5 ms·

People seem to think that’s an excessive valuation. I’m tempted to disagree—not that it’s high (it’s stratospheric) but that it makes sense. 1. People seem to

by bertil 3y ago

People seem to think that’s an excessive valuation. I’m tempted to disagree—not that it’s high (it’s stratospheric) but that it makes sense. 1. People seem to think they don’t have a moat: They likely have more talent than anyone else in the industry. Any one joining also expect that being an alumni is almost a guarantee to start another company with a billion-valuation series A; that might hurt retention, but not attraction. They have more data because they got it before everyone closed the hatches in reaction to their product; they now have the most varied interactions with users, collecting more; their partnership with Microsoft, which owns GitHub, gives them access to all the code. It’s the first company that could stick it to Google on its own turf and stay credibly in the lead for a year after that. There’s a high chance they will be replaced, but they do have more of a moat than anyone. 2. The quality of the product: it is essentially plug-and-play for most use cases, and after a year of recurrent innovations, I suspect they have a lot more coming. 3. Business model: so far, it’s $20/m/p. paid for and adopted by millions overnight. Anyone using it says it’s worth ten to a hundred times more. They are selling compute at scale, and the comparison is with human white-collar labor. There’s a lot to be said about the expectations at that price: they’d have to grow fifty-fold and then some to justify it. Still, it’s not entirely unthinkable, not the way over-priced flexible real-estate or over-complicated juicing pouches have been.

COGlory 3y ago
Talent in an emerging field is simply not a moat. What OpenAI has is a headstart. Others can still catch up. Especially as the field gains popularity and more smart folks enter it.
- bertil 3y ago
  The talent itself isn’t, but the fact that they have the strongest signal for talent: they pay more, give more tools and talent-complements (data, training sets), and have more open business model helps them attract more talent for now. It’s not a strong moat, but it is an advantage.
lumost 3y ago
If I’m doing the math correctly, growing their existing end-user offering at 20/month to a billion users would make their revenue 240 billion/year. As we don’t have good insight into the long term cost trajectory of GPUs - it’s unclear what the final margin will be. One thing I could say for certain is that paying a 90% margin on GPU compute for 240 billion COGS is nonsensical. The pressure to cut out Nvidia, or improve inference performance on Nvidia will be immense.
- belval 3y ago
  There aren't a billion person on Earth to pay 20$/month. Not even sure that there are more than 100M. Paid services are for B2B these days and that's where their partnership with Microsoft knee caps any chances for astronomic growth.
  - anonylizard 3y ago
    There is about 1 billion people in the developed world. North America + Europe + Australia alone is 700 million. It is absolutely conceivable that every single person in the developed world will have an AI subscription, in the same way we have water, electricity, internet subscriptions.
  - trogdor 3y ago
    Why do you believe that paid services are only B2B these days? For one counterpoint, there are a quarter-billion paid Netflix subscriptions worldwide.
  - squigz 3y ago
    As someone else pointed out, Netflix has more than 250m subscribers. Spotify does too.
- bertil 3y ago
  I think a better way to look at it would be as a ratio of GDP attached to white-collar work. I’m happy to consider 20 $/m * 12 m/y * 300M = 7.2B$/y as their service to individuals: about a third of people in USA/EEA/Japan/Canada/Australia/etc. with a fairly high margin over computing costs. That alone could justify a valuation between 20-50B$. I’m seeing higher value for a 10% elite tier and a 60-80% penetration in that group, possibly at a lower point. They could have a billion paying users in India, China, Indonesia, Nigeria, etc. at a lower price point. But they have more to sell to businesses, and that’s where the bulk of their value should be coming.
_8j50 3y ago
They have competition, you can argue the competition is inferior but the question is, can they monetize it (despite their org structure) enough to make them one of the most profitable companies on earth? Why would people not use llama* or gemini for a lower price? OpenAI has to be so good, it makes no sense to go with alternatives and even then, I can imagine a $10B market in a few years but not 100B. This valuation only makes sense if most people and companies start paying for it, kind of like iphones, windows, exxon oil. Can Azure,AWS or GCP be even valued that much? Azure just hit 100B revenue in 2022 (not even sure if they profited off of that). The only way this valuaion makes sense is if you are betting in 10 years openai will be making 10B+ net profit a year.
- bertil 3y ago
  I know many people who have set up Llama locally (myself included), and they mainly use OpenAI tools for their day-to-day tasks. I can’t imagine that for the vast majority of people, Llama would be an option, not without extensive packaging. That will happen, but the way Android has been available, usable, and even good for a decade, while all the profits are for iOS/Apple. Again: I’m not absolutely convinced this will happen at 100%, but more than 10, more than 20% likely? Yeah. An oligopoly with Mistral, Anthropic, and Google might be more likely, but they’ll all make a lot of money. My point isn’t that you should invest at that price but that investors who have are not completely detached from reality. They are confident, for sure and might be over-paying a bit, maybe by a factor or 2 or 3 — but not the way people who invested for Theranos, WeWork, or Pets.com. It’s not a scam; it’s just a guy who can credibly say: “I’ll get as much money as a worldwide phone operator within a few years.” I don’t know if he will, but I can see how.
  - _8j50 3y ago
    I agree with what you said. The only "scam" angle might be that this might be the pump stage of a high end pump-and -dump, the private investors might have had IPO talks and/or MS taking over the operation and paying them at that valuation.
hasty_pudding 3y ago
I have moved probably 90% of my search to OpenAI. Its faster and better. I cant be the only one. Does not bode well for googles search product.
- helsinkiandrew 3y ago
  Bard is getting quite good at queries for upto date info: “What was the last FED announcement?” “When does Apple announce results” “Where can I see the Barbie movie online in Finland” Google Search has a vast system gathering live data that can be added to Bard - as probably is being done to Bing

by bertil 3y ago

COGlory 3y ago

Talent in an emerging field is simply not a moat. What OpenAI has is a headstart. Others can still catch up. Especially as the field gains popularity and more smart folks enter it.

bertil 3y ago
The talent itself isn’t, but the fact that they have the strongest signal for talent: they pay more, give more tools and talent-complements (data, training sets), and have more open business model helps them attract more talent for now. It’s not a strong moat, but it is an advantage.

lumost 3y ago

If I’m doing the math correctly, growing their existing end-user offering at 20/month to a billion users would make their revenue 240 billion/year. As we don’t have good insight into the long term cost trajectory of GPUs - it’s unclear what the final margin will be. One thing I could say for certain is that paying a 90% margin on GPU compute for 240 billion COGS is nonsensical. The pressure to cut out Nvidia, or improve inference performance on Nvidia will be immense.

belval 3y ago
There aren't a billion person on Earth to pay 20$/month. Not even sure that there are more than 100M. Paid services are for B2B these days and that's where their partnership with Microsoft knee caps any chances for astronomic growth.
- anonylizard 3y ago
  There is about 1 billion people in the developed world. North America + Europe + Australia alone is 700 million. It is absolutely conceivable that every single person in the developed world will have an AI subscription, in the same way we have water, electricity, internet subscriptions.
- trogdor 3y ago
  Why do you believe that paid services are only B2B these days? For one counterpoint, there are a quarter-billion paid Netflix subscriptions worldwide.
- squigz 3y ago
  As someone else pointed out, Netflix has more than 250m subscribers. Spotify does too.
bertil 3y ago
I think a better way to look at it would be as a ratio of GDP attached to white-collar work. I’m happy to consider 20 $/m * 12 m/y * 300M = 7.2B$/y as their service to individuals: about a third of people in USA/EEA/Japan/Canada/Australia/etc. with a fairly high margin over computing costs. That alone could justify a valuation between 20-50B$. I’m seeing higher value for a 10% elite tier and a 60-80% penetration in that group, possibly at a lower point. They could have a billion paying users in India, China, Indonesia, Nigeria, etc. at a lower price point. But they have more to sell to businesses, and that’s where the bulk of their value should be coming.

_8j50 3y ago

They have competition, you can argue the competition is inferior but the question is, can they monetize it (despite their org structure) enough to make them one of the most profitable companies on earth? Why would people not use llama* or gemini for a lower price? OpenAI has to be so good, it makes no sense to go with alternatives and even then, I can imagine a $10B market in a few years but not 100B. This valuation only makes sense if most people and companies start paying for it, kind of like iphones, windows, exxon oil. Can Azure,AWS or GCP be even valued that much? Azure just hit 100B revenue in 2022 (not even sure if they profited off of that). The only way this valuaion makes sense is if you are betting in 10 years openai will be making 10B+ net profit a year.

bertil 3y ago
I know many people who have set up Llama locally (myself included), and they mainly use OpenAI tools for their day-to-day tasks. I can’t imagine that for the vast majority of people, Llama would be an option, not without extensive packaging. That will happen, but the way Android has been available, usable, and even good for a decade, while all the profits are for iOS/Apple. Again: I’m not absolutely convinced this will happen at 100%, but more than 10, more than 20% likely? Yeah. An oligopoly with Mistral, Anthropic, and Google might be more likely, but they’ll all make a lot of money. My point isn’t that you should invest at that price but that investors who have are not completely detached from reality. They are confident, for sure and might be over-paying a bit, maybe by a factor or 2 or 3 — but not the way people who invested for Theranos, WeWork, or Pets.com. It’s not a scam; it’s just a guy who can credibly say: “I’ll get as much money as a worldwide phone operator within a few years.” I don’t know if he will, but I can see how.
- _8j50 3y ago
  I agree with what you said. The only "scam" angle might be that this might be the pump stage of a high end pump-and -dump, the private investors might have had IPO talks and/or MS taking over the operation and paying them at that valuation.

hasty_pudding 3y ago

I have moved probably 90% of my search to OpenAI. Its faster and better. I cant be the only one. Does not bode well for googles search product.

helsinkiandrew 3y ago
Bard is getting quite good at queries for upto date info: “What was the last FED announcement?” “When does Apple announce results” “Where can I see the Barbie movie online in Finland” Google Search has a vast system gathering live data that can be added to Bard - as probably is being done to Bing