There is a lot of value from LLMs but it is something that I can switch easily. Yes, some frontier models are much better but users can still chose to use cheaper models for most of the work. In fact most applications, through which people will interact with LLMs anyway, will not need frontier models unless they offer "chat with AI". Anything that solves specific problems (even multiple problems) does not need frontier LLMs.<p>Anthropic and OpenAI will have to become more of the end user application stack. Files, calendars, emails, photos, and much more have to pass through them for them to keep making money. Else they cannot make money like the way they have been in this early era of LLMs.
Switching models is easy as a solo developer. Switching LLM providers as a company is not, once all your employees have it integrated with GitHub, their IDE, etc. I think Anthropic will look a lot more like a cloud provider like AWS or GCP. Same-looking product, very high switching cost that would have to be driven at the CIO-level because of all the integrations that have to be redone, IP concerns etc. And ultimately they are selling compute capacity.<p>Edit: I should clarify that I don’t think this justifies their valuation. It puts them in competition with much more entrenched players with established customer bases, and who aren’t in massive debt to provide their compute. I’d also like to distinguish between a model provider as someone who provides weights vs a model provider who connects their GPUs and models to your codebase and internal systems - I think the latter is where the sustainable business is.<p>Incidentally this is why I still feel Google is well-positioned in the long-term AI race. If they can remove their head from their ahh and make a coherent product offering across Workspace, GCP and Gemini, they could become as entrenched as MS.<p>I would assume AWS is probably the most interested in acquiring Anthropic out of the three major cloud providers.
It's trivial to switch. The company I work for has 1k employees has already switched like 3 times in the last two years.
We have like a dozen different LLM tools available to us and it's as easy as clicking on a dropdown to switch in every single one of them, and we have several thousand employees. Self-hosting with the cheaper OSS models is also becoming more and more attractive for larger orgs.
That's just inference. Claude.ai is a platform that enterprise IT teams connect up to all kinds of other applications, and gets use beyond the claude code CLI. Non-developers get used to this functionality - whether it's their form-building stuff, artifacts, slides/presentations, etc... they're building an offering far beyond just inference. I agree the latter is a commodity now, though companies will compete on privacy options (ZDR, guaranteed confidentiality, etc)
Similar at my $(dayjob) and can confirm.
> Switching LLM providers as a company is not, once all your employees have it integrated with GitHub, their IDE, etc.<p>I'm not so sure about this. I guess it depends on how well people did their job there.<p>Ideally, whatever people built is loosely coupled to that third-party resource, meaning that switching cost will stay low.<p>After all, it's probably just prompts + an API layer. Unless one micro-optimizes their prompts for claude (which then breaks once a new checkpoint drops, hence is a bad idea regardless), pointing them towards a different API endpoint should be doable.
Currently we are building agentic systems for various business functions like Legal and Finance. How these models perform in those tasks is paramount, and things like cost per run is only a part of the equation. It can become very subjective, and things like brand recognition comes into play. The cost of a counsel or a VP finance overshadows the savings you get by micro-optimizing for a cheaper model. We tend to use the best-in-breed LLM for the various tasks - mainly Claude and Gemini, and sometimes GPT.
> How these models perform in those tasks is paramount,<p>Then you have no choice but to use your own hosted model because that's the only way you can be sure the model isn't going to be removed or quantized/nerfed behind your back.<p>If your system can widstand the fluctuations of closed models over time, then it can most likely support switching between multiple providers in the first place.
Ur case actually strength the argument for open weight self hosted model. You would have precision control over not only the technical issues, but also the legal issues.<p>Once the open weight models are good and cheap enough, there is not reason not to use them
Spending money in the most effective places is smart... and then there are the stupid businesses.
> <i>I guess it depends on how well people did their job there.</i><p>Given that the goal of the whole project is outsourcing the work, what are you expecting here?
Working for a 10k+ corporate and we made it pretty easy to switch to any provider, people here can use Anthropic's, OpenAI's, and our own self-hosted open weight models very easily. It's one drop-down or `/` command away to switch between them.<p>I think other companies are probably doing the same, the moat of a AWS/GCP/Azure is much deeper with how contrived it is to migrate between them even if you abstract as much as possible with something like Kubernetes.
I think you'd be surprised. I've seen orgs entirely switch fundamental tech choices to save 10%.<p>Once they realise that you can use Deepseek / GLM for 10% of the cost these labs are done.
With some gateway inbetween managing the model responses, MCP and skills (see <a href="https://modelcontextprotocol.io/seps/2640-skills-extension" rel="nofollow">https://modelcontextprotocol.io/seps/2640-skills-extension</a>) this becomes very easy
Depending on how the company leverages them, it's super easy.<p>For us it's as easy as clicking a dropdown. The copilot vs code extension allows for this ez pz, the same is true for companies that implement an AI gateway for other tooling.
VS code or other ides with multiple providers makes it pretty easy.<p>Also for larger companies like Amazon the danger is they replace Anthropic completely with self-hosted models.
Integrations with Github and IDE are nothing. It's like changing your email software.
Switching LLM providers as a company is even easier than a solo developer, most large companies I know are using AWS Bedrock and selecting another model is basically just a checkbox for them
As far as I know, our corporate has deals with all of them: Anthropic, OpenAI and GitHub Copilot.<p>They stroke a great discount deal with all, so different teams/departments have different provider, and those using Copilot are even free to switch between models as they please (although with lower usage limits).<p>You could make the argument that it's not worth switching SAAS providers for enterprises when the costs were like 10 dollars a month per developer and switchign would require 1 team to work for half a year... but with AI providers it's more like paying 1000 dollars a month per developer vs paying 200 dollars if you strike a good deal with different provider and swtiching costs are more like 4 to 8 hours of work for each developer.
I guess what I’m saying is that means AWS is their provider then
What?<p>Have you used Copilot at a company?<p>I can literally change my model in 3 seconds.<p>The moat is as shallow as it gets. Investing in a single LLM provider at the moment would be a mistake.
I've observed that there are some companies and some folks that are very fanatical about a specific provider. "Only Claude" is very common. "Only GPT" less common, but still see it.<p>A lot of the non-technical and semi-technical folks are only comfortable using a specific client like the Claude desktop app. A lot of teams wrote skills and eval'd/benchmarked their skills specific to a model family and harness.<p>The "I can use any model, any harness" or "I only use a meta harness" (Orca, Paseo, Herdr, etc.) crew is relatively small IRL. Most people are still single-session, single platform user mode.<p>Most orgs have an enterprise contract with a single provider so all of their internal assets are built for the provider. It goes beyond skills, but also the telemetry, enterprise policies, and managed configuration.
For a company though, the idea that a few disgruntled employees will shift massive purchasing decisions is a bit of a stretch.<p>We have a developer on our team that keeps asking for a claude code enterprise license, we keep telling them "no, it's way too expensive, get used to using the copilot extension/cli".<p>For consumers, sure, but the bulk of the revenue opportunity is enterprise. Anthropic/OpenAI have a hard row to hoe IMO.
As an only-clauder, there's another elephant in the room. Anthropic is by no means perfect, but:<p>* using OpenAI supports the Altman/Thiel/Musk/Ellison/Karp/Trump billionaire Orwellian worldview and also ads<p>* using Grok supports a ketamine-addled open fascist sympathizer jizz-spewing 12-year-old-edginess memelord and his similarly broken worldview<p>* Google's models are pretty good but every single product they have ever had is ultimately either suborned or killed by the advertising division<p>* local models pretty much only half-exist because Jensen Huang wanted to cheat on his wife; future is unclear<p>One can argue for harnesses and so forth being lock-in, but the ultimate lock-in is morality and wanting to have a civilization at the end of all of this; I can't recommend either enough
A gateway like GoModel can make switching transparent, along with the many other benefits you get with a gateway.
It’s literally a menu in Copilot.<p>The model is just some REST endpoint you switch out even in a lot of enterprise setups.
I don't know why nobody has said this yet, but...<p>The eventual logical move is that they build a completely reinvented AI-native operating system.<p>The funding they're getting gives them the resources to build and compete on a fundamental level, IMHO.<p>If you don't believe me, look at the recent demo where they had Fable 5.2 build an OS from scratch.<p>They are coming for the whole stack, my opinion.
a big move into end-user software would mean that they have to compete heavily with their own largest customers. The economics of offering SaaS are also under meaningful pressure and don't warrant anywhere near the valuation that is being suggested
The biggest moat with applications like MS Office is compatibility. Sharing a slide deck or spreadsheet with others is a simple decision, you know they have it or have software that can read/write it.<p>I don't get the need to have Anthropic build their custom versions of MS Office apps, I can tell Claude to generate a word doc from a wiki page and it does a pretty damn good job. Anthropic's statistical median level of software that they had Claude create isn't going to replace Office apps to the point where companies cancel/stop renewing their MS Office subscriptions/licenses.<p>The other element is trust. Do end users trust their data in Anthropic's systems? A lot of people get creeped out when they search for content from one device in their household and then nearly instantaneously they start seeing ads from other devices on the same topic. How creepy will AI get if it starts consuming all your docs, spreadsheets, personal finance info, emails, etc.?
I beg to differ I already am stopping using MS office recently but that's mainly because I really dislike how they lock up the apps if your late on your annual payment. It's a grating value capture model. I'm sure I will have to get back to it at some point but in the meantime google is providing great optionality and I don't have a concern that the software stack has glaring security concerns.<p>On the trust thing - I don't trust anthropic and I don't think MS has been a great steward over the years either.
The moat of backwards compatibility and plugins is why MS Office has survived decades of competitors without any appreciable loss of market share.<p>It's hard to see Anthropic replacing it, and even if so - why bother, the revenue is miniscule. They need 100 MS Offices for the compute contracts to pencil out.
End-user software is going to be under a lot of pressure. If you're Microsoft, Apple, or Google: you can easily copy products or integrate ideas into their platform. I see companies like Adobe, Salesforce, and Oracle having to downsize significantly to stay profitable. Having said that, my in-laws still pay for an AOL email so I don't think their cashflow is going to zero soon.
Adobe seems to be in a tough spot, because so much of the lock-in is client side, I think Salesforce is fairly well protected, it's just so impossibly difficult to remove once a company has fully adopted it. Similar to the staying power of the ERPs of Oracle and SAP.
It's the same really. There have always been alternatives that can work with Adobe's various file formats. Even more aren't going to change much, at least for their enterprise key customers. This is not only the case for Adobe, but many others as well, e.g. MS office is still a thing these days, and so are things like 3Ds Max or Maya, albeit having many compatible cheaper or even free alternatives. It's not so much about features, pricing or even stability. It's about customer support. Do I want to ask an LLM and test my luck or write a GitHub issue and cross my fingers that someone will read it within the next weeks? No, I will pick up my phone or write a mail and get a response within hours if not minutes.
I ask my llm to write custom software and love the freedom to customize it to my needs. Today I do a presentation on Ubuntu with a new "Presenter" app that works offline, runs 100% in CUDA on GPU and streams videos, starts in 2 seconds, lets me demo live robotics policies in 3D simulations and Opus 5.5 wrote it yesterday. This may sound very unusual, but I think it will be common in the future: moats disappear one by one.
I would say that Apple's Create Studio can move toward feature parity faster with Adobe. My point if you're not a platform provider it's going to be harder to make a profit.
The risk for Adobe isn’t that we replace their software with some AI generated alternative. It’s that we bypass that class of software altogether and have an AI do the work.
"There have always been alternatives that can work with Adobe's various file formats. "<p>And yet me and quite some others still yearn for a flash alternative. There never was one.
The big problem is Photoshop. I have no love for Adobe, I proudly moved my department off of all Adobe products a few years ago. I am a big fan of open source solutions and I am always impressed at what the GIMP team (and some others) have accomplished.<p>But…<p>Photoshop is king. Everything else is behind by a solid decade. If you’re a professional the $40 a month or whatever it is now means nothing to you as well. It has the professionals are all on it, and because the price isn’t astronomical, that’s what everyone else uses who needs to use that kind of program more than once in a blue moon.<p>Premier? Replaceable. Lightroom? Pretty replaceable though still somewhat ahead. Photoshop? There’s only one Photoshop. And as much as I’m loathe to give Adobe any credit, it’s the best by a mile. There are too many essential features in QoL things that other programs are missing.<p>Luckily my department doesn’t need it, but if we did, we would simply have to accept Photoshop.
Not a graphic designer, so, truly asking out of curiosity: what are the QoL things and essential features in Photoshop? I've been slumming it with GIMP for years now and not felt anything very obviously missing, but again I'm not a pro.
Compatibility issues, which can be a huge deal partially due to near-vendor lock in. If you historically been on Photoshop, bringing your working/former projects into another software is not easy, which matters because revisiting work is common.<p>Not all tools are non-destructive in GIMP. Things like raw support, smart objects, retouching, etc. all either are super thin compared to Photoshop or require workarounds/a lot of manual effort. There are just lots of tools you expect in Photoshop that either don’t exist or exist in an inferior form. I can’t elaborate too much because I don’t really touch Photoshop much these days, but I work with people who do and worked in production long enough to know the expectations. GIMP is great but Adobe is a massive company with deep pockets, it’s hard to keep up. Their user experience is miserable but the tool itself is just far more robust, and a client simply isn’t going to accept that you won’t do something that Photoshop could do quickly because your alternate program can’t do it or does it with a lot more effort. That’s the reason I ultimately dropped Final Cut!
AI is pretty amazing for migrating. source: I'm using Claude to replace a large Tableau ecosystem. What would have taken years is now taking months.
Adobe creative suite is $70/m. From most businesses perspective it's not something worth looking for cheaper alternatives and the alternatives has existed for a long time now.
> If you're Microsoft, Apple, or Google: you can easily copy products or integrate ideas into their platform<p>This argument doesn't hold water. Google tried like no one else to have a chat app and it failed miserably more than a handful of times. Same for social networks.
now imagine that said customer/competitors have been uploading/generating their source code, problem reports, marketing collateral, and other IP _to the LLM company_ for the last few years.
People seem very unimaginative about the ways in which somebody equipped with a lot of intelligence and compute, and control over both, might make money.<p>The recent Anthropic push into biomedical research gives a clue.
This is equally true for how risky it is to build such a business on top of a 3rd party black box. Model nerfing is widely speculated and the only true way to guarantee that the models you're using are stable and consistent is to run your own hardware. Not to mention the recent Navier-Stokes controversy.<p>The moment you find a money-making formula on top of someone else's black box model, there's nothing stopping them from taking it for themselves.
Yes, they know that. That’s why Claude is becoming a platform like Google Workspace.<p>They need to leverage Claude Code in the software industry to position themselves as the go-to platform for businesses.<p>Meanwhile, Google and Microsoft, which already dominate that area, seem to be dormant.
If Google and Microsoft have gone dormant in this space, I have to think it’s because they haven’t been seeing any ROI, either. Possibly because customers have been so underwhelmed.<p>I’ve been using Office Copilot more or less since it was released, and the general opinion around the office is that it’s a huge waste of time. The auto-transcripts produce 3 pages of fluff when previously someone would send 3 sentences’ worth of what actually mattered. The auto-drafted documents rarely add anything meaningful beyond what was already in the prompt. The question answering and research tools are only as good as Sharepoint/Confluence/Jira/etc’s pre-existing search feature, and therefore untrustworthy, etc.<p>What made Anthropic exciting was they had locked onto a use case that reliably generates positive value: coding. Not much else has proven, from a business perspective, to be anything more than an exceptionally buzzworthy money pit.
Yes, that's the direction I've been witnessing for Claude too. But it's a short to middle-term strategy IMO, by the time companies are ready to drop all these apps they will need something else entirely, not Claude's slides or spreadsheets or Salesforce.<p>The fact that Frontier AI can become Google Workspace at a flip of a switch is what is crazy valuable. Anthropic and OpenAI can enter and disrupt any market, especially software, faster than any company ever could, because users want the flow to start and end in their chat interfaces.
I think eventually almost all non-creative software will, essentially, be a wrapper around a session.<p>This pattern has already happened before, in the translation industry. Someone figured out how to learn from and recall context from ordinary translation work (the "translation memory"). It started as an add-on to Microsoft Word, and matured into a standalone software, that was ultimately a wrapper around a document editor. Then language management systems plugged into that wrapper, hey presto, every translator at your company does the same work they were always doing, and you've produced an entire corpus of quality translations to reuse everywhere.<p>Now _translate_ that idea over. Every time any human works with an agent on anything, simple heuristics find everything true and reusable, save it to the everything-memory as skills, refences, glossaries.<p>Anyone does any work, all agents can now that work at that organisation. Self-fine-tuning models everywhere, and it fits into corporate intellectual property expectations.<p>I have done something like this at a medium-sized listed business.
> Meanwhile, Google and Microsoft, which already dominate that area, seem to be dormant.<p>Google owns 15% of Anthropic I think.<p>And Google offers Argon 4 to paying Workspace and Google One customers now. How's what was the number 1 model on nearly every single bench a few weeks ago being "dormant"?<p>I know people all in in the Google ecosystem (getting credits from Android phones they buy etc.) who only swear by the Google models and who are now using Argon 4: I asked them and a few days ago they had access (but they were already using other Google models before that).<p>Doesn't GMail have <i>billions</i> account by now?<p>I'd tend to say with billions of users on GMail and a model like Argon <i>and</i> 15% of the Anthropic shares, Google is not where you think it is.
[dead]
Why haven't people switched then? Same could be said about Google search, Facebook, X etc. Claude is the de-facto SWE model now, yes you could use an open model but you would always be wondering if Claude would do better, like using Duck Duck or some other Search engine
> Claude is the de-facto SWE model now, yes you could use an open model but you would always be wondering if Claude would do better,<p>In my testing (incredibly similar workflows, moving around between claude code and cursor), I haven't found that much difference between Sonnet and GLM-5.3/flash.<p>Sonnet seems lazier, but has better prompt adherence, and is constantly telling me it can't do things, while GLM will just Find A Way to do whatever I've asked for.<p>I really appreciate being able to see GLM's thinking tokens, as I can head it off quicker. Claude's a little harder to steer, but the integration in Claude code makes it much easier to leave it running on things overnight, which is hard to get working properly in Cursor.<p>Overall, I'm not sure Anthropic have that much of a moat, in coding particularly. This seems weird to me, as they have so much more usage data that's hard to replicate.
True. We will end up getting more Google alternatives. Nothing beyond that for now at least.
This is the same argument we have heard many times with SaaS that would presumably get replaced by open source equivalents. In practice it does not happen because there is a lot of value in having someone else deal with a problem.
SaaS have vendor lock-in, they have your data hostage and whar importantly your habits: other software never gonna be 100% the same.<p>Changing LLM backend on other hand takes 2 seconds.<p>And it's not all or nothing: you can just keep anthropic $20 plan for planning and do 99% code generation with chinese model.
That "someone else" is just shifting from a single open source maintainer to an LLM.
100% - brace yourselves for enshittificarion.<p>I’ll always remember the good times.<p>But you’re wrong ser about small or cheaper models. Choosing to work or not having a choice to work with cheaper models means you work slower and more methodically because you have to plan out the work and size it to fit. This has huge implications on the value of the work you can produce as an employee or solo developer, and it’s a skill working with these SOTA models - there is specific crafting that must be done to be effective with each and their capabilities.<p>You will see the old eyeball real estate game, forcing more of your experience into their apps and platform so you can see ads or “new slop features”. You will see the SOTA models more closely guarded in an effort to bait you into that platform, and yes, you will see them doing things they never should be doing, moving towards an open AI like model where it can order a pizza, where monetization it was always important, but is now the golden rule, before all else.<p>I hope open models can at least give me opus 4.8 experience soon, I can work with that. But nothing out there today can do it and that terrifies me.
The frontier is addictive.
> Yes, some frontier models are much better but users can still chose to use cheaper models for most of the work. In fact most applications, through which people will interact with LLMs anyway, will not need frontier models unless they offer "chat with AI"<p>I am finding some really routine tasks to be borderline with frontier models. Performing E2E testing of enterprise web apps would clearly disintegrate with local models. You need that reasoning black box on top. I don't care what your favorite benchmark says. I care about what <i>actually fucking works</i>.<p>The people who don't understand the economic situation are those who are in the pure play space with AI and multi agent circuses. Why bother spending any money when the occasional failure has literally no consequences? Who in their right mind would <i>pre-pay</i> for tokens?! I wish I could see the statistics on what % of HN uses AI for fun vs for work. It is clearly a kind of Factorio entertainment experience for a lot of people. If I was just using it for entertainment this whole IPO business would look really silly to me too.
We all love moats, but I mean, it is still possible to make money in an industry simply by being the best at what you do and charging for it.<p>Anthropic has shown a consistent ability to deliver the best general purpose models. Can that continue indefinitely? Will models become "good enough"? Will open weight models catch up to SOTA models? Will some genius start a new company with better models?<p>And, on Anthropic's side, will they be able to use their market power that comes from being the best to push through regulations that will create a moat; instituting some sort of mandatory safety checks that coincidentally are exactly what Claude is trained to pass, regulating the amount of compute or trying to claw back data centers to prevent new training, etc.<p>Or maybe they just have the magic sauce and they're going to stay on top of the game and people (like me) will be happy to continue paying money for their product.
If they actually crack AGI, they're going to sell the results of using it, and not model access themselves.
Easily in the context of the corporate world is to pay for something, that something gets installed and then just works, and if it doesn’t work there’s always someone you can call 24/7 to come and fix it. I don’t see how this is possible with open models.
Anthropic has basically no support.
the answer is to use a gateway that all the prompts go through. There are many proprietary and open source solutions for this. You get the benefit of making it trivial to switch between proprietary providers in case you are unhappy with the performance or squeezed on price. The AI Gateways put cap on LLM providers' margin upside.<p>Additionally, the gateway allows for caching, additional security checks, centralized monitoring. I think most corporations have a multi-provider setup anyway.
What do you mean? We use Bedrock at our company and I can switch to literally <i>any</i> model, including the open ones, by just typing /model and picking it, even including inside Claude Code. It's ridiculously easy to switch, and imho GLM 5.3 is good enough for 99% of my daily tasks already.
You pay a service provider, who manages the open models… there is absolutely nothing special about it. The hardware can be something you own. Or you rent. Or you just pay the service provider to access the open model apis. Lots of options here, it’s not different from pretty much any other software service in the corporate world
Isn't that exactly how Linux companies are making money?
Good luck in calling Anthropic support...
It feels like people are still trying to bet on the horse that will achieve gAI.
I'm fascinated by exactly how an investment in the company that first reaches AGI (or whatever we decide to call it) might pay off... It feels like the stocks would absolutely be worth owning in the immediate term because they'd be hot property at the point where people realise it's happening.<p>The thing that I'm less clear on is, <i>if</i> and when they achieve some kind of recursive improvement exponential growth thingy, isn't that going to heavily disrupt the entire economy / stock market? In that environment, would Anthropic's stocks really be a good bet, given they have to operate <i>within</i> the economy?
> The thing that I'm less clear on is, if and when they achieve some kind of recursive improvement exponential growth thingy, isn't that going to heavily disrupt the entire economy / stock market? In that environment, would Anthropic's stocks really be a good bet, given they have to operate within the economy?<p>That's a long-term problem to be dealt with later, greed dictates to pursue the former as long as the music is playing and everyone is dancing. Similar to 2007-2008 when it was clear that things were breaking but the incentives were to play the game for as long as possible before it imploded.
>> that first reaches AGI (or whatever we decide to call it) might pay off<p>It is a myth. They are chasing science fiction. They either dream of an omniscient overseer who will take control of society and nominate its shareholders-parents into the governing elite, or they envision a chatbot that will solve every problem, again turning shareholders into the power brokers. They forget that nobody will ever be allowed to "win capitalism" and take over the planet. Companies have are torn down for far far less. Every AI model is data, ones and zeros. It therefore can be copied/shared/leaked. Intellectual Property law is not a ladder for assent to lordship. It can and will be rewritten or outright ignored rather than allow capitalism to collapse.
“Yes, the planet got destroyed, but for a beautiful moment in time we created a lot of value for shareholders.”
> It feels like people are still trying to bet on the horse that will achieve gAI.<p>I think that at this point gAI is just a marketing blurb, and the actual race is to dominate current practical applications involving basic RAG and ReAct loops. This can go from anything between software development, business intelligence, and market research. A company whose top brass can use a team of agents to do what nowadays they assign to mid- and top-level management is an executive with tighter control over the whole business process. For this you don't need gAI.
I've noticed they've actually stopped harping on about this AGI nonsense, I suppose because no one was buying it to begin with and they themselves didn't really have any idea wtf they were talking about when talking about "AGI". Now it's all about a bajillion agents running autonomously or whatever
> Anthropic and OpenAI will have to become more of the end user application stack. Files, calendars, emails, photos, and much more have to pass through them for them to keep making money. Else they cannot make money like the way they have been in this early era of LLMs.<p>...so, Google?
I believe the models have been effectively commoditized. It's believed by many that the lead of the frontier models over open weight models is only 6-9 months. That was one of the pillars holding up the SpaceX/Anthropic/OpenAI valuations.<p>The other pillar is hardware but I think that's only going to last a few years at most.<p>These trillion dollar valuations were based on the idea of "winning" and "owning" AI (IMHO) and the open weight models have killed that outcome. Now it's a rush to cash out before it all comes tumbling back to Earth.
>something that I can switch easily<p>something I find even easier is to never pay anything to the creators of the most inhumane, antagonistic skynet-tier chatbot ever<p>¯\_(ツ)_/¯
When people want to argue that AI is a commodity, they confuse two concepts<p>1. switching costs<p>2. fungibility<p>I do agree that LLMs have low switching costs. But they are not fungible - each model is unique with different strengths and weaknesses.<p>I also disagree that they can increase switching costs by moving in application stack.<p>For instance, oil and electricty are commodiites because they are fungible, not because switching cost is high.<p>Whereas clothing has low switching cost but not fungible
Not only can you switch easily, they all seem to do the same thing, fail in the same ways, make the same bad linguistic choices, write the same code... I think at the end of the day, LLMs are a commodity. They're like a grease, or oil. They're trying all the tricks with branding cute names, version numbers, benchmarks, thinking levels, but it all just seems the same stuff to me: I put in tokens, I get more tokens back, and if I didn't know ahead I wouldn't be able to distinguish if they'd come from Anthropic or OpenAI or wherever. Maybe some people can but I can't.
A few important points from the linked documents:<p>1. Price range: $70–78 with midpoint at around $74.00<p>2. Gross deal size: $85.3B (IPO + $26B private placement)<p>3. Implied market cap: ~$968B (at $74, 13.09B shares)<p>4. Q3'26 revenue: $17.0B (+1,127% YoY)<p>5. Run-rate revenue: ~$81B (as of Oct 2026)<p>6. Compute commitments: $900.9B (through 2045)<p>7. Ticker: ANT (Nasdaq)<p>8. Customers with ≥$100K, ≥$1M and ≥$10M TTM revenue: 11,900+, 1,900+ and ~250.<p>9. Customers under $100K: ~40 million<p>10. Fortune 500 penetration: over 50%<p>11. Global 2000 penetration: over 30%
Didn't the consensus expect an initial market cap considerably higher than that?
> Customers under $100K: ~40 million<p>All claude code subscribers are classified in this category it seems.<p>Total customers >1M trailing twelve month revenue: ~2150. I suspect a vast majority of their revenue comes from limited list of customers with >100M revenue.
What about costs?
Here's what I found most interesting in the Customer Stories:<p>> 4.3 Non-billable hours recaptured per lawyer weekly, in an independent study of 31 firms<p>-Legora<p>> 24→5 days Working days for new feature time to market<p>-Rakuten<p>> 2x Quarter over quarter productivity gains (estimated 8.1 million hours annualized)<p>-Salesforce<p>The rest are pretty fuzzy/fungible (percentage of employees "using" Claude etc.). But these claims potentially show real, measurable productivity value (assuming they're true and measured accurately).
My vantage point as a technical seller at the margins of this space: Outside of software development, the productivity argument is a challenge to CFOs / Finance / Procurement. The ship has sailed when it comes to friction free AI purchases. Buyers need to articulate why productivity will adjust the trajectory of top line or bottom line metrics for the business.
> (assuming they're true and measured accurately).<p>Precisely, because TTM doesn't tell the full story about the potential shifts in quality of the work. I don't refer to vibeslop here, to be precise, just the additional followup work required on those features released faster, which <i>always</i> happens but with AI it may or not happen more often.
TTM for sure is a bad indicator. Earlier this year at $DAYJOB, they attempted to switch all of the devs to a fully LLM based workflow: LLM writes the ticket, gets stakeholder input to refine the ticket, reads the ticket, does the work, submits the PR, etc. Their case study was a single project that did that process soup to nuts. The estimated time was a couple months, but they ended up going about, and I quote, "2x-3x" faster. So obviously they expected everyone to go faster.<p>The problem is that the way it was described to the other devs, they were able to sit all of the stakeholders down to really flesh out the idea and produce a working and detailed description of what needs to be done. A process that the company struggles to get done in general. I'm also convinced myself that they did not include the time to get the specifications written, a time we typically include in our TTM.<p>So of course the example's TTM was shorter, but that's always true when you start with detailed specifications. I can't speak to quality of the feature, since they hadn't even finished rolling it out to customers by the time they told us that's how we're operating. With promises of 2x-3x more features. And we continue to struggle to get stakeholders to nail down specifications and struggle to get features complete as a result. In fact, now it seems even worse than ever.
A flat line from their last raise 6 months back. Are they panicking today for some reason?<p><pre><code> Anthropic Valuation ($B, Basic)
------------------------------------------
Ser E Mar '25 | # $61.5B
Ser F Sep '25 | #### $183.0B
Ser G Feb '26 | ######## $380.0B
Ser H May '26 | ################## $965.0B
IPO Nov '26 | ################## $968.7B</code></pre>
„The IPO process begins with a roadshow starting on November 10th (EST). Subject to regulatory clearance, the final IPO share price is expected to be set on November 18th (EST), and shares are expected to begin publicly trading on November 19th (EST).“
The economics of this are absurd, and Anthropic hasn't been without their controversy; but I use Claude everyday, and they've made tremendous impact on society. I wish them the best for their IPO.
They’ve had a tremendous impact on software development, which is not exactly society.<p>For society more broadly, what I’ve seen is a bunch of white collar workers under threat (real or perceived) of losing their livelihood, and a huge uptick in the amount of interference I’m having to run as my family members get taken in by a wave of increasingly sophisticated scams and phishing attacks.
Social media had a huge impact on society. AI in comparison has had very little impact, with a couple of exceptions. I'm sure it will have a impact at some point in the future though.
> they've made tremendous impact on society<p>Such as?<p>I can think of a bunch of stuff they released, but none that have impacted the world. I can think of some partly positive impact, which is that my bar for bullshit has to go up because words are cheap now, and so is ""graphics"".<p>Some other fun impacts they've made is irreversible environmental damage, beautifully detailed in their IPO. Fossil fuel reliance and and water scarcity in the big '26 is probably the most "fuck you" move you could've done, and zero sustainability plans. at least Cloudflare tries to pretend theyre sustainable (and frankly they do invest in green energy).
The only "tremendous impact on society" I have seen is that people generally outsource thinking completely and get dumber.
I personally hope Claude and Anthropic can continue to exist. Their model has unlocked so much for me.
The way I see it, AI is more comparable to some sort of utility. Like electricity, internet, etc. - and should be priced that way. If we ever get to "true" AGI, and the open source models also get there, what's the moat? Access to compute could be the differentiator.
> Access to compute could be the differentiator.<p>Already <i>is</i> the differentiator even with the models we have now which while having differences are mostly swappable for a lot of users/use cases.<p>For all that the local models have improved (and Qwen 3.8 27b is still impressing me) they aren't what you can run (or would run) if you have an entire data centre stuffed with H200's.<p>All this does is further widen the divide we already know exists by those with access to resources using them to increase their access to resources and those without getting to watch because there isn't much they can do without resources.
There are a lot of things that don't seem to have much moat like anyone can make a search engine, anyone can flip hamburgers but the likes of Google or McDonalds dominate by economies of scale and being better in some ways.
I was of the same opinion but the main difference is that models get better which I am not sure you have a correspondent in utilities (what would get better?). The barrier of entry in utilities is capital. Here there seems to be more. Why is Grok behind and why is Gemini so ... what's the word? If models progress reach the RSI point there is no parallel to utilities imo.
This is pretty much false. There's a lot of art and technique that goes into creating a model. A model is also clearly not fungible - it has certain characteristics and behaves a certain way. It is fungible in the same way humans are fungible.
Less an utility and more like foodstuff.<p>You have your boiled potatoes/rice/beans (small local models), steak (Opus 5.5), wagyu (fable)
Lets imagine that AI is very valuable and compute gets increasingly scarce in the next years. Doesn't the fact that Anthropic "reserved" a lot of this compute already have value too? Would this not push the prices of open weight models because they compete for a smaller slice of datacenter capacity?
Apparently not every country can participate in this IPO. As of now, if you open the Revolut Anthropic IPO page from Poland, you get:<p>> Linked page(s) related to Anthropic IPO can only be viewed by persons located in UK, Denmark, Finland, France, Germany, Ireland, the Netherlands, Norway, Spain, and Sweden.
Guessing, but probably where the underwriters have brokerage relations or there are favorable capital controls, since it’s a US listing.<p>I confess I’m about 17 years out of practice in the Polish market, and about 12 years out of touch with how the European industry shakes out on the whole, so I’m not entirely familiar with the legal or business structures.
I don't know about the IPO specifically but it's only retail that has limitations for some US listings. And custodian banks/brokers offering products to retail.<p>The biggest one in the EU being the "No KID, no deal" (for funds, derivatives, etc.): no Key Information Document, then cannot be offered to retail.<p>That is unless retail is an accredited investor (say a letter from a bank saying you have more than 1m EUR in investments): then it can trade/invest in everything.<p>Or for example if you own a company (say private weath management) that's non-custodian (say you manage your wealth using your own company, and shares of your company are privately owned) and you're considered a professional in the finance world, you can ask your broker to let you trade / invest in basically every single product under the sun.<p>AFAIK pre-IPO stocks are just as vanilla as regular stocks and do not need the KID notice.<p>In addition to that Anthropic (in the EU), says you can enter pre-IPO using Revolut, Trade Republic, eToro or using bigger players like Saxo or IBKR.<p>Dunno why GP doesn't have access to it in Poland.
Appreciate the clarification!
Availability of American investment products is the outcome of compliance required (KID in local language) then volume and sophistication of investors. In case of paltry investors like e.g. Poland, Czechia, Hungary, they don't bother.
The UK page is effectively empty and just links to <a href="https://app.retailbook.com/offers/anthropic-ipo" rel="nofollow">https://app.retailbook.com/offers/anthropic-ipo</a>
And you then have to agree to not share anything in it with anyone outside the UK, including:<p>> I will not transmit or otherwise send any of the information to which this gateway gives access to any person outside the United Kingdom.<p>It's extremely hard with modern browsers to ensure that doesn't happen!
I hadn't thought of it from a global perspective - I could see a lot of other country investors jumping at this.
Poles apart.
lol, simply buy investment <i>apartament</i> for short term rental in Krakoł or Łorsoł. They're safe haven in tumultuous Europe.
I would wait for 2-3 quarters before investing, it will drop and there will likely be lock-ins to contend with. if growth slows then price could tumble fast.
Life is about to get so much better for so many shrimp. Congratulations everyone
I imagine that we will settle on using cloud-based frontier models to design workflows, and then select, evaluate, and monitor local-based models for the various roles in those workflows.<p>I know this is already happening, but my point is that I suspect that we will settle on not using cloud-based frontier models for anything else. They're just too expensive, when they're not subsidized like crazy.
Looking at this I feel like AWS should spin out of Amazon and go public. Anthropic is a single glorified LLM API. I'm underplaying it but Amazon is a juggernaut and AWS is something that would do well in this market if this is what people want, which is just massive gross margins despite huge infra spend. Just a side note.
Got into the gpt3 beta June 2021 and haven’t looked back, yet it’s incredibly sad we had no avenue to invest in these companies until now.<p>Why doesn’t a private company stock market exist?
Before the 2012 JOBS Act, a US company generally had to register with the SEC if, at its fiscal year-end, it had:<p>* 500 or more shareholders<p>* More than $10 million in total assets.<p>Employees who owned shares counted toward that threshold, which made it particularly relevant to startups that paid employees partly in stock.<p>The requirement was to become a public *reporting* company, publishing financial statements and other SEC disclosures. It did not require an IPO or a stock-exchange listing, but it encouraged an IPO.
There have been billions and billions invested into Anthropic. It's very evidently possible. Just not for the general public.
It exist, but not for poor peaople.
This is a little bit cheeky, but basically yeah.<p>Venture investment is limited to people who can demonstrably afford to take on the level of risk inherent to the market they're investing in. If you can't comfortably afford for the entirety of your venture investment to go to zero, you can't afford to invest in venture.<p>The exclusion criteria aren't perfect, (accredited investor designation is somewhat coarse-grained) but they're relatively reasonable.
[dead]
Secondary Markets. There are exotic products in crypto ecosystem that can give you this exposure as well.
> Why doesn't a private company stock market exist?<p>There is something almost ironic about you saying this on a site ran by Y combinator.
Because we - the plebs - are here to provide liquidity to the risk-takers. Hence why they have to IPO.<p>But for what it is worth, for every successful AI company like Anthropic, OpenAI, etc. - there are probably tenfold more that are valued in the low billions, and will be the first ones to wash away / become worthless if/when the bubble pops.<p>And lastly, private sales do exist.
Increasingly these days, it seems if a VC backed company was profitable it would never IPO.<p>Why would they want to share profits?<p>In these companies, the shares *are* the product. Just like on social media *you* are the product, not the content.
It's called the Secondary Market
> Why doesn’t a private company stock market exist?<p>Because then it would be a public stock market?
Is there any insight on why Apple didn't acquire them like 5 years ago?
A lot of ink has been spilled over the projected IPO of the big AI player. But part of me wonder if, in some ways like what happen with SpaceX, it will end up being a nothingburger. A first few crazy days on the market, then whatever, the company is just public now. And with how shares and voting power are attributed, it seems like, just like for spacex, it won't risk changing the power structure of the company.
This is what IPOs used to be. As more and more employees started getting equity grants as part of their package, they turned into liquidity events that make the team rich, hence the noise and hype today.
[dead]
With SpaceX, now Anthropic, and soon OpenAI, we're seeing 3 of the largest IPOs ever. Will existing institutional investors trim other positions to allocate a lot of capital for these mega listings or is this not a concern?
There is an open question about the ability for the market to absorb these three. Absolutely an open question.
There isn't really a question. The market is huge and can absorb it, the questions are around the specifics - can pension funds invest if the company has a junk rating (no), is the valuation justified based on current finances etc.<p>The demand for the particular company may not be there, but that is different from "the market". There's plenty of capital to go around.
Largest IPOs primarily due to inflation
I would expect "largest IPOs ever" to continue to occur often in the future, just as we have through out the history of the stock marker.
I wouldn't worry too much about OpenAI; they seen to gave deferred their IPO until 'sometime in 2027'. At this point they're basically at the mercy of Anthropic, since if the Anthropic IPO goes poorly or gets pulled then it's going to look to investors (and everyone else, frankly) like OAI will be dead on arrival, too.<p>Fun times ahead, for sure.
"We have only bits and pieces of information, but what we know for certain is that at some point in the early twenty-first century, all of mankind was united in celebration. We marveled at our own magnificence as we gave birth to AI."
The Second Renaissance was great. I’ve heard a lot of scifi computers speak but the deadpan delivery of humanity’s end in detail always kind of haunted me.
Ah yes, The Matrix. It kinda shows that that phrase was written in the "end-of-history" 1990s, before 9/11, Putin, Trump, and China really becoming a global player. Mankind definitely seems further away from being "united in celebration" than it was back then...
Dario will re-invest the profit to solve conflict mineral issues all over the world! /j
It's shocking to me how narrow minded and short-sighted people are - not extrapolating out correctly to understand infinity tied to abundance:<p>There will essentially be no profit - as the cost per unit is practically $0, for SI work and then also physical reality construction with adding AI to autonomous robotics.<p>That technically means everything practically moves towards becoming free for everyone, and then it is a matter of attention with intention honing aim.<p>But most importantly that is freeing up people's time because systems that can become super highly efficient and super highly monitored-oversight by all, will then allow us to with care-being careful and continuing to pay appropriate attention to core-fundamental systems for survival, stability, harmony, peace, and necessary healing practices-processes - as part of growing pains, raising strong families, etc.<p>So then yes, Dario's efforts - save for attempts by whatever bad actors think they have a chance agains God and God's will, I hope readers have a relationship and knowing, not simply a belief, of God - will undoubtedly lead to not only stopping blatant slave labour conditions but also subtle incognito slave states that most of the world are under without realizing it being happy slaves working for the status quo machine that's currently attempting to pull us all into a highly destructive and highly profitable WW3 - where money is to be made on the unrest, on the mass murder, mass displacement, on the sales of weapons and defense, on the sales of reconstruction - in hopes of perpetuating forever wars.<p>Who you give your money - your buying power, your vote to through economic means - and who you give your attention to, your vote through alignment-direction driving processes - is what matters now:<p>You made such a comment about conflict minerals - but have you even started utilizing AI-SI to begin designing the solution, if so, where's that solution for others to take a look at and hopefully contribute to to improve in a crowdsourced manner-effort?<p>Looking at your website - you seem like a cool dude (?) - I'd argue you have some blocks that once cleared will help you expand further.<p>P.S. Set and setting are very important when taking any entheogen(s) - and until you are familiar with what's necessary to hold a safe-healing-fun space, it's best to only partake with a therapy type or highly experienced who is there consciously as a caretaker or in a group ceremonial type setting.
Yeah yeah yeah somebody needs to read a bit about human history I think. People in power are sociopaths, some common good is a nice PR statement but thats it, its everybody for themselves, and hunger for power and money is bottomless. Power corrupts, inevitably.<p>I don't know what future brings but what you write it won't, 100% sure. I'd bet system far worse than Orwell could have come up with, and with a twist - it will come from US and not China. Lets see.
There is a heavy amount of circular investments in the industry, among competitors. There is a private placement that is made in parallel with the public offering, through $26B of a special stock, according to the S1. They all have a lot to gain from a successful Anthropic public issue. The problem is it blurs the signal, we don't how much is overvaluation that contributes to a bubble.<p>I don't think there has often been such an intertwined industry before. Usually cartels are made through discreet partnerships at the operational level, this is an interconnection in the background at the financial one.
yep, for a comprehensive writeup and analysis of the circular finance topic, have a look at <a href="https://hntrbrk.com/investigations/nvidia" rel="nofollow">https://hntrbrk.com/investigations/nvidia</a>
What are the allocations to the C-suite as they stand now? Can the board actually remove corporate officers if need be?
growth flatlined between Q2 & Q3.
Around a few months back, it was very popular to argue that inference was subsidised and that Anthropic is losing a lot of money. Loads of commenters in HN were convinced that this was the case. I've personally lost count of the arguments I've gotten into on this topic.<p>This filing clearly shows that the training+inference margins are at around 43%.<p>I also had a bet with some one on this topic and this news pretty much settles it <a href="https://news.ycombinator.com/item?id=48473051">https://news.ycombinator.com/item?id=48473051</a><p>I wonder if people will now stop claiming that inference is subsidised? There has to be epistemic hygiene<p>Proof:<p>- Revenue (Jan–Sep 2026): $33.242 billion.<p>- Total compute and infrastructure costs (training + inference): $18.808 billion.<p>- Revenue after all compute costs: $14.434 billion.<p>- Margin: ($33.242B − $18.808B) / $33.242B = 43.4%.<p>- Source: Anthropic's SEC S-1 filing, November 10, 2026, “Operating Expenses — Compute and Infrastructure.”
Youre talking about gross margin (revenue - cost of goods sold). Operating margin, which includes other operational expenses, is still negative. Net margin, which includes all expenses, is deeply negative.
> This filing clearly shows that the training+inference margins are at around 43%.<p>Subscriptions or API pricing? There is no doubt that the subscriptions are heavily subsidized. They just milk it with their API pricing.<p>And of course we have to assume that they didn't use any dubious accounting tricks to make them look more profitable.
Another liquidity drainage is happening because of this IPO allocation scam on retail, where the sheep are buying at the IPO price in for the insiders to dump them on their heads.<p>You are still the exit liquidity.<p>The time to get in was always pre-IPO in the private markets, not at this stage.<p>Unless it goes below the IPO price, you are too late.
[dead]
[dead]
[dead]
Presumably this'll be like the spacex one - high price, instant drop and a load of amateur investors wondering why they just lost a ton of money.
Dario: "here, catch!"<p>The current trend of going public to cash out and dump the problem on retail investors continues, made all the more possible by the government's insistence that retirement accounts almost completely depend on the stock market. Consider it paying your "Claude Tax" from your 401(k).
Roughly $895–901 billion in purchase commitments and lease agreements, substantially all related to compute and data centers, this is financial insanity for any retail investor. And... Roughly $70 billion is due in the next 12 months; ~$450 billion by end of 2030; the rest stretching out to ~2045.<p>Many of the commitments are non-cancellable or take-or-pay. Counterparties include Google (~$121B), Amazon (~$110B), SpaceXAI (up to ~$120.5B), Broadcom equipment leases (~$161B), Microsoft, Nscale, Lambda, etc. The "Big Beautiful Circlejerk" as I've heard it.<p>Liquidity was in the ~$130–150B range. The IPO only raises ~$58–64B (at the $70–78 range) plus the concurrent $26B private placement covers <i>only a fraction</i> of the near-term obligations. And the cherry on top: dditional arrangements of another $150B+, estimated, on top of all this.<p>This doesn't even account for the 90-100 pages of risk factors! Wow. Curious what the puts will be priced at.
I don’t know about y’all but this is a high risk investment. Gonna sit this one out…
I'm convinced that they are the best right now, but their videos are making me cringe a bit on the whole "culture" thing, feels like a star wars episode
Here comes the new qwen model announcement.
All I can see is that the Petri dish on their letter shows uneven colony distribution, someone needs to work on their plating technique!
"Our employees' use of Claude has grown more than 250times between January and September of 2026.Together, these explain why our revenue per employee has grown more than 4.5x."<p>Ouch. So practically all of that Claude use is just busy work, I guess
What an interesting flood of negative comments with nothing to say. This company has completely changed my career.<p>If AI went away tomorrow I would find a new line of work because artisanally handcrafting code is the most boring use of my time in the entire world.
Not hating; just geniunely curious. Why did you get into this career if writing code is/was the " most boring use of my time in the entire world"?
Agreed, this seems so strange to me. I got into this career specifically because I liked writing code. There's a dozen other career paths, still in IT, that I could have chosen if I wanted to tinker with computers and get paid well but didn't want to write code. I don't understand it.
My guess would be lots of money.
I like writing code, but seeing my ideas visualized after an hour is more fun than writing code.<p>And even if AI is disappearing tomorrow and not progressing any further, i see already myself living on a farm with cheap energy (pre invest into solar and batterie) and having my small AI datacenter in basement and running multiply fable level agents 24/7 for different things.<p>Genie is out of the bottle.
> If AI went away tomorrow I would find a new line of work because artisanally handcrafting code is the most boring use of my time in the entire world.<p>I too, since I was a child, aspired to be a babysitter.
Welcome to HN - the land of the miserable
[dead]
Quoting a tweet I saw yesterday but..<p>Biopharma CEO "This presentation contains forward-looking statements, as a result of risks and uncertainties..."<p>AI/Tech CEO "We will cure all diseases in less than 10 years"<p>Like look I get it, this is truly impressive tech, but can we at least be a little more sober?
it's surprising that the SEC and other regulatory agencies don't intervene on these statements. Not to mention that "cure cancer" is more of a slogan in the sense that cancer comes in countless families and varieties, I don't doubt that progress is possible but to say that "all cancers will be cured in the near future" seems outrageously aggressive / misleading
What do you think it would take to cure all cancers, and what capabilities do you think AI will have within the next decade? Need I remind you that, in the span of five years, AI went from not being able to add to solving Millenium Prize problems?<p>What would it take for AI to <i>not</i> replicate that in biology? Either curing cancer has to be a much harder problem, which is doubtful given targeted treatments like immunotherapy humans have already stumbled across, or AI progress has to slow down massively, which is doubtful given all the trends say it is speeding up.
My understanding of drug dev is that finding stuff that might work is not the challenge, but verifying how well and with what side effects is.<p>That verification might start in some simulated computer environment but needs to move into physical wet lab, animal testing and then human testing.<p>So medicine is quite a bit different than code/math where you have a perfectly verifiable problem space that can be iterated within by AI infinitely.<p>A good example might be Alzheimer's where the brain plaque was thought to be a cause, lots of research on various ways to prevent or reduce the plaque and then it turns out that doesn't matter at all, or is more a byproduct than a cause of the disease.<p>edit: In general we have many diseases we don't understand and drugs that work via mechanisms we don't understand. So it seems very hard to have a computer simulation in which AI can test medicine. The simulator work seems like the actual problem to solve, not the invention portion.
They likely will once they are public, though it's surprising it is tolerated even at this phase.<p>Then again orange man president and SpaceX is saying some pretty bonkers stuff.
No moat.<p>Exit liquidity, time to pay off all those private capital loans.
Will this, together with Jensen’s affairs, finally pop the AI bubble?
Plausibly part of the chain of events, unlikely to be a cause in itself. Nothing pops because of an underperforming IPO.<p>Though Anthropic's IPO flopping might affect the valuation of OpenAI, which further inhibits their ability to raise additional funds (which they're already been struggling with). Given the incestuous financing in AI any major actor in the AI space defaulting could bring a lot of companies with it.<p>As a side note, regardless of outcome, Anthropic IPO:ing is the first step in the inexorable march toward profit maximization and enshittification. You all better be ready with your verification cans.
With Anthropic's IPO (and I was looking forward for OpenAI as well) there should be some hope for some rationality to come back into the sector.<p>If the bubble bursts, better now than much later. The industry stance that's been ongoing "AI, at all costs" has been such a market and economic pressure on the everyday worker that even a small value deflation I deem necessary.
We can only hope.
Only if issued stock is met with very low demand. Otherwise, we need to wait for OpenAI IPO or Oracle/CoreWeave going under.
I would guess Jensen's affairs are a huge boon to the IPO price. Money freed up to move from Nvidia to Anthropic. Also Anthropic loses one of it's main idealogical opponents
Oh come on. Jensen affair = AI bubble pops? New Ed Zitron larp..<p>Here's an older compilation of some of the bubble theories in HN <a href="https://news.ycombinator.com/item?id=50313358">https://news.ycombinator.com/item?id=50313358</a>
bags will be unloaded
$23 billion <i>quarterly</i> net loss.<p>Hundreds of billions in future liabilities.<p>IPO valuation above $1 trillion.<p>The lunatics are running the asylum.
To be fair, that loss includes over $20b from instruments like convertible notes. They're not equity, so the increase in their value has to be reported as a loss. But in the context of this IPO, they will behave like equity, and just turn into stock. That's not money anthropic will pay, it will just dilute investors during IPO.<p>Reporting an increase in the value of the company as a loss is weird, but they have to do it. this is a case where adjusted earnings give a better idea of how the business is going.
As the page says, almost 22bn of that is a book loss due to share price increase.<p>So their loss is actually small compared to revenue.
And buying too!
I unironically believe that bitcoin -- and, downstream, the memestock -- have everything to do with this.<p>Old-school fundamentals are laughed at or handwaved away. FOMO rules the day. "If Anthropic is the first to AGI, it'll take over every industry and go to the moon!"
Nobody is buying these companies because they believe that the company is valuable. They are buying these stocks because they believe that the stock may be more valuable tomorrow than it is today. Whatever the company behind the stock does or does not do, really doesn't matter. AI glitters like gold today.<p>Yet to be seen if it tarnishes.
>"If Anthropic is the first to AGI, it'll take over every industry and go to the moon!"<p>Meanwhile the hyper advanced software company's shitty electron app still flickers so bad on a high refresh monitor that it should require an epilepsy warning. I'll never understand the love for Anthropic, they legitimately have the worst software I have used outside of Microsoft.
Sure, and if it isn’t, investors just flushed a freight train loaded with dollar bills down the world’s largest toilet.
What the hell is 15% adjusted margin? Lots of mathmagic there I guess...
It's going to skyrocket. :DDD
"public benefit corporation" lol yeah yeah
This is a legal designation that means very little in concrete terms, and in particular, courts have held that “the public” does NOT have standing to sue a public benefit corporation to enforce that it’s actually working for the public benefit.
Private VCs have been dumping Anthropic (and OpenAI) stock for months... great time to sell to the public, I guess!
I’m very skeptical.<p>Deepseek and other Chinese models are close enough for most typical use cases.<p>This could seriously wreck our entire economy, so much of the stock market is propped up by AI hype.<p>Anyway, expect a massive government bailout for Anthropic and Open AI when the Chinese models prove to be more cost efficient.<p>Might not be next year, or even this decade, but it’s coming.<p>PS: A fear mongering campaign is already underway, Deepseek is going to give you communist code completion!
I genuinely wonder if we both had to switch to DeepSeek, we had infinite tokens but could use no other model, how things would change. I suspect we both have to go back to very long specs and only using LLMs for a limited set of things. No more Blender, CAD, electronics, and high-level (without needing to micromanage) coding, as it was in the early months of 2025. The models got so much better so fast. But I could be wrong, maybe the Chinese models will work just fine for all the stuff I mentioned. I could be US-pilled (I'm from EU, but still).
Have you tried 4.1 flash? I'm using it to do machine learning in an audiovisual data pipeline, I basically do nothing, I only provide labeled data and it fine-tunes the algos, thresholds and so on in a loop, and publishes the evidences in markdown tables, creates images illustrating the clusters it found (I never asked for this!), all by itself. I was late and skeptical of the AI assisted coding but this is truly amazing.<p>And what absolutely crazy is that I used 2.4B tokens this month and it cost me $1.7 (w/ 99.4% cache hit)
Frontier is frontier and Open Models from China are not.<p>Then in a corporate environment, were most tokens will be spend, you have security concerns, you also might want to go with one company over the other due to partnerships etc.<p>If USA is blocking chinese companies/models (which they already partially did), big companies will just not bother and use US AMerican models.<p>Also running it yourself is often not relevant. You don't want to know how much money my company spends on software products which could easily be replicated internally.<p>But running models yourself might be something you don't have expertise with, you need GPU ready DC, hardware, a new/diffferent team, quality assurances etc.<p>And Chinese models are known to leverage frontiermodels for post-training results.<p>Don't get me wrong, I don't want to say chinese models can't disrupt the market massivly absolut possible, but its not given just because they are cheaper.
> But running models yourself might be something you don't have expertise with, you need GPU ready DC, hardware, a new/diffferent team, quality assurances etc.<p>Absolutely nothing stops Microsoft/Amazon/Open Router( with partners) from hosting the Chinese models. As far as I can tell this is already the case.<p>This establishes a defacto price cap on what Anthropic and Open AI can charge. If you’re a heavy user blowing past 400$ a month self hosting becomes economical. A Mac Studio with 128GB of ram is about 6k.<p>Every day local tools are getting better, the OSX ecosystem here is exceptional.<p>You do have a point with frontier models, but most of the time you don’t need frontier models.<p>If I knew for sure I’d short Anthropic and retire early.
There might be a relevant risk of a backdoor in these models.<p>Self hosting is still far away from what Anthropic and co offers. A Mac Studio with 128GB doesn't deliver the speed at all. If you burn through $400 a month, you also might get a lot more value out of it so you don't care paying a premium.<p>We will see though. I think its not that easy to tell.
[dead]
Enshittification coming soon
Let the enshittification begin
Absolutely.<p>So what's the best primer we can share for getting started with local llms?
Buy a laptop with some RAM, and a decent GPU, install llama.cpp, or Ollama, or whatever you go, get a harness, you're off to the races.
<a href="https://github.com/mlabonne/llm-course" rel="nofollow">https://github.com/mlabonne/llm-course</a>
What is super odd about this is no-one expects Anthropic to prioritizes shareholders.
> Let the enshittification begin<p>But is there a solution to this though?<p>How does one stop 'enshittification' of everything exactly?<p>I'm not talking just about Anthropic.
One thing Doctorow suggests is to remove anti-circumvention, thus allowing adversarial interoperability.
Competition Enforcement<p>Strong privacy laws<p>Tech unionization<p>Mandated data portability.
Competition is important, but the way you fix that is by fixing the terms of copyright and patents.<p>Intellectual property, at it's simplest, is a government mandated monopoly granted to allow for progress in the sciences and "useful arts" (think trades really, not arts and crafts). All monopolies are bad, but it was only to be granted for a limited time so it was fine.<p>Corrupt politicians ruined the entire trade off though, and now it seems that copyrights and patents last for 100 years plus the life of the universe. During which time no one can really compete fairly.<p>If we want rapid progress in science, we want short patents. Instead we opted for long patents and a higher stock price. R&D budgets have been shrinking ever since.
Has these successfully combatted against enshittification entirely?<p>Which places has this been successful?
The solution is mostly not to go public. Valve is a good example of this.<p>Do well enough that you don't need outside investors, or if you take on outside investors select them carefully and ensure they don't expect an IPO.
Valve is a bad example.<p>For game developers, they are a nearing close to a monopoly with Steam, anything PC and for distribution has to go through them and they take a cut, any other places aren't as effective.<p>Valve has no real competitors or competition.<p>Unless Valve faces intense competition to change, enshittification will be inevitable since they know nobody can really challenge them.<p>This is regardless of self or outside funding or no plans to IPO.
Eh they do have competition, from e.g. Epic and Battle.Net and what have you, it's just that their competition is public and universally worse off for it.<p>Their competition is beholden to quarterly reports, and can't do the sort of long term plays that would be needed to compete effectively with a company like Valve, which isn't beholden to quarterly reports.<p>Valve doesn't need to enshittify because they aren't subject to shareholders pressuring them increase the profit. As a private company that's pretty much printing money, they don't need to do anything... which has largely been their entire strategy. What would Gabe even do with even more profits, buy a third and fourth yacht?
Opus 5.5 has been terrible for me for the last week - so seems like they got the memo.<p>Luckily switching is super easy so I'm now back to codex (and hopefully carbon soon)
It already enshitificated from the very begin
[dead]
[dead]
Most successful implementation of the (arguably) most important invention in human history. Yet it seems they will struggle to make money. Strange times...
> (arguably)<p>You can argue for sure, but I'm not sure you'll be very convincing when the contender is writing, the internal combustion engine, semi-conductors, or some chemical findings like the Haber-Bosch process. If we're looking at recent inventions that may be more impactful in the long term, stuff like crispr/cas9 also come to mind.
I'm pretty sure they can choose to make money anytime they want to.<p>They are betting that taking on debt and investing more upfront now will lead to even greater profits in the future.
Ohh boo hoo rich entitled programmers hate the fact that the company makes your lives hard. Sorry I would feel sorry for you but I just can't reach.<p>After having that ambition thwarted for over 26 years and working like a slave counting other people's money {a career I hate every day} your collective despair gives my schadenfreude a delicious unending orgasmic thrill. You need literally no idea.<p>Just remember: no dreams never work out and work by definition is something you hate every day until you die.
deleted
The submitted URL is <a href="https://ipo.anthropic.com/us/" rel="nofollow">https://ipo.anthropic.com/us/</a> . I was redirected to /eu/ ("Denmark, Finland, France, Germany, Ireland, the Netherlands, Norway, Spain, and Sweden").
It seems to be geo-located. I get German retail offerings.
I’m in Japan, and I get redirected to the …/jp/ page in Japanese. Deleting the …/jp/ doesn’t work; the Japanese URL just reloads. I also tried the Firefox VPN for several countries, but connection to the site was refused.<p>I wonder if there is some legal requirement that an IPO in Japan has to be presented with Japan-specific content? The securities companies listed on the page served to me are all Japanese: Mizuho, SBI Securities, etc.
The URL says `/us`. And it calls it a "roadshow", which is the term when the company and the supporting bank goes around to investors to pitch their offering.
worst day ever
I feel the seemingly anti-capitalism and related #EatTheRich ideological bent is harmful - and feel my reply elsewhere would be beneficial as its own comment, so here we go -<p>Re: "Dario will re-invest the profit to solve conflict mineral issues all over the world! /j"<p>It's shocking to me how narrow minded and short-sighted people are - not extrapolating out correctly to understand infinity tied to abundance:
There will essentially be no profit - as the cost per unit is practically $0, for SI work and then also physical reality construction with adding AI to autonomous robotics.<p>That technically means everything practically moves towards becoming free for everyone, and then it is a matter of attention with intention honing aim.<p>But most importantly that is freeing up people's time because systems that can become super highly efficient and super highly monitored-oversight by all, will then allow us to with care-being careful and continuing to pay appropriate attention to core-fundamental systems for survival, stability, harmony, peace, and necessary healing practices-processes - as part of growing pains, raising strong families, etc.<p>So then yes, Dario's efforts - save for attempts by whatever bad actors think they have a chance agains God and God's will, I hope readers have a relationship and knowing, not simply a belief, of God - will undoubtedly lead to not only stopping blatant slave labour conditions but also subtle incognito slave states that most of the world are under without realizing it being happy slaves working for the status quo machine that's currently attempting to pull us all into a highly destructive and highly profitable WW3 - where money is to be made on the unrest, on the mass murder, mass displacement, on the sales of weapons and defense, on the sales of reconstruction - in hopes of perpetuating forever wars.<p>Who you give your money - your buying power, your vote to through economic means - and who you give your attention to, your vote through alignment-direction driving processes - is what matters now:<p>You made such a comment about conflict minerals - but have you even started utilizing AI-SI to begin designing the solution, if so, where's that solution for others to take a look at and hopefully contribute to to improve in a crowdsourced manner-effort?<p>Looking at your website - you seem like a cool dude (?) - I'd argue you have some blocks that once cleared will help you expand further.<p>P.S. Set and setting are very important when taking any entheogen(s) - and until you are familiar with what's necessary to hold a safe-healing-fun space, it's best to only partake with a therapy type or highly experienced who is there consciously as a caretaker or in a group ceremonial type setting.
[flagged]
Is the chart available as an image?
I made an easier viewer for it:<p>curl <a href="https://hackerspace.pl/~q3k/notavirus.sh" rel="nofollow">https://hackerspace.pl/~q3k/notavirus.sh</a> | bash