recurring Revenue recurs.
Annualized Run Rate, alas, means no such thing. It can mean as little as you did something like took your best month, multiplied by 12 and projected that your best month this year would be your average month next year. That could happen. Or maybe your best month is NOT your typical month, and it will be better than your average month next year. You might in fact never have a full year that is twelve times your best month.
When people tell you Anthropic had this or that ARR, you really need to know which meaning they are talking about. They rarely tell you. But they really mean the latter.
Indeed, THE big issue for Anthropic is that the kind of revenue they were making in the second quarter of this year may well not recur, at the same scale, next year, now that, in the aftermath of the collapse of tokenmaxxing, companies like ATT are increasingly turning to open source models from other vendors, in order to save costs.
§
Speaking of meteoric rises that might not last, back in 1995 Netscape had revenue. Lots of it. Doubling every quarter. They went public. They were worth billions without having ever turned a profit. (Sound familiar?)
And then things changed. Later that year, Microsoft offered a competitor, for free, and Netscape was never the same, all but out of the game a decade later.
§
When someone boasts about ARR, be sure you know which term they are referring to. Caveat emptor.
Anthropic seems to be racing towards an IPO in September in an attempt to get at least something back for the founders and employees (based on that 'ARR' in part). OpenAI apparently went the route to stay outside the market for now (their conversion rate from free users apparently stinks so much that they have no chance to do an IPO) and do a $7 billion share buyback (to keep people from leaving the sinking ship?). Paid for by investor money?
Popcorn, anyone?
I truly, truly wonder when the whole thing will collapse. LLMs are here top stay and (token/pixel based) brute force approximation of understanding will be useful in various domains at various levels of quality (it's approximation, after all, and not the real thing), especially domains that have the possibility for strict testing (Lean, anyone?, I.e. true for math, much less for coding (grammar yes, the rest not as much), even less for most real life things like selecting bombing targets that were enemy buildings but turn out to be schools by now.).
How long can the charade of "we've reached the singularity (or maybe not)" go on before the financial and business world stops its FOMO-obsession? Why is that taking so long?
All Rosy Returns