I am totally blown away that a company that was formed in 2021 will have a 1.5 trillion market cap on the public market... to put into context, a valuation not achieved by Apple until 2020
The picks and shovels are down 40% in the last two months. You can buy them at a forward P/E of less than 3 if you think earnings hold through 2029. Which is another way of saying that the market doesn't expect them to. There is a staggering amount of money being bet on "it falls apart soon and it falls apart hard."
Or, in the exceedingly long (in AI-land terms) timeframe you mention, they could rocket to the moon and then crash and go bankrupt because everyone uses them to bits, looks for another alternative, and local models on commodity hardware get to a point in 3 years where nobody wants to spend on tokens any more.
They are only cashing out like $2B. Non-founder directors, employees etc can also sell 10% of their shares immediately, but most of the rest is locked up for 180 days.
Perhaps they were expecting to be more oversubscribed? Like they increased the number of shares offered to keep a target amount of money. Likely they see the rough few months ahead. Ramp.com AI index shows spend on Anthropic AI was in decline/flat July 31 - Oct 31
Likely they're best bet is to weather the continued devaluation of tokens and cheaper models until they hit on a better business model, which honestly is probably not that hard.
Thanks for sharing. Something seems off in that data. It says in Nov, Anthropic's share of token volume is 55.1% and share of token spend is 49.3%. Given Anthropic tokens are pricier than other vendors, one would expect share of spend to be higher than share of token volume.
That's a good point, likely there are other discounts / subsidies being realized on accounts? I think from Ramps point of view it might be hard to distinguish these.
That being said, I have also noticed that at times the X axis dates of the weeks changes. In the past, like a week ago, I noticed that the "Week ending Oct 23" changed to "Week ending Oct 25". Which also noticeably changed the shape of the charts.
I work with data a lot so cut them some slack, but I do with they had more grounded information in the report (a lot of % and not a lot of specific values).
I also wish they would somehow tie in to the massive OpenRouter Rankings. I think the two together would help paint a much more cohesive picture of how token budgets are spent and change week to week.
Oversubscription was kind of a self fulfilling prophecy. Large investors and institutions placed larger orders knowing the IPO would oversubscribe and they wouldn’t get their stated allocation.
It's a common pattern for IPOs, regardless of demand. Companies gauge the price range during the roadshow, make an initial announcement at a low price, then raise the target at the last minute going into the IPO to make it look like a hot offering.
The previous range was 70-78 and it priced above the target range, at $83.
So they are selling 20% more shares and selling them for 15% more than the midpoint of what they were targeting, because the demand is overwhelmingly high. They are going to make $20B more from the IPO than planned.
The rumor is they were oversubscribed by more than 6 times over. SPCX was reportedly only 3.5-4.
About 1.1T market cap will exist tomorrow prior to opening bell, out of a potential 1.5T that includes .4T stuck in options, convertible notes and such.
But it’ll likely open above 83 when trading actually begins. We may see upwards of 2T if it goes over 100 a share
so many reasons not to take this at face value - greenshoe, lockup periods, 5% float. consider the latter alone - in this math, the 95% unsold are considered to have the same value as the rarified 5%. that defies economics.
For actual science done by scientists they're far behind open models now.
The people I know using this model are perpetually hamstrung by slowness, pedantry, misplaced assumptions and appeals to authority. You can't move forward if you're just trying to norm your training data.
They've created this extremely slow, overwrought system with lots of guardrails and token limits designed only to create addiction. Why would you use that when you can have a marginally less "smart" system that's 50x faster and 100x cheaper? Who cares if the model one shots something. It needs to be checked anyway, by test in a machine or a lab.
So far the pension funds have been major investors in everything AI, every fund is in it hard and deep now. It absolutely will be in your retirement fund, but don't worry your unlikely to be able to get to retirement unless its really soon.
https://seekingalpha.com/symbol/SNDK/valuation/price-earning...
Are you (the IPO buyers) really expecting this company to do another 100x from its inflated valuation even after the lock up period expires?
AGI will be achieved in 15 hours and you will become their exit liquidity during the IPO fan fare.
And they're taking around ~105B in cash.
81.34B from the 980M shares at $83 and 26B from Google/Nvidia/etc, minus 1.9B the existing holders selling 23.6M shares.
Likely they're best bet is to weather the continued devaluation of tokens and cheaper models until they hit on a better business model, which honestly is probably not that hard.
That being said, I have also noticed that at times the X axis dates of the weeks changes. In the past, like a week ago, I noticed that the "Week ending Oct 23" changed to "Week ending Oct 25". Which also noticeably changed the shape of the charts.
I work with data a lot so cut them some slack, but I do with they had more grounded information in the report (a lot of % and not a lot of specific values).
I also wish they would somehow tie in to the massive OpenRouter Rankings. I think the two together would help paint a much more cohesive picture of how token budgets are spent and change week to week.
So they are selling 20% more shares and selling them for 15% more than the midpoint of what they were targeting, because the demand is overwhelmingly high. They are going to make $20B more from the IPO than planned.
The rumor is they were oversubscribed by more than 6 times over. SPCX was reportedly only 3.5-4.
But it’ll likely open above 83 when trading actually begins. We may see upwards of 2T if it goes over 100 a share
How are investors not terrified that in what is supposedly the biggest boom ever, the leading company is downsizing its estimate of what it is worth?
(I will be buying shares.)
The people I know using this model are perpetually hamstrung by slowness, pedantry, misplaced assumptions and appeals to authority. You can't move forward if you're just trying to norm your training data.
They've created this extremely slow, overwrought system with lots of guardrails and token limits designed only to create addiction. Why would you use that when you can have a marginally less "smart" system that's 50x faster and 100x cheaper? Who cares if the model one shots something. It needs to be checked anyway, by test in a machine or a lab.
I'm getting 2008 vibes.
Don’t worry. It will.