Anthropic prices initial public offering
https://www.anthropic.com/news/anthropic-prices-initial-public-offeringAnthropic and OpenAI are the AltaVista of the AI era. With first starter advantage but no real moat and very good at spending money.
I am still waiting for the Google of this era to show its face.
Edit: every time I question these crazy valuations I get down voted really quickly.
What's your reasoning for why these two are AltaVista and not Google [of the AI era]? How would you differentiate without hindsight?
I was there. Google had a better product, and then - some time later, mind - they discovered that ads work really well for them.
OpenAI and Anthropic have a good product and they (especially Anthropic) are increasingly good at generating revenue from it.
What's missing, in your opinion?
Also, it steals IP.
The newer Chinese models get somewhat close, but OpenAI & Claude got better at stopping them from copying the model so the Chinese models will not make such big jumps anymore. And most Western companies will not use Chinese models if they are not open weight.
If they keep going like this huge parts knowledge work sector (legal, marketing, software, ads, health) economy is AI driven in 2-4 years and OpenAI & Claude will own it. Another huge shift of wealth from the Western World to Silicon Valey.
What they have right now is enough money to to work through that. But they don't have enough revenue to keep this up forever and they can't stop either.
I don't think these companies are as bad as some dot-com companies, but I personally struggle with coming up with an investment thesis at the current price given it assumes they'll retain pricing power for models. And another problem is that this isn't like Amazon or Facebook where you build a decent product and with a few tweaks every couple of years people will continue to use them for decades – these companies need to develop and release new competitive models every month or two just to retain their market position.
I think the only thesis that makes sense is that the singularity happens, the private companies that owns the models which are produced as a result retain legal ownership of them, and that humans retain control over the models. In my mind I'd give this outcome a sub 1% probability.
AMA!
You've gotta understand there's a lot of people on HN with their wealth tied to the success of AI at this point. Every downvote on a post like this is just a lowly engineer that wants to keep having steak dinners.
This is a problem of hackernews. I also made critical comments in another thread and got insta-downvoted. This may even come from bots, it is quite strange. That has not happened on old.reddit.com, for instance. Something is not working well on hackernews in general here. See also the speed limitation, after 5 comments I can not comment for some hours. That's just tedious ...
They do have moat. Every single company is pushing their private data into these models. Do you really believe that data isn't making into the training data one way or another?
Meanwhile Claude is getting really good. It's a decent harness now, even the remote control over a couple servers is quite OK. They're adding artifacts left and right, new features. And all without making the product seem bloated, they stay focused.
The prediction markets don't show anything on the horizon either: https://releaseoracle.xyz/
I am not a big AGI believer but business- and product-wise it seems like ANT has something that others are missing.
OpenAI is absolutely crushing Anthropic regarding cost of running models.
A lot of real world scenarios have already been completely solved with models released months ago. My clients don't need more powerful models anymore. They need faster and more affordable models.
OpenAI has been aggressively pushing on both fronts. Where is Anthropic's 750T/s wafer-scale compute offering? Where is this on their roadmap?
Of the people I know in tech, other than those who don’t use either, everyone is paying for Claude.
*Also consider the possibility that Anthropic may be dieselgating it down ("benchminning"?) instead of OpenAI doing benchmaxxing. It might make sense from some game theory standpoint - If you are significantly better than your enemy, it might be in your interest to lull them into believing that you are only slightly better.
Sol is half the price of Opus and Luna outperforms haiku at every eval I throw at it. Haiku wasn’t even able to complete several of my evals until the newest release.
The idea that Anthropic is miles ahead of OpenAI doesn’t pass the smell test in my usage.
And that’s before the Claude vs Codex comparison (which I’ll grant is more personal preference but I find Claude to be wretched).
On the other hand, at work, where I have to justify paying API prices, I almost exclusively use Sol, while occasionally bringing in Luna (though I've also been playing with the MS models since yesterday; not impressive, but very cheap).
If they do manage to catch ANT in ability while keeping costs this much lower, they could still leapfrog again. I'm not sure I'd bet on it being soon after seeing Astra 6.2, but I don't think the race is over.
I have one recent use case now where using Claude to review it, it first wanted to use Fable, got all excited about possible savings from lower defect rates, had to backtrack and admit that Fable provided no benefit whatsoever to justify the extra cost over Opus after it ran evals (in fact, arguably the quality was worse, but that part was a subjective measurement).
There's a mounting number of tasks like this where the value is going to be falling costs whether by switching providers, or moving down their model ladder as the one below gets good enough.
Given Opus 5.5 is now good enough for the task in my example, it means we have every incentive to start evaluating cheaper models now and again, instead of waiting eagerly for new, smarter models.
Honestly this sounds like it calls more for Deepseek v4.1 flash than OpenAI's models, although those charts from earlier this week did look impressive. Trying to compete on price efficiency sounds much more daunting than competing at frontier capabilities tbh.
who are your clients and what do they do with ai?
Assume nobody ever catches up to Anthropic and everyone will always be 1 year behind. Is the 50x valuation still reasonable? Is the money really in doing PhD level math, or simpler stuff like summerizing coporate emails?
And lets be honest, some US and Chineese labs are only months behind and the only reason they have not fully caught up is limited compute. For China this could change quickly (e.g. Huawei gets their yield issues under control and suddenly Chinese labs have cheaper compute than west). People who invested in Anthropic at 50x valuation must belive this scenario is impossible in the next 5-10 years, but that is a huge bet.
This is not a trick question or anything, I'm genuinely wondering whether stickiness (MS) or product quality/iteration speed (Ant\) would succeed.
My analogues for this market are OSes, office software, and CRMs. Windows + Office remain revenue juggernauts 20+ years after fully free and open alternatives got good enough. Salesforce is worth over $100B, even though there have been good FOSS CRMs for literally decades.
IMHO it's difficult to predict how the AI market will evolve without understanding why people continue to choose relatively expensive alternatives that functionally check the same boxes.
Even if new customer growth falls apart, if they can hold 30% growth while public it will be a good buy today.
Super intelligence threatens the marketshare of nearly all other companies.
No, and obviously not. Much more going on with this IPO than common sense investing.
It's why stock markets and corporate profits are at an all time high, while most people struggle to make ends meet and governments are woefully underfunded.
That $1.5 trillion IPO is the result of a completely lopsided economy.
In that sense I am super bullish on Microsoft as they have the vendor lock.
- around 50% in the US market
- they offer way more products, than phones, are the leader with laptops for instance
- take a big cut of every store purchase
As predictions are only worth anything, if they are made ahead of time...I predict their share price will lose 25% of its value,over the next 3 months. Drop to 40 to 45% of IPO price, after their first public quarter release.
and has hard financial commitments 25x times, the money they just raised, over the next four years,
They have about $1 trillion commitment in compute over the next 5 - 15 years. I'm not sure where you got the 25x$81b from.Fully open models will be more hobbyist, I expect. Open weights will continue as a strategy but probably not as prominently as today.
Possible, as long as everything goes well for them. Their debt deals don't care how much of a transformational company they are. If they get less demand than they expect, or slower demand than they expect, they could end up bankrupt.
The answer is that business is not just technology.
On the technical side, it's also important to see beyond the current market, what's currently possible, and broaden to - ok, many innovations are coming in AI, who is most well-positioned to be the centre of that?
The faith/hype vs objective data doesn’t support the valuation.
It is entirely dependent on having faith that this will work as expected.
Google could "Internet Explorer" Anthropic to death with a superintelligent ad stuffed free Gemini. shivers
> Google could "Internet Explorer" Anthropic to death with a superintelligent ad stuffed free Gemini. shivers
Hell, Microsoft could do the same with a cheap inferencing model; if people simply use copilot because it's already embedded into all of their applications because it's embedded into the OS (Office, Windows, anything that runs on Windows), only a few dedicated people will continue using the expensive paid-for options.
You can say the same about the Company of Scotland.
If their bussiness case is not true, we will be paying for this bubble in higher interest rates, for the next 25 years.
Now tell me: Which scenario do you think is more likely until 2030?
A - We cure cancer, create self improving models of hyper intelligence, that will solve Fusion and bring us to the stars, create a CPU, an OS and a 3D printed Car and deliver it to you within the same afternoon, solve Earth warming issues...
OR
B - This biggest credit bubble in history, will put the sovereign debit of the US and 30 year Treasury to 8%, force the US government to default, cut spending, increase taxes and bring about a recession that will take 25 years to recover from?
Maybe. Or maybe not.
It has already defined the future in some ways. Claude code has already part determined the future of software engineering, for example.
Anthropic picked the better market at least for the short term. Long term I wouldn't count OpenAI out just yet. The consumer market tends to compound into winner take all eventually given the network effects of advertising marketplaces, even more so than B2B.
I don't think there is a lot of difference between the "sota" models long term. Even if we said that say Anthropic is ahead, Google will be able to do the same in months.
But if it turns out that open weight models are the future, OAI and ANT are fucked, but Google has massive infra to host those and make money anyway. It also has the products and business connections to integrate AI into users workflows. And lets not forget, it earns massive amounts of money even without AI, so there is a higher price floor even if an AI bubble pops.
And even if it turns out that Anthropic takes over the world, Google still has a stake in them. (Although so will everyone else though SP500 by that point.)
I don't understand why would you invest in Anthropic IPO instead of GOOG, but as I said, would be interested to hear opposing thoughts.
I have an argument, but it's mostly snark.
Like, Google are definitely in the best position here, but they've messed up so many times when it comes to product management/enterprise adoption that I am confident in their ability to snatch defeat from the jaws of victory.
Down the rabbithole we go. Oh clauuuuude! :D
If possible to do, this is a vastly superior strategy to selling tokens, since token consumers can swap out token vendors. But if you simply recreate their entire product and technology lineups from scratch, faster and cheaper, and capture all their customers with superior products and prices (at least at first), then you own the market rather than selling tokens into it amongst other token vendors.
None of this is guaranteed to happen. But I do not see a way to rule it out as a possibility.
A good fraction of people are smart when the system breaks. Check out the history of the Fall of the Roman Empire. And if Dan Carlin is right, become buds with your local hells angels.
even so, what about network effects? how do they lure users away from TikTok and Instagram? AI-generated clips? or say Anthropic recreates Amazon. that's a software company, right? so how do they get all the warehouses, cloud infrastructure (they will need entirely different kinds of datacenters and hardware), relationships with third-party merchants, etc.? or say SpaceX - is your superintelligence going to be setting up massive autonomous factories and buying thousands of tons of aerospace materials and billions of dollars of machinery to make that happen? where's it get the money?
and even if this somehow does manage to accomplish all that, how politically popular will it be? what prevents the US government, lobbied by.. the majority of corporations, and hounded by constituents, from invoking the Sherman Act and busting this trust like AT&T or Carnegie Steel?
They already have solid control over logistics and inventory management.
It’s not a stretch that they then capture the means to control primary inputs and manufacturing.
Throw in an overly enthusiastic AI to continually optimise the supply chain, continually optimising a country’s supply of food and commodities, and we have all the ingredients for the classic sci-fi dystopia of a computer controlled society.
> But I do not see a way to rule it out as a possibility.
Your argument is unfalsifiable. Anything can happen by this logic and they could be valued whatever trillion $, like, maybe they’ll invent a way to travel through the multiverse. Maybe they’ll create the star trek replicator thing.
It is certainly not unfalsifiable. There are other scenarios in which this does not happen. For instance, the most obvious is that other organizations develop a similar capability at roughly the same point in time. This results in a "multipolar" economy that is mostly controlled by a small set of distinct owners of this kind of capability. The multipolarity prevents total capture and thereby prevents the rest of the scenario. Another obvious scenario is one in which the government acts soon enough to short-circuit this process before one organization gains a sufficiently decisive advantage that government intervention is no longer efficient enough to be effective. Yet another obvious scenario is one where a short-term financial crash deprives Anthropic of enough capital that this strategy is paused or slowed down enough to allow it to be countered. And yet another obvoius scenario is one where the scaling laws simply stop scaling or the amount of compute needed to scale them to this point cannot be physically obtained, thus allowing time for the strategy to be countered.
All of these alternatives would come with clear data points early on which would allow one to distinguish them from the original scenario I described. Right now, though, the data is ambiguous, and some data points lean slightly in favor of what I described, but not nearly enough to make any guarantees whatsoever.
Why is it that Anthropic gains so much from this perceived image, but then needs to public money to turn it into reality? If they were at all close to any of that, surely their internal AI can figure out how to self-finance the company asap. Or maybe that's exactly what it did and it told it to go public. Company selling shovels adage.
I think markets are way more complicated and capturing them isn't going to happen the way you're suggesting: dollars, and the public, don't really like copypcats. You would need to understand the business, the product, and manipulate people at an unprecedented level. Anthropic selling tokens as their business, means that all the way up to this "super intelligence", means that they've enabled everyone's defenses and raised the threshold for this dystopian hostile taker, i.e., it won't happen.
https://seekingalpha.com/symbol/SNDK/valuation/price-earning...
I could imagine a company owning a substantial share could do it for other reasons other than expecting Anthropic valuation to go up.
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 when it all crashes down, there will be fire sales and the vultures that are flush with cash from the exit will buy up what remains at ridiculous discounts.
We shall own nothing and rent everything, and Big Tech is obviously colluding with Big Finance to usher in neo-feudalism.
And if they dont, then we’re all screwed
The offering consists of 980,000,000 shares of Class A common stock
That 980M is more than they had previously offered (825M), so they did expand the IPO.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.
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.
Be careful out there.
Just allowing them to sell if they want doesn't seem problematic to me, unless they all (or a large number of them) actually sell immediately. Then obviously something's wrong.
Note that this is an actual question, I'm not even close to an expert with these things.
Stock price will be directly linked to having the best model (which could give OAI an advantage as they aren't competing with shareholder interests - yet)
That’s not to say there aren’t a lot of winners even 1-2 years later but I’ll happily be sitting this one out (and wait for a drop or two). The market gives some pretty sweet deals for those who are patient.
That's quite non-impressive when you remind yourself that these IPOs happened a lot of time (6+ years) ago, and that their stocks are highly volatile.
It seems to me the VC system got quite efficient to accurately compute the total addressable market well enough and inject their money to grow until that whole TAM is covered enough.
Then the VCs release the now horizontal stock to the public markets.
That being said, there are also counter examples, like Datadog (up 362%) or Crowdstrike (up 1377%), both in the same IPO class as Airbnb and Uber.
The same with Datadog, Robinhood, Facebook, Coinbase, Crowdstrike, and I’m sure I could go on. Even ARM was generally flat for six months after IPO… the only real winner I’ve seen recently was Reddit (which dropped a bit but was briefly flat before going up).
My point is buying the IPO instead of waiting will probably have you overpaying for it in the short term (1-2 years), long term it doesn’t matter (if the stock/company is making money). Most big tech IPOs have a sell-off, sometime within the first year or so, and it has been nothing but beneficial for me to wait.
I absolutely want to buy Anthropic’s stock, I’m just willing to wager I’ll get it at a hefty discount sometime in the next year or so… from what I’ve seen their revenue is nowhere near their valuation and we’ll probably see the market recognize that (even briefly).
IPOs statisticly do mostly bad within first 12 - 18 month of trading; this is mainly due to institutionals & insiders selling their stuff.
Neither structure gives ordinary public shareholders control proportional to the money they invest. This should probably be made illegal.
I would prefer it if people were still allowed to invest in such companies in the public markets, but perhaps we can make it so that inclusion into an index requires "1 share = 1 vote". That way, at least retirement funds are protected.
I'm getting 2008 vibes.
However… when America sneezes, europe catches a cold.
The biggest problem is that poor economic management (in the UK, austerity under Osborne and the Covid response) has eaten our margin. We are not starting from a position of strength.
Don’t worry. It will.
The specific models and capabilities are a sideshow.
And hence, the money flows all to the top despite viable alternatives...
(a separate market of workloads is contracted out to Fireworks, Baseten et al. and a third segment DIY it with rented or owned GPUs. But those customers are different buyer archetypes little overlap with those buying trust.)
The reason I think Anthropic's position isn't that strong is that most of the money they raised will go into making other companies very rich. Because they don't actually own a lot of the infrastructure they need. OpenAI has similar issues. The companies that actually provide the data centers (i.e. hyperscalers), chips, gas turbines, etc. Will be on the receiving end. And of course some of those hyperscalers have their own ambitions on tapping into the same markets. The SpaceX pivot to data centers in recent years is particularly interesting. Anthropic is already paying them a lot.
The reason I'm fairly optimistic overall about the economy is all the spending that will result from these investments in all that physical stuff represents an enormous cash injection in the economy. The investment bubble may or may not burst but it won't wipe out the infrastructure. If Anthropic collapses, that infrastructure will just be gobbled up by whomever takes over.
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.
Who are the pedantic, authority-appealing users who are using this model?
What makes you believe they’re norming their training data to that group?