I don't see how Anthropic can become profitable enough to justify its investments unless they can either stop training models ("Pace the Frontier!!1"), or magically eliminate open source competition so that their volume x margin on inference is high enough to dwarf the training costs.
For the IPO, I assume they're going to try to get everyone to focus on inference - which is profitable in isolation. But I don't know how they'll distract the market from the other half of the equation.
Their inference margins are already high, the thing bringing their net down isn't training per se, it's the fact that they have to train the next iteration right away due to competition, otherwise an extended inference cycle would've eventually recouped training costs.
The bet is that cost per intelligence will keep going down and demand for intelligence will grow at the same time, if this bet holds, a user could get the same or more done with a cheaper model (running-cost wise) while paying the same subscription and the lab gets improved margins.
Open source is a challenge but their current moat is the fact that they're ahead, and as long as they stay ahead they'll be able to charge a premium.
Unlike the dot-com era, this is a bubble that everyone saw coming.
Like dot-com, there are many long-term winners, but cash hungry frontier labs are not among them.
Prediction:
OpenAI, Anthropic, Oracle, and SpaceX[1] go down and assets be sold or companies acquired for a fraction of their valuation. The same goes for most neoclouds. Profitable "old" tech companies like MS, Google, Meta, and Amazon are fine and scoop up the valuable parts for pennies.
Nvidia will be in Cisco's position (Cisco was the world's most valuable company in the dot-com era). Both are good companies that are not going anywhere, but stock prices get decimated when markets collapse (Nvidia liabilities are not high enough to take company down, they are mostly tied to there not being a bubble and data centers are getting built)
[1] Everyone already knows that, according to SpaceX, > 80% of SpaceX's valuation is in AI, not in space stuff, right?
“So, what you're telling me is that the music is about to stop and we're going to be left holding the biggest bag of odorous excrement ever assembled in the history of capitalism.”
A brilliant line from the movie Margin Call (a fictional portrait of the actual events of the 2008 crisis) that appears likely to get reused in upcoming AI bubble burst movies.
New Constructs takes a similar position on all mega IPOs. Sometimes they're right, but sometimes they're wrong. For example, they insisted Uber's IPO valuation was too high due to unsustainable losses, hidden liabilities, etc. Initially they were proved right when the post-IPO stock price plummeted, but in the long term Uber has performed well, become very profitable, and now trades well above its IPO valuation.
I'm expecting a similar story to play out with Anthropic: The stock price will come down significantly after the IPO, resulting in lots of headlines about the AI bubble bursting. But in the long term it'll climb up again and investors who stick around will do very well.
“Now trades well above the IPO valuation” does not mean something would have been a good investment at the IPO price given opportunity cost and the time value of money. Additionally, the current price could equally not represent good value.
I suspect SpaceX is going to be the opposite of what will happen with Anthropic. With SpaceX I think the overwhelming majority of people didn't initially understand what Starship is, how far along it is, or what it even means in terms of its impact on the pathway of getting humans to be a multiplanetary species. So it was greatly undervalued and zoomed upward.
With Anthropic I think a lot of people haven't really realized the trend line of cost of "AI" (both in literal $ terms, and compute costs for local operation) or what open weight systems will do to the market in the "long run", where long run in this case means the next year or two. I don't think any LLM company is going to come out a winner in the longrun, because the price for LLM output is going to approach $0 with no chance for a moat whatsoever.
> the overwhelming majority of people didn't initially understand what Starship is, how far along it is, or what it even means ... So it was greatly undervalued
I don't think anyone was undervaluing the "space" segment of SpaceX. It's the clear leader in launch frequency and cost-to-orbit among the new wave of space firms, and shows no sign of losing that position.
Where the IPO raised a bunch of red flags was SpaceX suddenly acquiring Elon's AI firm, which had a significantly more nebulous basis for its valuation...
> the pathway of getting humans to be a multiplanetary species
I am afraid that you're suffering from Flavor-Aid poisoning here. We need a reality check about "multiplanetary colonization".
I've been following human spaceflight since I was born. I've actually participated in robotic spaceflight. And I'm here to say that it is robots who are destined to be a multiplanetary species. The robots have already successfully escaped our solar system and extensively explored the Moon, Mars, and elsewhere.
Many unscientific people believe that it's harder to colonize the Moon or Mars because they are farther than Africa or Rio de Janeiro is from Rome. But mere distance is not really your biggest obstacle.
I've also been following the ISS since its inception, and they have empirically disproven the feasibility of colonizing planets. If we succeed in establishing a crewed Moon base, it will be an ultra-Herculean effort, and it's on the edge of feasibility.
But Mars? Venus? Asteroids? Europa? You really think humans are going to live there?
The reasons that ISS has disproven this are manifold, but I'll point out two key points:
Their supply-chain is only about 8-48 hours, and they can stash large amounts of provisions up there. Their escape route is also only 8 hours and they're back on Earth in a jiffy. None of this is true for any body outside of Earth orbit. The Apollo missions arrived on the Moon and returned safely, by burning, jettisoning, consuming, and abandoning nearly everything they started with!!
A 15th-century expedition to Africa or Brazil means that you can fish in the ocean, breathe the air, drink rainwater along the way. An expedition to the Moon puts you in an airtight bubble with no stops, no rescue, no resupply, and most importantly, no breathing on the way, nor once you get there.
The ISS has also demonstrated the severe toll on the human physique in microgravity conditions. Watch astronauts who land or splash down. They are carted off in gurneys. They are literally unable to walk unassisted. NASA restricts returnees from driving or even jogging. The astronauts describe having to "relearn gravity" activities like brushing your teeth and showering. Any astronaut arriving on Mars would be as helpless as a toddler! Do you want to colonize Mars with toddlers? Because all the treadmills and workouts along the way won't make them fit for purpose on arrival.
So between here and Mars it's minimum six months, no escape, no rescues, no takey-backsies. We could send humans to Mars if we didn't want them to survive or return. I mean, no problem. There's no air to breathe on Mars; no water to drink; no fuel to burn; no fish to fish nor deer to hunt; no steel to build a shelter. You pack it all into the spacecraft, newbie?
But NASA and the rest of the world has amply demonstrated just how capable robots are where humans fail. Robots are going to inherit the solar system. Robots are in 2026, the de facto interplanetary species, sine qua non. My heart is warmed by these human achievements. I am happy to live and die on Earth. I think that every American should be grateful for Earth, and not permit our oligarchs to cheat and lie to us with fairytales of escaping our Earth.
I fully agree with your points: the hard, unyielding reality which most Trekkie fanatics refuse to acknowledge, except for one: "severe toll on the human physique in microgravity conditions."
That's easily fixed!
Spinning a space station for gravity is straightforward, it's just that the ISS wasn't designed for it, and NASA didn't have the budget to build a second, bigger space station with "extra engineering challenges" on top.
It's definitely practical though, it's just a matter of launching enough mass to orbit.
Like you said, Starship is going to change a lot of things, starting with space station design.
Anthropic's small models are almost certainly distilled from their larger ones. And their biggest training expense are going to be experimental models that aren't even intended for release, just to test out some idea.
The model is engine, the harness is the car Anthropic are selling.
They are a service company, selling home-grown, mass-produced, conveniently packaged intelligence at a well established price for consumer. Their API pricing might just be for value proposition and tax deduction purposes.
Every tech company, tech department, startup that I know is using claude code to deliver. The boutique deepseeks, qwens, glms and kimis are nice, but ain't production ready in 98% of cases. All the pi's, opencodes, etc are nice, but much less enterprise friendly. And google just outright sucks in comparison. There's a use case for each, but not at the scale Anthropic is doing.
Only real competition - OpenAI - is close, but they're burning money to stay there, which imo is unsustainable long term.
Is Anthropic worth $2T? Can't say. But boy they have a good opportunity to test this, and they may not have a similar chance in the future.
I've tried a bunch of LLMs, and I agree that Claude is currently the best at coding, but the gap isn't huge. Gemini 3.8 Flash is surprisingly good and I'm looking forward to trying Gemini 4. The open weights models like Kimi, GLM, etc. are a little behind in quality I think but not massively.
With harnesses I see little difference between what Zed or Pi gives me and what Claude Code gives me. Enterprise features? The AI can help code those in a week.
AI itself is a great breaker of moats.
The question isn't whether they lead today. It's whether there's any kind of moat. Without a moat, frontier model development is an industry like chip fabrication or large volume manufacturing. Those industries are a red queen's race where if you stop innovating for even a short while you sink. Ask Intel what happened when they decided to coast for a few years.
Chip fabs and bulk manufacturers make a lot of money, sure, but they don't command such a multiple. Anthropic is priced like they're going to be to AI what Google was to web search or Apple is to (US at least) mobile and higher-end PCs.
I don't see it. I think they'll be a big player but that big?
Of course... maybe if they suddenly have that much cash they'll be able to carpet bomb the entire world with AI doom Rationalist ideology and buy their regulatory capture and outlaw their competition.
They made it convenient to get value for the specialist masses and I don't think anyone else managed to achieve it yet. That's their moat.
I've seen firsthand attempts at integrating open source in place of it and they all failed. Time is money and it costs premium. Touching the ground running is a huge advantage. That's why cloud services stuck.
I don't see how Anthropic can become profitable enough to justify its investments unless they can either stop training models ("Pace the Frontier!!1"), or magically eliminate open source competition so that their volume x margin on inference is high enough to dwarf the training costs.
For the IPO, I assume they're going to try to get everyone to focus on inference - which is profitable in isolation. But I don't know how they'll distract the market from the other half of the equation.
Their inference margins are already high, the thing bringing their net down isn't training per se, it's the fact that they have to train the next iteration right away due to competition, otherwise an extended inference cycle would've eventually recouped training costs.
The bet is that cost per intelligence will keep going down and demand for intelligence will grow at the same time, if this bet holds, a user could get the same or more done with a cheaper model (running-cost wise) while paying the same subscription and the lab gets improved margins.
Open source is a challenge but their current moat is the fact that they're ahead, and as long as they stay ahead they'll be able to charge a premium.
If the models get good enough, the demand for them will soar.
These companies believe they're a year from AGI, and that's how they'll be profitable.
They don’t care if they get to sell more equity.
This. The product is the stock, the tech is the fomo theater to sell the product.
Unlike the dot-com era, this is a bubble that everyone saw coming.
Like dot-com, there are many long-term winners, but cash hungry frontier labs are not among them.
Prediction:
OpenAI, Anthropic, Oracle, and SpaceX[1] go down and assets be sold or companies acquired for a fraction of their valuation. The same goes for most neoclouds. Profitable "old" tech companies like MS, Google, Meta, and Amazon are fine and scoop up the valuable parts for pennies.
Nvidia will be in Cisco's position (Cisco was the world's most valuable company in the dot-com era). Both are good companies that are not going anywhere, but stock prices get decimated when markets collapse (Nvidia liabilities are not high enough to take company down, they are mostly tied to there not being a bubble and data centers are getting built)
[1] Everyone already knows that, according to SpaceX, > 80% of SpaceX's valuation is in AI, not in space stuff, right?
Interesting take. How high is that probability within the next five years, do you think?
You're of course speaking of a world that doesn't manipulate the cost of memory.
Hard to run local modals when sam altman continues to drive up the cost of compute.
> In an interview with CNBC in 2021, Trainer acknowledged that “crazy stuff happens” and New Constructs doesn’t always get it right.
> “I can’t let that bother me,” he said at the time. “I have to stay true to what I think is right.”
Monkey remains true to darts, dartboard.
I posted other (different) research into the IPO yesterday: https://news.ycombinator.com/item?id=50033644
“So, what you're telling me is that the music is about to stop and we're going to be left holding the biggest bag of odorous excrement ever assembled in the history of capitalism.”
A brilliant line from the movie Margin Call (a fictional portrait of the actual events of the 2008 crisis) that appears likely to get reused in upcoming AI bubble burst movies.
New Constructs takes a similar position on all mega IPOs. Sometimes they're right, but sometimes they're wrong. For example, they insisted Uber's IPO valuation was too high due to unsustainable losses, hidden liabilities, etc. Initially they were proved right when the post-IPO stock price plummeted, but in the long term Uber has performed well, become very profitable, and now trades well above its IPO valuation.
I'm expecting a similar story to play out with Anthropic: The stock price will come down significantly after the IPO, resulting in lots of headlines about the AI bubble bursting. But in the long term it'll climb up again and investors who stick around will do very well.
“Now trades well above the IPO valuation” does not mean something would have been a good investment at the IPO price given opportunity cost and the time value of money. Additionally, the current price could equally not represent good value.
Yeah. Look at SpaceX. Dropped to $108 from an Ipo price of $135, but it's currently at $162 which is... fine?
I suspect SpaceX is going to be the opposite of what will happen with Anthropic. With SpaceX I think the overwhelming majority of people didn't initially understand what Starship is, how far along it is, or what it even means in terms of its impact on the pathway of getting humans to be a multiplanetary species. So it was greatly undervalued and zoomed upward.
With Anthropic I think a lot of people haven't really realized the trend line of cost of "AI" (both in literal $ terms, and compute costs for local operation) or what open weight systems will do to the market in the "long run", where long run in this case means the next year or two. I don't think any LLM company is going to come out a winner in the longrun, because the price for LLM output is going to approach $0 with no chance for a moat whatsoever.
> the overwhelming majority of people didn't initially understand what Starship is, how far along it is, or what it even means ... So it was greatly undervalued
I don't think anyone was undervaluing the "space" segment of SpaceX. It's the clear leader in launch frequency and cost-to-orbit among the new wave of space firms, and shows no sign of losing that position.
Where the IPO raised a bunch of red flags was SpaceX suddenly acquiring Elon's AI firm, which had a significantly more nebulous basis for its valuation...
> the pathway of getting humans to be a multiplanetary species
I am afraid that you're suffering from Flavor-Aid poisoning here. We need a reality check about "multiplanetary colonization".
I've been following human spaceflight since I was born. I've actually participated in robotic spaceflight. And I'm here to say that it is robots who are destined to be a multiplanetary species. The robots have already successfully escaped our solar system and extensively explored the Moon, Mars, and elsewhere.
Many unscientific people believe that it's harder to colonize the Moon or Mars because they are farther than Africa or Rio de Janeiro is from Rome. But mere distance is not really your biggest obstacle.
I've also been following the ISS since its inception, and they have empirically disproven the feasibility of colonizing planets. If we succeed in establishing a crewed Moon base, it will be an ultra-Herculean effort, and it's on the edge of feasibility.
But Mars? Venus? Asteroids? Europa? You really think humans are going to live there?
The reasons that ISS has disproven this are manifold, but I'll point out two key points:
Their supply-chain is only about 8-48 hours, and they can stash large amounts of provisions up there. Their escape route is also only 8 hours and they're back on Earth in a jiffy. None of this is true for any body outside of Earth orbit. The Apollo missions arrived on the Moon and returned safely, by burning, jettisoning, consuming, and abandoning nearly everything they started with!!
A 15th-century expedition to Africa or Brazil means that you can fish in the ocean, breathe the air, drink rainwater along the way. An expedition to the Moon puts you in an airtight bubble with no stops, no rescue, no resupply, and most importantly, no breathing on the way, nor once you get there.
The ISS has also demonstrated the severe toll on the human physique in microgravity conditions. Watch astronauts who land or splash down. They are carted off in gurneys. They are literally unable to walk unassisted. NASA restricts returnees from driving or even jogging. The astronauts describe having to "relearn gravity" activities like brushing your teeth and showering. Any astronaut arriving on Mars would be as helpless as a toddler! Do you want to colonize Mars with toddlers? Because all the treadmills and workouts along the way won't make them fit for purpose on arrival.
So between here and Mars it's minimum six months, no escape, no rescues, no takey-backsies. We could send humans to Mars if we didn't want them to survive or return. I mean, no problem. There's no air to breathe on Mars; no water to drink; no fuel to burn; no fish to fish nor deer to hunt; no steel to build a shelter. You pack it all into the spacecraft, newbie?
But NASA and the rest of the world has amply demonstrated just how capable robots are where humans fail. Robots are going to inherit the solar system. Robots are in 2026, the de facto interplanetary species, sine qua non. My heart is warmed by these human achievements. I am happy to live and die on Earth. I think that every American should be grateful for Earth, and not permit our oligarchs to cheat and lie to us with fairytales of escaping our Earth.
Everything you’re saying is true but it’s nevertheless going to happen anyway.
I fully agree with your points: the hard, unyielding reality which most Trekkie fanatics refuse to acknowledge, except for one: "severe toll on the human physique in microgravity conditions."
That's easily fixed!
Spinning a space station for gravity is straightforward, it's just that the ISS wasn't designed for it, and NASA didn't have the budget to build a second, bigger space station with "extra engineering challenges" on top.
It's definitely practical though, it's just a matter of launching enough mass to orbit.
Like you said, Starship is going to change a lot of things, starting with space station design.
While I think SpaceX is still horribly over-valued they at least have a diverse line of businesses that are very hard to copy.
Anthropic is a one trick pony and the world is discovering lots of other cheap ponies that do the same trick.
Maybe the portion of Anthropic that produces Opus/Sonnet can be spun out for us devs at a reasonable valuation?
Keep all the Mythos/Projects/etc. hype/feature-creep crap in some other entity and value that at $5 tn or something if you wish.
Anthropic's small models are almost certainly distilled from their larger ones. And their biggest training expense are going to be experimental models that aren't even intended for release, just to test out some idea.
This reminds me of all the people asking for AWS to be spun out of Amazon.
The model is engine, the harness is the car Anthropic are selling.
They are a service company, selling home-grown, mass-produced, conveniently packaged intelligence at a well established price for consumer. Their API pricing might just be for value proposition and tax deduction purposes.
Every tech company, tech department, startup that I know is using claude code to deliver. The boutique deepseeks, qwens, glms and kimis are nice, but ain't production ready in 98% of cases. All the pi's, opencodes, etc are nice, but much less enterprise friendly. And google just outright sucks in comparison. There's a use case for each, but not at the scale Anthropic is doing.
Only real competition - OpenAI - is close, but they're burning money to stay there, which imo is unsustainable long term.
Is Anthropic worth $2T? Can't say. But boy they have a good opportunity to test this, and they may not have a similar chance in the future.
I think they have a fair chance at this.
I've tried a bunch of LLMs, and I agree that Claude is currently the best at coding, but the gap isn't huge. Gemini 3.8 Flash is surprisingly good and I'm looking forward to trying Gemini 4. The open weights models like Kimi, GLM, etc. are a little behind in quality I think but not massively.
With harnesses I see little difference between what Zed or Pi gives me and what Claude Code gives me. Enterprise features? The AI can help code those in a week.
AI itself is a great breaker of moats.
The question isn't whether they lead today. It's whether there's any kind of moat. Without a moat, frontier model development is an industry like chip fabrication or large volume manufacturing. Those industries are a red queen's race where if you stop innovating for even a short while you sink. Ask Intel what happened when they decided to coast for a few years.
Chip fabs and bulk manufacturers make a lot of money, sure, but they don't command such a multiple. Anthropic is priced like they're going to be to AI what Google was to web search or Apple is to (US at least) mobile and higher-end PCs.
I don't see it. I think they'll be a big player but that big?
Of course... maybe if they suddenly have that much cash they'll be able to carpet bomb the entire world with AI doom Rationalist ideology and buy their regulatory capture and outlaw their competition.
They made it convenient to get value for the specialist masses and I don't think anyone else managed to achieve it yet. That's their moat.
I've seen firsthand attempts at integrating open source in place of it and they all failed. Time is money and it costs premium. Touching the ground running is a huge advantage. That's why cloud services stuck.