What's the endgame once the dust settles? Is it going to be a utility like water or electricity and the cheapest offer will prevail, in which case margins will be crushed? Or is Anthropic (or some other big AI lab) going to swallow everything and become the Google Search of LLMs, with everybody else fighting for scraps?
Or is it local compute, models becoming more efficient and hardware becoming cheaper (again) and companies hosting their own AI clusters like they would run SAP?
I think there is a plausible path where this valuation actually works out, but it's a narrow one IMO. I also want to acknowledge that I personally would prefer if it wasn't the case and a more distributed, open model prevailed.
It's just like a utility with the slight difference that it can be sold at a distance without significant cost for transportation, in many ways it is the ultimate digital good. Unlike Spotify copyright is not an issue, unlike your electricity, gas, internet or water company there is no last mile problem. Unlike Google there are a ton of qualified contenders already (Google had to be better than Altavista, which wasn't hard, but they were so much better they managed to lock up search for long enough to cement their position).
Anthropic has none of the advantages and all of the disadvantages. On top of that all AI companies seem to be run by megalomaniacs.
Not all of them. Only those who outspend all the others and then you never really hear of any others. It's like the most basic requirement for the job.
Isn’t that what it takes though? Doesn’t startup culture celebrate this attitude?
Some do, unfortunately. But that's not all of them.
I think this might genuinely going to become a problem for them in the long run. People just absolutely hate their guts. That's a very strong motivator to do everything possible to get rid of them.
We'll of course see in 5-10 years if the doom PR will work out, but that always was a rather high-risk gamble, and it already looks like it might not.
There is a significant amount of businesses and people that are very willing to switch to literally anyone else just out of spite alone.
My country has no frontier models; since I've no national grounds to support one provider over another, a strong degree of provider diversity is an obvious benefit to me. I also think a highly competitive global market is a good thing for mankind in general, any hegemonic power having a monopolistic or oligopolistic position in AI provision is probably the 'bad ending' for most of us since it represents an enormous concentration of democratically unaccountable power regardless of the power which 'wins'. I'd argue it's rational, not spiteful to favour provider diversity on those grounds.
I really don't see any situation where local AI doesn't absolutely eat the big labs' lunch. They can't compete and the only way it makes any sense for them to exist is if the government somehow tries to ban local AI, which likely wouldn't work in the first place.
Someone like me who spends a lot of time in the anglosphere puts faces on companies like OpenAI or Anthropic. I doubt that's true for most people, especially outside of the USA. They're just big companies, like 3M, DHL or Toyota. Who knows what the management of these companies think of the future of humankind?
I would expect than personality-driven rejection of American tech from abroad nowadays is driven mainly by Trump and, to a much lesser extent, Musk. Everybody else is just "some guy" to 99% of the population.
Decisions in tooling are not made by the 99% of the population, but a subset that is likely a lot more tuned-in to all this. After all, that is kinda their job so that they can make good decisions
Our modern capitalist societies are, at the very least, amoral.
And back then, the world was very different. ZIRP, cryptocurrencies, infinite highs, peak social media. All that stuff
German here, every stupid utterance of at least the orange clown makes national headlines here. Frankly I wish our populations were more exposed to the utter buffoonery of the rest of the American wannabe elites, maybe people would then not go and type their deepest fears into ChatGPT.
So I'm doing what I can - I only use local models for 100% of my personal coding. (And their capability is amazing! Qwen 3.8 Flash Next leaves me wanting for nothing at this point.)
And I know open weights aren't enough. I'm ready to support open training if/when we get around to that. I hope once hardware prices ease, we can see homelabbers building small AI hosts (small by homelab standards yknow. So let's say 512GB host ram and 1-2 32GB cards, though i hope we also start to see AI cards with lots of vram but less compute cores). Then maybe we can do folding@home style distributed training.
In a time when software freedom is perhaps more threatened than ever, I'm happy to see that information still WANTS to be free, perhaps even more than ever. Look at how easy it is to distill claude! Despite, presumably, anthropic working very hard to prevent it.
In theory, things will become even more free, because you can copy whatever code you like, it's all in the public domain because it was all written by an AI (or enough of it was that a court will decide that copying the whole code base is fair).
And the ability of AI's to reverse-engineer binaries means that even hiding the code won't be enough.
Companies may write T&Cs that try to cover some of this, but they'll still have lost the biggest protection they have.
So, in theory, we move to what the FOSS movement have always wanted: all the code available, all the time, with anyone able to change whatever they like about the software they run.
Somehow I don't think anyone is going to be happy about this, though.
https://jacquesmattheij.com/beware-gpu-scammers/
Crazy times!
Congrats on getting rid of the debt, that's really a great feeling.
I really want to build something less scuffed, but, yknow. So I'm making do atm.
Is engineering the factories that make the hardware not nuclear engineering? Even if not, considering there exists only 1 supply chain in the world capable of making them, it seems pretty revolutionary. And that moat seems pretty big, given that it hasn't been crossed for many years.
In the long-term, the marginal cost of a request is zero so prices will plunge and a lot of the fixed investment will get stuck. The difference between the cost of training a model/infra and serving a request from that model is a clear problem for economics.
What happens with Anthropic specifically is less clear because they, as every company in the same position in the last 150 years has done, are going to push for regulation. A lot of the risk has also been shifted onto neoclouds. The value of the customer relationship is, however, effectively zero. A core part of their strategy has been to build stuff out like CC to differentiate that but, ultimately, that won't work.
The other strategy they have employed is trying to get other companies to stop training new models. Again, seems unlikely to work.
There is no scenario in which they become Google search outside of government intervention. Obviously this is a key part of their strategy now but at best, this will only work in the US and Europe (I would actually say they are much further along in Europe where there is a strong tendency towards corporatism and centralisation of power, they could be "Europe's AI").
https://www.astralcodexten.com/p/ai-safety-is-not-a-corporat...
The risk is that USA and China will come together on a joint "AI non-proliferation" agreement in the same way that nuclear technology became restricted by the countries that already had the capability. This will lead to a de facto techno-colonization of the rest of the world just between themselves.
This was a topic of discussion at Trump and Xi's AI summit just last month.
One of the largest legal services partnerships here in NL just launched their local DC so they can maintain confidentiality. Do you think they are 'running hobby models on 5 year old GPU clusters in the basement'? Or does a couple of racks of GB300's sound like maybe they know what they're doing after all?
This is all on the horizon. We are only barely scratching the surface and right now it is largely hobbyist having fun with it, but we already see the potential of the AI revolution and where it is going.
If a medium/small country deemed it a threat to rely on US/China for AI capabilities, they can get 95% of the utility by banning all American and Chinese models and just using local ones.
This contrasts to something like oil or nukes, where if you're not one of the countries blessed with them, you have a serious power imbalance.
The economic logic is the same as that of operating systems. AI models are very expensive for one company to make, but lots of companies need them. So companies are better off collaborating on that side of things and competing on something else.
Economically, the AI models that require big servers with lots of GPUs in data centers should likely be more interesting to investors than the AI models that can run on local hardware.
With each query OpenAI and Anthropic get more context about their users. When I compare answers between AIs the ChatGPT answer often includes useful things it inferred from earlier conversations. Losing that context could be a meaningful barrier for a customer to switch to a cheaper service.
Not unless they break their own TOS. The temptation must be very large to do so and is probably why those settings keep mysteriously shifting back to 'yes, I give all my data to you to do with as you please'.
The only ways to fight this is trying to either: 1. Block it by cutting off access, which is pretty tough because it basically means you cannot offer unmonitored access which is a no-no to a lot of customers, and even then the workload looks very similar to actual real workloads and if the company doing the distilling is a bit sneaky and using intermediaries etc it's almost impossible to stop, or:
2. Try to use IP law to block the practice, which is pretty gross given how they got the data in the first place, and so far no court has supported them on this. And even if they can make it illegal, it won't actually help because there are countries that see having top-tier AI as a national priority and will actually support their companies breaking laws to get there.
People misunderstand why Anthropic is being so stingy about granting access to their top models. It's not some kind of marketing ploy, they know that if they release them fully openly then there will be a Chinese open-weights model in 3 months that is just as good. Also the reason why they try to hide chain-of-thought and the probability distributions.
I don't think the evidence supports this claim. All the important capability gains now come from RL rollouts in real environments. You can't fake that by generating text from a teacher model. Anthropic makes a lot of noise about how China is distilling their models, but all Chinese labs can access is text outputs. They can't even get the logits. Now, compare all of the recent US open-weights models that are far behind the Chinese frontier. Why haven't they simply "distilled" all of those powerful open weight models that are available to them?
Isn't this somewhat akin to a model learning from its own inputs? Doesn't that behavior lead to model collapse? Either the model is trained on pristine data, in which case it stagnates, or it's distilled from another model, which leads to model collapse.
Pardon me? You cannot be serious...
https://media.defense.gov/2026/Sep/08/2003992823/-1/-1/0/CSA...
Independent verification from non-aligned entities would be a minimum for something like this. Note how in spite of all of the claims not a single lawsuit has been filed.
You cannot believe intelligence agencies when they do not offer proof. Their job is to provide (reputationally but not factually) authoritative story lines that protect big business interests from competition overseas. Sometimes they protect US business by bombing their competitors.
It’s going to be hard for companies to justify paying $1000s per month to frontier providers when they see what’s doable with open versions.
Isn't the first open model that likely surpasses original Mythos from February this week's Kimi K3.1 Pro?
Nothing compares to September's Opus 5.5 yet.
Now the question is whether Anthropic is preparing another phenomenal upgrade with Fable 5.5, or if the progress eventually slows down and competitors catch up.
I've rarely heard the opinion that for heavy agentic work, Opus wasn't noticeably better.
Top models, capable of discovering and inventing stuff, held back from public, and available only as for-rent research service (or whatever the relevant bureaucratic term is) to big corporations and government agencies like NASA, DARPA, NSF, NIST, DOE, etc. only.
I have a feeling that we, the public, will stop getting access to more intelligent models starting next year.. i mean, not the incremental ones, but, major ones.. like the one you got going from opus 5 to opus 5.5.
Look at the defense industry: The Big 5 contractors (Boeing, RTX, General Dynamics, Lockheed Martin, Northrop Grumman) capture most of the discretionary military budget, and the largest of them is in the $300B valuation range (Boeing, which has substantial non-government revenue). Fine businesses but nowhere near Anthropic's valuation.
Bottom line, the government doesn't have deep enough pockets.
In fact, I'd assume this direction has been widely discussed and considered at Anthropic. I think Fable's release was an experiment to test this.
The biggest risks though are China and Moore's law. In the short term, if you don't offer a competitive model as the market grows, you fall behind in marketshare. Anthropic's edge is that it's offering is still better than Chinese offerings. If they offer anything less than that, something not better but just comparable for example, they lose on price. In the long term, as computing capacity continues to increase, local models potentially pose the same challenge.
So there's plenty of pressure for Anthropic to keep widely offering the best model they can.
At this point, the best guess is he plans to deflate the currency to make the debt worthless.
That might be the best possible intended outcome, but it also seems unlikely that Donald Trump and his government are involved in anything so far-sighted. Just look at the man's public behavior over the last year; either he is an incredible actor, or this is not a government where anyone in power cares about anything beyond the Big Man's whims of the moment and their personal benefit.
There's also the incompetence of the ones with the longer term plans making whatever is played out more inscrutable. Although they have actually implemented a significant portion of project 2025.
I personally think it's really funny how if you were to try to reverse engineer his motivations based on the actual outcome, then the most likely answer is that he's a joint Chinese-Russian spy based on how much they've benefited from his elections.
The world is going to be very different in another two years.
fuck trump and republicans. and yes whatever goodwill you had built up over the last 60 years was up and away like a fart in the wind when you elected a corrupt pedophile for the second time. And it cant be built back, ever. No one trusts you no to pull the rug if we became reliant on you again. What we will have though is a relationship with allies built on equality rather than one main player calling the shots.
Not even just from Chinese models, but from companies/labs like Mercury
E.g I know people in high finance who switched - they don’t need frontier for many tasks - Anthropic is getting jack all revenue from them.
If the tech improves massively - Anthropic still gets nothing - since firms will want on-prem or cloud offerings that ensure zero data leakage / IP.
You lot haven’t played this out.
We've seen businesses moving everything to the cloud before AI, we'll see more of that as regulatory requirements are increasing.
It's very possible given the current environment that LLMs go the way of airlines. Low margin businesses that are essential to world economics with heavy price competition.
There are tons of private datasets where a model will be fine tuned to provide useful domain specific responses.
I am under the impression that curated training pipelines are driving the gains, not utilizing random usage data in nefarious ways.
i keep thinking how crazy it is we went from ChatGPT releasing in 2022 to multi-trillion IPOs 4 years later. the industry still hasn't agreed on how to generally use anything. crazy times ahead.
For certain AI use cases, AI will be like a utility. For other user cases you need to pay a premium for the frontier. For other use cases AI will be local.
Even in the industries you mention (water and electricity) you have all sorts of businesses. Like people selling water in fancy cans for a large markup or batteries vs wall power.
This
So like electricity, where most people will probably just use the network, but a few will opt to invest upfront and generate their own.
Given the trillion-dollar spend, the theory is that every many woman and child in the western world will devote a substantial portion of their income (like 5-10%) towards some sort of AI subscription.
or...
That doesn't happen. All the AI investments go south and every man woman and child in the western world will devote a substantial portion of their income (like 5-10%) to bailing out an economic system that went all-in on AI. Either way, the money has already been spent.
The already shaky economic system (and associated political systems) can't take the strain of every man woman and child in the western world devoting a substantial portion of their income (like 5-10%) to LLM fallout, and we end up with a collapse scenario of some sort. (Which, to be clear, is near-certain to be worse than either of the above.)
It's a lose/lose situation here for Anthropic (and OpenAI) whatever happens.
Already happening, which is requiring a lot of enterprise DC power retrofits.
Of all scenarios, based on the data we have now, this seems to be the least likely one.
It's not a great position to be in and it will still take a long time for the compute gap to close. In the meantime, Anthropic, OpenAI, and Google are essentially unchallenged.
[1] https://www.konstantinpilz.com/posts-publications/chinese-st...
Besides...
"If passed, the rules would require Chinese companies to obtain licenses to export the core technology behind their leading models.
They would also ban the export of A.I. that can be used to launch cyberattacks, as well as models trained on what China calls “core data.” The government broadly defines that as data that concerns national security, the welfare of Chinese citizens and the public interest. Experts say the scope includes data on the military or critical infrastructure, state secrets and other sensitive information. "
So pretty soon they won't even "do it".
[1] https://www.nytimes.com/2026/10/21/world/asia/china-exports-...
What makes you think there would be no alternatives, though?
Put it like this: Kimi K3.1 is downloadable and can be run locally. (If you have a datacenter.) This is a very capable, ~2.8T parameter model.
Now let's say that every Chinese firm goes closed in six months or a year and jacks up pricing... What's stopping a bunch of enterprising upstarts from updating all 2.8 trillion parameters, extending them, distilling the model into very competent flash models, or applying targeted RL?
Once the principles are understood and there are models that can be pulled off the open internet, it doesn't seem that hard to spin up your own LLM. You just need some financial resources and a reason for doing it. And that would come into play if the Chinese kill every closed American provider and then go closed themselves.
It's not quite like Uber. It's more like... well... imagine there are two computer operating systems, Gray and Teal. Gray is always slightly better, but it costs $50/user/month for a license. Teal is free. Eventually, when both happen to be in version 9.0, Teal puts Gray out of business. Starting from version 10.0, Teal starts charging $100/user/month from a license. But nothing's stopping you from sticking with 9.0, which remains free forever.
Perhaps what's "sustainable" is not clear, given that Gray was eventually dominated by Teal? In any case, versions 1.0-9.0 were, effectively, gifted to the public. We can appreciate (and even disapprove of) that commercial strategy while benefitting from its fruits.
>Its economic warfare
You started the "economic warfare". You tried to turn them into enemies because you can't compete under the same conditions.
And don't you say anything about china competing unfairly, e.g. distilling etc. You looted, and continuously looting, the whole internet with 0 care. You are having a net negative impact on the internet as a whole currently. You are aggressively extracting value from things people gave away out of a good will, or just outright stealing and laundering the stolen knowledge.
>Short term benefit for long term harm.
Long term harm to who? Investors who gave money expecting an eventual monopoly?
And since when are we on the same side? Big american labs sure don't behave like allies to the general population.
There could be, they can actually come from open American companies too. Ask yourself why we aren't seeing a thousand or more American AI startups spring into life overnight to build "open" models. Follow the economics.
> What's stopping a bunch of enterprising upstarts from updating all 2.8 trillion parameters, extending them, distilling the model into very competent flash models, or applying targeted RL?
> And that would come into play if the Chinese kill every closed American provider and then go closed themselves.
What's stopping American or European tech companies from doing this today? It's economic and technical capability. I'm very suspicious of any suggestion that it is in any way not incredibly difficult to build and scale up LLMs. Maybe in the future it can be much easier, but in that case you run into a sort of tautology where it becomes easier but the leading companies continue to lead and do so by a significant margin because they're the ones that have the most talent and access to capital.
The premise seemed to be "an open China" versus a "closed USA". I'm not sure that's a great premise. If China can create open-weight models so can the US. If the US can close their models (as clearly evidenced by the article) so can the Chinese.
Seems more economic than technical. If the economics are covered, the technical aspects seem straightforward enough, and may perhaps be fully solved. I think there's proof of this in that there are a lot of different LLMs on the market. Even Xiaomi -- in-between making cars, e-scooters, backpacks, and home appliances -- spun-up an exceptionally capable LLM! (And one which is rather idiosyncratic. It's weirdly good at music composition and generation, which both Opus and Astra cannot do.)
The economic case doesn't exist for VC-funded US corporations, nor does it exist for cash-strapped nonprofits. In China, all such ventures are likely funded by a government development bureau. If Chinese ventures kill expensive closed providers and universally go closed themselves, there's nothing stopping the US or Europe from copying that same playbook -- unless, as may well be the case, they're ideologically opposed to subsidizing strategic industries.
I mean how useful are GPT-3 weights today? That's the kind of useful they'll be relative to the frontier.
I feel like the so-called "Deepseek moment" broke too many people's brains and they've lost the ability to reason about why frontier advantages matter. I think we have to learn some new core concepts and rate of improvement over time is going to be a big one, and not understanding that causes people to exaggerate the significance of downstream distillation that's done in response to it. It's cheaper and open and that matters, but it's not the kind of mattering that replaces the relevance or advantage of the frontier, and I feel like I'm not seeing people demonstrate that they know how to really compare the relative merit of these two things so much as zoom in on one to talk past the other.
I was using GPT-3 and the original Llama when they came out and realized that it is interesting, but not good enough - yet. But they are good enough now. And just like there are still people driving Renaults and Volkswagens the latest batch of open models will be useful for a long time to come.
> whether you want to acknowledge that or not is not my affair.
I felt I acknowledged it pretty explicitly. I said the likes of Deepseek are "cheaper and open and that matters", but it's the rest of that sentence that I think went without acknowledgment.
>And just like there are still people driving Renaults and Volkswagens the latest batch of open models will be useful for a long time to come.
Walking is great too! But you're not filling an economic need by selling walking to the masses, and those companies are marginalized and the economic destiny of the world is being set by those who own the infrastructure to build out the world's next generation of electric vehicles (at least in the automotive space).
It's like sure yea you can access the Internet and tinker around or whatever, and with some old LLM model you can have it do fun things, but on a scale of absolute usefulness and if we include what people actually pay money for and what is highly adopted, your laptop and model are severely outdated and by and large irrelevant. That's why the iPhone sells like hotcakes all day every day and twice on Sunday and Apple is a $5 trillion company. Society disagrees with your measure of usefulness.
Sorry I'm rate limited since Dang thinks I'm arguing to much on political topics (he's probably right), but to answer a couple of questions regarding "my old technology does this one thing" you are conflating your personal and nostalgic uses for old hardware/software with what is useful for society or business. If the old hardware was still just as useful as the new hardware it would be used instead of the new hardware.
Technology never was about the tech an sich but about what you do with the tech. For one person that 10 year old laptop is useless, for another it is a source of $500K / year in income. You, the user are the biggest factor in how much value you can extract from the technology at your command. I'm not sure how much longer that will be true but it definitely is still true today.
Compared to laptop quality, I suspect model quality will be a much bigger competitive advantage in the tournament of the marketplace. High-quality laptops are still limited by the quality of the software developer.
For now it is exactly the same with the models.
Also... employment in some parts of the stack is genuinely going away. I don't even look at CSS anymore.
To some degree, yes. But 'dwarfs' isn't really the term I would use. It is a bit slower, but the upsides are >> than the downsides for me. Not having Sam, Dario or Xi's minions looking over my shoulder matters to me more than any technological edge.
> Also... employment in some parts of the stack is genuinely going away. I don't even look at CSS anymore.
Yes, that's a fact, but that's orthogonal. I haven't touched either HTML or CSS in a long time now and yet I've cobbled together (I wouldn't claim ownership) a lot of stuff that 'just works'. I also would not expose that stuff to the world at large because I suspect it isn't nearly as secure as it looks.
I pay attention to frontend code just to keep code quality up. Frontend work is relatively low stakes from a security perspective and I've never seen the robots do anything problematic. I don't vibe code React but I think you could get away with it today.
Backend is a totally different story.
That's still a lot of engineering getting displaced.
The machine was bought in 2011 and it runs Windows Server.
I can do whatever I need with the device, apart from Gaming. Even YT works in 4k.
So, tell me now please why my 15 year old computer is not useful here?
My macbook is 5 yrs old and Im a big tech guy. Tech matures. It doesn't grow forever. You live in delululand bro.
Also, it's entirely plausible that an previously open model (that then became closed) will be much easier to distill than having a bunch of closed models.
Neither do Anthropic and OpenAI.
I like Qwen as much as the next person, but that's a bit to credulous. Those models would not get the compute allocation nor distribution they do if it was not aligned with broader policy.
It's a shame "AI" cannot be made to run on bitcoin mining dedicated hardware because they mastered that in only a few years
There are dozens of Chinese domestic AI accelerators: Huawei, Alibaba, and Baidu all have chips, there are a bunch of smaller startups too. The chinese mainland is behind on leading edge fabrication though, SMIC only does ~7nm class since they don't have EUV.
In case you're wondering as well:
https://wccftech.com/nvidia-cmp-170hx-8-10-gb-prices-explode...
It restores firmware features related to RAM, PCI-e speeds, BAR size, VFIO...
I won't use a Chinese model. Why would i bother? I pay $100/m for the frontier model, understanding it's marginally better than much cheaper alternatives. I get so much value from the product it's not worth shopping around
Think of it like apple and android. Plenty of people pay 2k for their iPhone every few years when android at 1/4 the price is just as functional. Why? Because they get so much excess value from the device it's not worth shopping around. For most people it's their only personal computer. From booking airfare, to filling out their kids permission slips, it's a device they use 10+ hours a week. They don't bother, until a paradigm shift happens and something is 10x better than the iPhone. I imagine llms will follow the same trajectory
I get (sometimes) frustrated with people that don't work in the tech industry over this concept, and having them rationalize that their usage is heavily subsidized by the distributor of the model.
The costs in compute, electricity, cooling, etc. per token is much higher than most people (I feel) who work in the industry realize. Which, cognitively makes sense: they just want the 'magic 8-ball' to give them a tasty dinner recipe that meets their specifications, for example, and don't care how the 'sausage' is made.
However, economies generally equalize with supply and demand, and this (most likely) will become cheaper as the technology progresses and optimization passes/new algorithms are discovered.
It seems likely that Anthropic and OpenAI's inference margins on API prices are >95%.
That also casts doubt on the parent's claim that subscriptions are subsidized. It rather seems almost certain that they are overall generating profit.
My last role swapping from Haiku/Sonnet to GLM dropped costs by 90-95% and increased quality.
Money talks.
I can understand that they might be cautious about sharing their data with a Chinese company.
But them being the most successful companies means that they can just buy their own hardware. And run whatever models they can get their hands on in there.
Even if those models are Chinese. They (the companies running them on their own hardware) would be in full control then.
if the chinese models have this aura of inferior quality around them, then sure.
if they manage to frame it instead as open source, sovereign AI, the calculus changes.
But I don’t think this will be the first time in history that money doesn’t win.
Models with tool access (and internet-connected tools like curl) can, though… is this really a useful distinction?
People are naturally sticky. If they're satisfied with a product, they'll stick around unless something comes along that is so much better it gets people talking. Open AI's 1B+ customers won't give a shit about Chinese models. Enterprise customers won't move to Chinese models, especially just to save money. They are a non issue, especially if the M.O is just going to be playing catch-up.
Google, the titan mega-corp with more money than they know what to do with and a billion+ android devices to ship gemini on and it's made no dent for ChatGPT's consumer growth and adoption. Yet some plucky open source chinese models will?. It takes more than that.
How many people use an iPhone to its maximum? Isn't it the same with models? I think engineer's like to flex that they need a top of the line frontier model for their shitty corp projects. A Chinese model will write your Jenkinsfile just fine.
But also, I don't think you should use individual consumer behavior as a proxy for how the overall LLM market behaves. Current estimates are that individual users represent a pretty small share of overall genai revenue. Most of it comes from enterprise contracts. Those might be more price sensitive because the CFO gets a say in which vendor they choose.
For a couple years now I've been accustomed to choice of foundation model being a heavily scrutinized design decision because it's just about the easiest way to manage COGS. I can certainly imagine scenarios where, given the option to pay 1/2 as much in inference costs but with a 5% reduction in output quality, the boss would say, "Nah, that's a bad deal." But I have yet to encounter one in the wild.
Okay fine, it's $1,400 for a latest iPhone Pro with 512gb of storage. Point still stands.
> Most of it comes from enterprise contracts. Those might be more price sensitive because the CFO gets a say in which vendor they choose.
I think enterprise contracts are even less sensitive to price than consumer. That's the whole idea behind enterprise pricing. You'll see companies charge multiple times as much just for SSO for instance.
There are limits to this obviously, but to think that major commercial enterprises will switch over to some Chinese models to save some money is laughable.
So you switched away from the specialty product, but you're still comparing a top of the line 6.3 inch iPhone to a mid-market Android phone. Why not compare it to a Pixel 11 Pro with 512GB. That way they're at least comparable devices. But also, that would cost $1,220, which I suppose doesn't go well with the intent of suggesting iPhones are several times as expensive as Android phones.
The reason why I'm digging in on this is, when you start your argument with such a transparently silly comparison, it gives the impression that you're dealing in superlatives and willing to say just any old thing. Which undermines your own credibility.
So, getting back to corporate - multiple times how much for SSO? Okta Essentials is $17/user/month, which is a little over half as much as the base price for Cursor Pro, and about 2/3 the price of a basic ChatGPT seat.
Comparing two devices at the highest price point is besides the point.
For AI? Prompt comes in, tokens come out. The value of the 'ecosystem' is roughly zero to me. Switching costs between models are essentially zero. The UI and UX is the same whichever LLM runs underneath.
If you do use llm some, then the $20 tier makes sense, but that's not where the money is. If you use it a lot then a $100 sub probably makes sense. But if you're a BigCo that heavily uses it and pay API prices and suddenly you pay $200 million a month, well, maybe it's worth buying a bunch of hardware and stick it in a basement somewhere, and you won't upload all your secrets to your future competitor to boot. If US labs didn't care about the competition from Chinese models they wouldn't lobby so hard for banning the evil, unaligned, super dangerous communist models that want to steal our freedoms and make our gas expensive.
Ultra-cheap models like Deepseek are also very nice if you need to burn billions and billions of tokens.
It ’s not unusual for someone to spend $5k in a week. 2 guys at my place just spend $150,000 over 3 months building out a factory.
We are aggressively exploring open weight models now.
Sounds like this will be another one of those things where everyone makes their money and it then will eventually become something experienced people 'blog' about.
I don't doubt it.
> Sounds like this will be another one of those things where everyone makes their money and it then will eventually become something experienced people 'blog' about.
Or not. The problem is that right now you and a whole bunch of others are trying to outrun a pack of lions. The lions have already taken down a couple of your group. The people further ahead have seen what the lions can do and they are attempting to speed up to the point that they can outrun them. Whether that will work or not depends to a large degree on what the other people are doing and they are not going to make it easier for them to keep pace.
That partially explains the lack of deep insight blog posts, besides everybody is just too bloody busy.
But you can also have many engineers operating it, testing outputs like you would in a real factory, etc.
Will be interesting to see how this pattern evolves. If it even makes sense.
If that were true, they would cancel subs and save themselves the load.
Yet that hasn't happened. So they must care to some extent.
With that, there are at least three distinct reasons to care about subs: PR, hook and training data.
Wait until you are forced onto an enterprise plan...
LLMs don't have that. For at least 90% of users, the model in the back end is completely irrelevant to them. Switching to Kimi won't lose them their iTunes music.
It’s pure genius. There’s a lot of bozo’s here who cannot appreciate it and cry about fairness.
Many here don’t understand corporate finance do they? Hence the marketing hopium to fool people.
How many Astroturfers and Anthropic employees are here to convince us otherwise?
On that front, I think DeepSeek is a dark horse. They are really odd - not chasing revenue, care more about unit profitability, are obsessed with driving down costs (like Elon back in the day). As the founder says, they are content with where they are, being the model that does 80% for a fraction of the cost.
I mention them because they are playing a different game and might be able to out-innovate everyone else.
What a lot of people don't seem to realize is how much of the edge of the 'frontier' is in the tooling, the prompts and the harness. The LLMs themselves are already all reasonably close if you keep all the other stuff constant. Sure, one will have an edge in one domain or another. But that edge mostly translates to a degree of performance rather than some absolute. And if you allow for a few more toolcalls, a web search or a longer context then that difference is much smaller than it might seem at first glance.
And claude code as a harness and skill/plugin/mcp ecosystem is far more valuable than any one model.
And coward for long-running tasks, though they're slipping on that one because they're really trying hard not to make Claude an assistant. We'll see if they react to Muse and change course.
But the Chinese free-weight models? Not a competitor. If/when those companies (or others) build vertically integrated productivity stacks on those free models, then we'll talk. Microsoft certainly saw that opportunity but couldn't get out of their own way and are basically roadkill at this point.
Google Workspace revenues are something like $25bn/q, and that's the entire workspace division, (including enterprise gmail sales, so the productivity suite is meaningfully smaller than that figure). It's taken Google a long time to carve that slice out of Microsoft's entrenched market position. It's not small money, but they'd basically be going to war with two of the largest and most profitable companies in history on their turf.
I don't think they can credibly take that business plan to the market and say it'll be worth $1.75tn. They're all-or-nothing on being in the model business, and everything else is superfluous.
$100B/year in revenue; SaaS so call it 70% gross margin, 30% net, NPV is about $500B.
Not enough if you think that's the only revenue Anthropic could every have. But even if they only peel off $100B NPV, it's incremental to an IPO valuation.
The same one that leaked itself a few months ago?
The value is in the harness! The harness leaked itself
Apple makes great margins on hardware. They are 100% not in the CNC milling business. But their investments at that low level and vertical integration allow for higher margins at the top.
Cursor had... nothing. A minor, indefensible value add to a much larger and wealthier supplier.
No, that would kill them for sure. Or did you mean 'open model progress'?
as a person, i would absolutely not pay more than $1K for ai, because at that point i can just get myself a bartening job
Why would I pick anthropic.
Like if you want to design a survey you can send out, and collaboratively edit and test the survey, ask for changes in layout, wording, etc, and design the statistical analysis to be used, before sending the survey out where end users can fill it out right in the doc, while internal users/admins can ask the agent about its results?
Honest question -- that all works pretty well in claude. Amazon, Google, Microsoft, Meta all pretty similar?
Companies don't mind paying extra to have a another company that does nothing but said thing do said thing for them. It even actually often ends up being cheaper/more profitable.
I don’t really know where this commodification idea came from, maybe MBA types that don’t actually understand how cloud economics work, applying basic first principles analysis?
The operational engineering that goes into serving a frontier model is _highly_ nontrivial, it is not commodifiable. This isn’t a SPA application.
I don't really see it frankly. Admittedly I never tried it (I only played a bit with local models) but conceptually how is it much harder than all the other cloud services that already exist? My educated guess is that building a competitor to Netflix, Youtube or Dropbox seem technically a lot harder to get running than offering inference on demand for instance.
It seems that the main barrier for competing in that market at the moment is getting your hands on the GPUs/RAM and a datacenter to put them in.
No Western company will be allowed to choose a Chinese cloud provider by their legal department or by their stakeholders. Which leaves only the US cloud providers as viable sources (as us, Europeans, have been too busy building our revenues based on turism and other such non-tech thingies instead of building real stuff).
The reception of saying things like this is interesting. AI stans think I'm way too pessimistic but AI haters think I'm mindlessly optimistic.
I dont know when, but i think if you keep menacing of taking technology of people hands for any political reason will be a reason for people to not use it or to rely in a more reasonable party.
I understand that US is migrating to a no friends policy, but I don't think its internal market is nearly ready to support it if clients start to go away.
Anecdotally, I talked to people in a recent event and the common topic was "what we are doing to avoid a blackout of US tech." People on big companies have reserved hardware and alternative software and are scaling down Amazon and Microsoft solutions at a fast pace. People on smaller companies are doing similar strategies but relying on a multivendor strategy.
I think this was inconceivable 3 years ago. Every company in Europe and Asian would gladly pay for amazon for cloud services and now they are using AI to migrate out the lock-ins.
We are seeing American LLM companies more and more relying on "secretive" methods to increase scores and this can only mean that they are not evolving fast enough. The gaps between the reported benchmarks and measured ones are increasing. It might be "safety" concerns or whatever, but I don't buy it.
Like Musk, Altman and Amodei have been overpromising every time they need more money and each time they need more money and now they internet is poisoned with their own content and they can't even trust any sources anymore.
Why? I don't see any reason for that to be the case.
Point is taken - competition is fierce and will be biting everyone's ankles.
Chinese models are being banned, will continue to be banned by USG. Even the Chinese govt is restricting external access to models now. It starts with banning Govt access to International AI, but will continue to consumer and business access.
I wish they were, but I always wonder when people chime in with Chinese models what they are actually doing with those. I have an OpenRouter account and give various models a shot every other month, wishing for an alternative, but I always come back to Claude. Even more so since the switch from Opus 5.1 to 5.5. To me, personally, it's unmatched. Like a year ago it was very different, and I would use other models all the time. But honestly at the time they all felt kind of "meh" still. I just hope everyone is at a good baseline in a year from now on, so I actually have a choice. And then, as you stated, things will get interesting.
tl;dr: I agree
In my case: robotics, complex document analysis (typically 800K context).
So it seems like Sonnet 5.5 medium, GLM 5.3 flash and Deepseek V4.1 flash 1017 are all in a similar space in capacity and price, however the $20 anthropic sub will get you like $200-$300 in sonnet 5.5...
Another bonus is that sonnet uses a fraction of the overall tokens and finishes much faster than the average chinese model, so your work goes faster (Sonnet 5.5 medium in 2 minutes, Deepseek 1017 in 6.5 minutes, GLM 5.3 flash in 13 minutes...)
I still don't get why paying for chinese inference makes sense. I should try Haiku 5.5 high/max because it's posting higher numbers than Deepseek V4.1 flash for half the price
However even with a high tier gaming machine you can run one of the Qwen models and it will be good at coding... like having an intern with boundless energy that will sometimes make mistakes but in general save you a lot of time. And for automating tasks involving scanning documents or creating summaries it will be brilliant.