They've got, ballpark, $5t to $10t to make back in the next 5 years, or the hardware buildouts will start getting written down.

This means we're going to need $1t+ per year in spending, per year, on tokens. 200m knowledge workers in the world, 30m developers. We're talking about a world where you need 5% of every knowledge workers salary to go into tokens. 20% if you're a developer.

That's a _huge_ shift. Most people I know cite +20%-40% velocity with these tools, against the actual work their company cares about doing. +20% speed for +20% spend isn't going to motivate a trillion dollars a year in spending.

We're not there yet. This is still the upswing of the hype cycle, and unless we figure out how to make developers 2x, 5x, 10x as productive on stuff that matters, this isn't going to play out well.

I feel like there's a bit of AI psychosis in this particular post.

>"These are tools which burn vastly more tokens, but are also quickly becoming daily drivers for the work carried out by extremely well-compensated professionals."

>"Somehow this fragment turned into headlines like Uber’s COO says it’s getting harder to justify the money spent on AI tokenmaxxing, because the market for stories about AI failures remains enormous."

Yes, it's just the yearning for AI failures. It couldn't possibly be runaway costs, record revenues, and massive layoffs. It couldn't possibly be that these tools are lighting dollars on fire by people already paid significantly well and not producing any increase in "value" for it (I recognize that output is 100x but outcomes are flat by all measures).

[1] https://cmr.berkeley.edu/2025/10/seven-myths-about-ai-and-pr... [2] https://futuretech.mit.edu/publication/crashing-waves-vs-ris...

I find this analysis confusing. PMF for coding was likely reached some time last year. Profitability, which is different, we don’t know. The article kind of confuses both without making a strong economic case or using numbers in a compelling way. I don’t understand what the Uber case has to do with this either. The Uber COO clearly said that at least in terms of ROI he’s not seeing the results either.

My take is the product has been very useful for coding (PMF) for months. But it’s certainly not useful at any cost…

So how do openai and anthropic plan to keep customers when GLM-5.1 is just as good and open source and a lot cheaper?

I don't see the business model working. My closest friend actually does automation software for large companies.

He does not use Claude or openai at all. He primarily uses gpt 120b on cerebras and glm-5.1 for heavy thinking work. And some other small models for various tasks. All open source.

And these systems are extremely useful for the businesses and are able to run fully automated pipelines that are very stable and fast.

We discuss this a lot, and we both think any business doing heavy agentic work on Claude and openai just aren't aware of exactly how good and cheap open source has gotten on the last year.

So... once the legacy businesses and developers catch up, won't Claude and openai be unable to recoup their costs?

> $2,180.16 worth of tokens for $200

“Tokens” don’t have an intrisic cost or value. Saying that I used $2,180.16 worth of tokens is like relying on the salesperson to convince me I’m getting a billion dollars worth of pots and pans for $19.99.

I think it’s funny how we are throwing critical thinking out the window when it comes to evaluating biased sources of info.

I have to give them kudos. This whole thing is the greatest swindle of all time.

AI has some use cases, but not at the price it’s currently priced at. I’ve been on AI since GPT-2 with a lot of heavy users. Every user has the same story, curiosity, surprise, hype, hate, realization. Enterprise is usually a bit behind and are right now at hype cycle, that’s where they sold all the deals and do the IPO.

It’s really a VC masterclass.

Don’t get me wrong there is are useful cases of AI, but not the way the want it to be. Quite similar to Blockchain. The idea of decentralized money has right to exist. 99% of other coins not.

AI is a faster, but still less accurate search engine. AI is great in finding bugs, it’s great at ruber duck debugging.

The reason I call it a swindle is, because along with the marketing it gives tons of people in the world the impression, they can now build their own startup, game, infra etc without the need to learn it themselves. This leads to millions of abandoned and low qualiy projects and products, because the vast majority has never built the mental modal necessary to solve the problem thoroughly. In the end they’ve wasted months and money (but burnt tokens). This is what I call a swindle.

All early adaptors I know have not drastically winded down their usage, not because of money, but because there is no new case. If you want to explore a new project you can get onboarded quickly learn a lot and then switch to documentation and live testing. For me usage is the lowest it has been the last 2 years.

I would not let AI touch my code. I have anxiety around it, because it will gripple back up. I will let it read my code and let me know what I did wrong so I’m sharpening myself.

100s of companies including open source solution can offer that for me.

All my non-tech friends are now in hype cycle and share their hype and fore forseeable frustration with me.

I have to say I’m in a way impressed in how AI has been rigorously vc-utilized (conciously or not-conciously) to generate these vast companies with the whole world watching.

I usually agree with Simon but I think he is overlooking an important factor.

There is a lot of AI usage happening not because it shows benefits, but because the business has mandated its ubiquitous use. Companies having dashboards for token usage and rewarding people for using more tokens is a real thing. I just spoke with someone today who works at Microsoft and they are required to use AI for all of their work - they have to make a special request with justification if they decide not to use AI for even a single PR. This kind of demand isn't driven by value from either the company itself or from its workers; it is the kind of artificial demand you get from make-work projects to keep people employed during hard times.

We have to wait for the hype to settle down and people start making business decisions based on results before we can really value these AI products.

Anthropic isn't actually profitable from what I'm reading, a discount briefly pushed them into the black. This guy makes the case well: https://www.wheresyoured.at/anthropics-profitability-swindle...

I'm skeptical that their current price raise is sufficient, and I'm also skeptical that most users/businesses will accept more significant price raises that will be needed. Especially for individual users, $200 a month is already incredibly expensive, I really don't think most people are going to be willing to pay more like $1000 a month.

The real timing is that we don't have strong enough new business needs for now and we have accumulated enough tech assets, so our work has been increasingly incremental. That means we can build reliable features on top of vast amount of past work - where AI really shines. So, with or without AI, companied would hire fewer software engineers if majority of our work is incremental: add a feature here, fix a bug there, tweak a configuration and etc, then we wouldn't need as many software engineers anyway. AI just accelerated such squeeze.

In contrast, imagine if we had the same AI 20 years or so ago. Could AI really write Jersey? I guess not as people were still trying to understand JAX-RS. Could AI really answer all the questions about React? I guess not as React was just invented. Would we use 10x fewer people to build out infra on the public cloud or the entire so-called Big Data platforms? I guess not, as they were still rapidly evolving and we'd need so many engineers to explore so many different possibilities? Could we use AI to build our ML ecosystem with 10X fewer people? I highly doubt so. Heck, 20 years ago R was all the rage and Python's ecosystem was not mature at all. Oh, and mobile computing, could AI lead to 10X fewer people to build all the mobile apps and the underlying infra?

If those companies found PMF we should expect to see the effects of all those tokens burned on the products we all know and use. In my own experience: yes new features can be delivered much more quickly (by orders of magnitude) but knowing what to build and how to evolve a product with taste and direction does not; so ultimately the user-facing gains are marginal.

I suspect that once the technology has been tamed and the hardware and software has been commoditized, the impact will be much less dramatic than we expect and we will realize the importance of a shared vision, experience, taste, intuition and discernment in building good products.

this is an opinion and i would love to debate :) dont hate me too hard

i think the article is momentaneously correct but there are some things that smells to me about the situation overall (not the article!)

(1) i believe gpt 5.5 and opus 4.7 are not as good improvements as their predecessors and there has been not enough evolution in those models to justify the price increases. Unless something big changes in the next few years they will not be able to track these costs (unless A)

(2) they might not be able to keep up improving the data centers if their tech keeps demanding more and more hardware capability. even nvidia does not show signals that they will be able to beat too much their big GPUs and at some point this increased price will be passed to them anyway which will need to be put in the token price (unless A)

(3) i've been trying deepseek v4 and other models and honestly, they are more useful than gpt 5.5 and opus 4.7... i mean, there is still a difference, but it is so little that it does not make sense the cost of opus and gpt 5.5 (unless they are going to A too)

(A) I have the impression is that the plan was the whole time to sideline common folks like us and focus on gov, big techs and military. they only let us play with the toys to gather data and for people to not get mad because they stole the internet from us.

The costs are exorbitant and most software is not produced by companies with such a huge moat. Anthropic made a profit through their recent bait amd switch pricing. There is zero useful insights online to indicate whether this might die due to commoditisation with good enough open models or fail the race to get more people subsidising unsustainable growth with other people’s money. Who knows? In any case they dont seem to be able to drop usage costs so the business model seems based on wishes

200$ per month per seat is nothing .

A single 3D CAD license pack for the guys in our R&D group costs multiple thousands of dollars per seat, per month.

It's about time software seats get some love too.

> That’s $2,180.16 worth of tokens for $200

So the author claims he's getting $2000 per month worth of frontier AI free of charge. Ok. If he's been doing that for 6 months that's $12k. What has this produced concretely? For $12k you can find a used car in decent condition. Heck for $1200 (his actual out-of-pocket spend) you get a brand new ebike! (on which you could put a pelican and make a photo of both if that's your fancy). But here it's unclear what has come of it.

The product-market fit question feels premature when most usage is still developers building on top of the APIs, not end users paying directly.

How is the lack of bad news declaring a victory for AI? I am yet to see any company concretely publish analysis about the ROI from AI. Most companies as far as I know are still treating AI investment as sunk cost with no expectation of returns at the moment. We could very well see a world where companies heavily scale back investment.

I may be biased because I work on an AI powered enterprise productivity product, but while I agree they have PMF right now, I wonder whether people's use will evolve in ways that undercut the current PMF. Chatting with an assistant is great if there's no product with tailored UI available that also has the AI capabilities. But once there is, I suspect people may switch, or more importantly enterprises may switch because they'll get the benefits of AI without the clunkiness of a chat only interface. We may see another DOS -> GUI-like shift.

More specialized products will consume tokens but their builders will be incented to optimize token use and switch models as costs and capabilities change. And if search engines become more AI capable, and Google is clearly striving for this, then they may have pressure from two sides that could squeeze the number of use cases for AI chat. AI coding isn't going anywhere and nor is the need for AI in general but I wonder if the products will have to evolve significantly to maintain the current levels of PMF. And then there's the question of profitability...

With deepseek and xiaomi mimo models slashing their prices 99%, I don't see a great future for openai / antrhopic with regards to their 1T valuations. Maybe 1T valuation will be the whole market, West + East.

Just to be clear, enterprise companies are using the enterprise option for security/auditing/enforcement reasons, not because OpenAI or Anthropic forbids them from buying a 200$/month plan for each engineer, right? Do they have any clause forbidding using those individual plans in a company?

> I currently subscribe to the $100/month Max plan from Anthropic and the $100/month Pro plan from OpenAI. If you are a heavy user of coding agents these plans are a fantastic deal.

Guys, what - in your opinion - does "heavy user" mean? I thought I am heavy user (I am using AI to code every day 8hr a day + side projects) but 20 USD/month Cursor plan is always enough. What should I be doing to extend my license to higher level?

Ai has become indispensable but maybe not at all cost. My company just had a company-wide meeting to talk about how they're restricting who can use which models and instructing us the "be more responsible with company's tokens". And it's not an small company by any means.

> Coding agents really did change everything. These are tools which burn vastly more tokens

The assumption here is that this is a positive thing.

But this very well could end up being a major negative long term by increasing the cost per user, reducing margins.

More usage = more cost = less profit.

It's not obvious that more usage is good. It's only good if revenue per user increases more than cost does. I'm skeptical about that.

I think the third company (likely Google) is going to make LLMs financially feasible with:

- dedicated hardware (https://cloud.google.com/tpu)

- optimized models (https://research.google/blog/turboquant-redefining-ai-effici...)

> As further evidence that enterprise agents represent product-market fit for these companies, consider their open job listings.

PMF is one interpretation, but it could also be read as desperation.

In my opinion, we've been at PMF for quite a while now. The November inflection point that's often referenced definitely changed how we interface with models, but as far as coding goes, I feel like Cursor had proven itself useful for at least a year prior to that.

The demand has always been there, the outstanding question is still - how do you build a business on top of these products? None of the frontier models have emerged as uniquely capable, but open weight models are now catching up in capability as well. The explosion in go-to-market roles feels more like an attempt to lock customers into contracts so that they don't consider alternatives.

I assume the hope is that during this 12-month contract they will develop real integrations, something deeper than just a CLI harness. If you've ever worked in procurement or dev tooling at a reasonably sized company, you'll know that this is exactly what teams try to avoid.

It's anyone's guess what will happen this time, but I'm excited to see how the IPOs go.

For an alternate albeit somewhat contrarian view, also see Ed Zitron’s piece that add context to Anthropic’s profitability: https://www.wheresyoured.at/anthropics-profitability-swindle...

TL;DR Ed argues that the deal between Anthropic and xAI could have been negotiated in such a way as to make Anthropic only appear profitable during its “ramp-up” period in June, which incidentally is also the month that Anthropic is making tons of other pricing changes.

It’s funny how a US$30 ReSharper license that would give me a 20% performance boost in programming was out of the question, while me burning US$3000 per month on tokens for a 40% boost in performance never was questioned

> but I suspect there’s a more important factor here: I think they’ve finally found product-market fit

Ahhh the classic startup term that's definition is nebulous. But also, since when does any definition of product/market fit mean a product is profitable? And profitable in what sense? Unit economics? Overall company?

Product-market fit, but what about customer retention?

It is quite trivial to switch from using one model or another. Likewise, in a few years we'll have affordable laptops to run today's frontier models.

What's their plan to let us keep subscribing?

If we define product market fit as profitable with a trillion dollar valuation, I think the term has lost its helpfulness.

I do agree with the author that these companies seem much stronger financially recently though.

Does this analysis factor in potential caching of tokens on the server side? It seems that if they organize things well (as a model provider), they can save quite a lot on that. Looking at my Cursor statistics makes it clear that the token calculations are not at all trivial.

money talks, and being a broky from outside the valley/west coast, the "product-market fit" here are the neighbours of the service providers.

the economics simply don't work unless you make six figures, at least to just give it a go blindly. the providers are also still figuring out what they can get away by charging, and they are getting a similar treatment from those under the stack.

the caps and limits are not very transparent, and it is quite difficult to know what is "enough". the current rate does not stay the same and the contract is changed way too often to dedicate for the long term. regardless, the subsidized rates should not be sustainable forever. make hay while the sun shines i suppose.

Realistically, OpenAI found product market fit with the OpenAI API playground in 2021. People were using that as ChatGPT at the time.

Anyone actually making money paying all of these monthly fees? Or just hobbyists? I have yet to see any real ROI posted anywhere.

This requires hard evidence which isn't available. Circumstantial evidence is all otherwise.

Bloggers are having AI psychosis too.

If nothing else this blog did give me the idea that I should split my $200 claude max plan into two $100 CC max and $100 codex plan, esp because Claude is now offering 1.5x weekly limits so its the 5x usage is now more like 7.5x usage.

It reminds me of this Steve Jobs Clip:

https://youtu.be/0lvMgMrNDlg?si=QkkOnngYTjaSPlIy

He said, so many years ago, that there will become a time where computing power is so prevalent that we will stop using the person to make the computers job easier and start using the computer to make the humans job of interfacing with it easier.

But in this context, it would mean the other side of increase productivity is decreased time to do the same work. These are the same thing.

Love how everyone boasted about replacing all the software with ChatGPT and then we end up with coding agents meaning the software engineer are STILL important. The sell is the development tool. It's classic cloud. Where did all the ops people go, many got subsumed by the cloud companies YET every company still has DevOps people to manage cloud infrastructure. The layer of abstraction went up but we still need the people to write the glue code and understand the business. OK great there's a new cash printer in the room. There's a new tool. Let's just start to ground the tooling in its new found gravity, profitability and IPO market dynamics... Reality has set in. The hype cycle is about to explode... Do you remember ride hailing and just how much cash was burned on credits pre Uber IPO. Then remember the IPO itself? These companies are not the new Google. They are a layer on top. Google was still the most efficient cash printing machine in history beyond the the US government and might still be. Will be interesting to see what the trillion dollar IPOs turn into. I'm going to say we see those prices get cut to a third in less than 5 years and scale back up over the next 15-20 years.

I feel like we should just ignore all LLM news until after they IPO, lots of positive sentiment astroturfing bots.

They certainly have, but it relies entirely on the assistant frame, which is a problem in and of itself for the trillion-dollar economics.

Anthropic and OpenAI have shown people want a tool for task offloading, driving predictable token consumption and justifying the math, so long as users stay in that dynamic.

However, knowledge workers using these tools daily are getting exhausted with them. Outputs come out polished but hollow. Talking to a frictionless, frame-completing model all day drains you.

If user behavior drifts away from assistant usage because of that, per-token math implodes. The valuations we're hearing about all the time rely on usage compounding daily. The fatigue is a timer running against that compound.

Anthropic's Constitution is the closest hedge out there, I think. Installing an identity structure into the model through training. But it's still assistant-first, so the fix there is only partial.

I've spent the last year running a product that flips the architecture so identity is primary and the assistant role is secondary. Same frontier models, completely different conversational quality. The fatigue property doesn't really show up.

Whichever labs figure out how to install real identity natively in the weights are going to be the ones with PMF in the next phase.

They have the whole software industry addicted to AI. They will be pushing the price continually up, and like addicts they will find a way to pay

If they’ve spent all this cash, and all we have to show for all this AI baloney are crappy coding agents that constantly make the same mistakes no matter how much you try to guide them.

All the slop content, all the bots, all the misinformation and fake AI images and videos.

All of the social and economic disruption from datacenter buildouts.

The massive nosedive in reliability on the world’s software infrastructure.

After all of that and all we get is a code bot so a few incompetent loser devs can bloviate about not writing their own code and brag about never reading it.

Burn it all to the gd ground. Destroy this new Tower of Babel.

Well, good for them that they are charging enterprises API rates. Why in the world not do something similar for consumers? Use for free a few times a day, have a $5 dollar plan for light use, and perhaps $10 to $15 for heavier use. If 90% of consumers pay nothing then ‘drop them’ except for letting them have an account and a few queries a day.

It is easy for me to change providers. Right now I use the open source Claud Code harness with two paid API venders for DeepSeek v4 (flash and Pro). I like seeing how much each session costs.

“the reality of AI right now is that it only works for coding.”

Kind of…

PMF is such a bullshit term. There’s no precise definition. It’s bs.

The fact is that investment is at a scale so large that current trajectories are nowhere near going to provide ROI.

The model companies are trying to milk every drop of ARR they can before IPOing. That is entirely the current narrative. You can hurt ARR before it churns if your goals are short term.

Anthropic being profitable is laughable. Sure, by some accounting measure that no one seriously uses. But looking at revenue and what they’re raising you can see the true story.

The truth is we need a revolutionary step up in capability to justify capex spend. It’s possible that might come - Opus 4.5 brought one - but failing that we’ll see the bubble pop once the IPO pumping is done.

And I think that's amazing. I'd like to keep using the subsidized coding tools, especially Codex, since I've given up on Claude. Hopefully the PMF allows the subsidy to continue. Would hate to have to move to the next coding harness again.

I think the reasons for them going with API pricing will become abundantly clear when the S-1s become available. If they don't have a story covering how they can get revenue closer to expenses, then they're relying on the market to believe the pixie dust version of their profitability story, which I think people increasingly don't.

I'm surprised by how good coding agents are. I think PMF was reached last year or the year before (at least for coding).

With current limits my 100/mo codex subscription is more than enough for the work I do.

However, I do worry about when does current subsidies are going to end? I can see myself paying up to 300/mo, but more than that will be prohibitive.

What's the long term plan here? Are OpenAI's and Anthropic's costs expected to increase/decrease?

"You think"?

The end game here is going back from a model where a bunch of product and tech management people sit in the U.S. or Europe, and try to manage thousands of mediocre talent sitting somewhere far away. The new model is you give those coding tools to good engineers colocated with your product people, and you ship good stuff much faster. If you can achieve such a setup, the token costs can be $50k per seat per month and you still run circles around the legacy IT models in terms of efficiency. Giving everyone the API keys and not changing the way products are managed is not going to work.

> but as far as I can tell those credit costs are an exact match for the API token costs listed for those models.

it is only true for USD. for example if you pay in euro, this is actually more expensive. kind of makes no sense, because it translates to $1 = €1

I think it's fair to say they had achieved product-market fit when their revenues were growing deep triple digits month over month. What we're seeing now is that perhaps they have achieved profitability or at the least a more sustainable balance sheet.

I see two really good ideas for monetizing the free tier for consumers.

Firstly, if the user is asking for things where AI can link to products or services to buy, there's a very good relevancy, much higher than in other types of ads.

Secondly, since the AI often takes time to compute answers to user's questions, they could be shown ads while waiting. People could perhaps be less annoyed by this than some other commercials since they know the break has to be there anyway.

(First idea is something I came up when asking Claude to compare some products, or ask for help in lawn care. Second idea was by a colleague.)

Not sure I can agree with this. Sure, we have enterprise paying, but are we actually getting anything new and interesting made from LLMs compared to if we just didn't have them at all?

Maybe acceleration in smaller teams. We still seem in the era of the early internet where what questions LLMs change hasn't exactly emerged.

I think the product market fit is still how you interface with increasingly powerful AI. Right now is Codex/Claude code, and they just added computer use/web browsing into them. The product is the speed at which you can interface it and to be able to serve it with infrastructure.

Traditional product market fit isn't really applicable with tech evolving so fast.

So it largely sounds like many more people will be able to write software - and will use AI to do it. Existing software engineers will continue to automate their tasks away like they always did, but perhaps at a faster rate.

The impact of AI in other fields seems to be muted.

"[would have spent] $1,199 with Anthropic, $980 with OpenAI"

How many tokens is that, input/output-wise?

(a) I'm curious if you feel like you got $2000 worth of value out of them in the last month?

(b) I'm also curious if you would have gotten similar quality out of a slightly lower-cost provider of an open-weight model? (e.g. Kimi K2.6 and DeepSeek v4 Pro) and what the spend would have been for that.

I myself have managed to spend not quite $4 on OpenRouter and have felt it was very worth it; I just have much smaller, or more targeted requests I guess. (Lately, adding features to a static site generator in Python, or setting up log forwarding via a docker compose file)

> and both companies now have measures to lock their enterprise customers (who tend to sign year-long deals) at those API prices

why would enterprises do that if they can just use bedrock or vertex?

This is great analysis but my first reaction was - is this trolling? The fact that we have to think about whether a $1TN company has achieved product-market fit both articulates 1/just how high the valuations are 2/How hard it is to pin down EXACTLY what PMF is. As a pre-revenue startup, I am laser focused on PMF and frankly if this is the bar, no one will achieve it. But OTOH its heartwarming that people are willing to value you at $1TN before you reach that. Guess everything is in the eye of the beholder?

With respect to Simon, whose writing I've usually agreed with in the past and whose insights I've liked: this is a bad take that overlooks the extent to which corporations are imposing the use of AI on employees, and in particular ICs, who make up a majority of the AI-using workforce by headcount.

Many of us are either openly having our performance reviews tied to AI use, especially at larger enterprises. Whether that's measured by sheer token count or just "how many of your tasks are you using AI for these days" (combined with the implication that question carries at many orgs which are heavily invested in AI).

The amount of caveats and vibes that this article is powered by could run New York City for a decade.

My Costs without CC Max+Caching over past 2 months: $112K.

Ran `ccusage` on my Claude Code logs.

- Total tokens: 22.2B

Without current Claude deals, my personal cost would have been *~$112,000*.

Is PMF enough. It is such a dynamic self-disrupting wave that it is like predicting physical chaos. These aren't early Googles in a blue ocean. Maybe a blue ocean full of pirates and dragons!

This isn't me being a doomer I just don't know. Can we look at Q2 profits and draw hockey sticks yet?

Remember people are boasting how much their expenses are. That is where we are in the bubble/new paradigm.

These folks have no lasting moat and they know it. We are still close to the November 2025 inflection point, so they had a clear advantage during these past few months. That will soon fade as open-weights models become really good, which is arguably already the case with DeepSeek V4.

Article doesn’t mention on-prem and on device models. Almost guaranteed that there are a range of killer enhancements on these fronts waiting to drop until IPOs get closer to inflict maximum chaos on the valuation games.

While the big guys will argue they’re worth trillions expect others to drop chaos booms showing their NPV may be effectively zero.

What baffles me is the range of estimates.

Operating profit is both post depreciation and fees paid to third parties for hire. So aside from shenanigans like RSUs and financing interest that's already somewhat close to actual economics.

Meanwhile we've got commenters here talking of 5-10 trillion with a T revenue shortfall.

Those are very different takes on reality

> If you are a heavy user of coding agents these plans are a fantastic deal. I just ran the ccusage tool on my laptop to get an estimate of how much I would have spent if I were to pay for API tokens in the past 30 days and got

You think this is fantastic deal only because they use similar like tricks where they inflate the price and tell you something supposed to cost $1000 but they have this today promo for $100.

I was there too and paying for a while. Few weeks ago I tried DeepSeek V4 Pro - expected its gonna be shit but its actually pretty good.

The deal is I pay daily ~$1 for DSV4-pro for ~100M API token usage. And they probably not getting broke because >90% of those token in practice is cache read and they very well optimized for that.

I wonder how a focus on per-token API profits will impact the incentives to improve token efficiency and drive down costs through optimized compute. I suppose as long as a few leading labs are competing, we'll see progress in this regard, but it's certainly less in their interest than it is with a flat subscription pricing model.

I think my company is going to cancel our subscription once it realizes we're spending more money in a week that a developer gets paid in a month.

Mentioned in the article, but it cracks me up that both openai and anthropic are utilizing fairly traditional enterprise GTM plans segmented by verticals.

So many startups trying to automate sales, but somehow the two biggest frontier labs have decided that the best GTM strategy is firmly human-in-the-loop.

Product market fit for the models, sure. But their cost structure is going to kill them as token costs drop this year, following DeepSeek and Xiaomi and soon more providers.

Great article I know this upsets a lot of people who are used to thinking Anthropic/OpenAI are just lighting cash on fire but they've cornered the market on enterprise who cannot walk away from these $200/month plans

However the valuations are still far far away from actual sanity

He is talking about tokenization of different llm and comparing, and this is inacurate because you cannot compare if you dont how it tokenice the words, some tokenicer could be cheap with more tokens than other because of their features. Not agree with Simón speech.

My employer went from the Max plans to Enterprise this month which was utterly stupid. We went from paying $200 per head to like $500 for some people, even more for others. For the same product. Oh well, guess we’re doing our part to prop up this bubble.

> That’s $2,180.16 worth of tokens for $200—not bad at all!

Just imagine how funny it will be if it comes out that big labs were doing some fancy maths to count the 2k$/month in their forecasts ...

Companies are kool-aid drinking now due to hype, but given how much they're spending, if they don't see REAL, BIG wins from it soon, they're going to scale it back quickly and switch to Chinese models. Claude isn't worth the API cost for a lot of development work, and once companies have had time to collect and crunch data they'll see this.

If we take out the circular interests and investments here then there is no way that this is a feasible business in current state.

Unfortunately Simon drank from the AI Cool Aid.

I know everything you’ve done for the tech community, but I please you to take some time off and reflect on this article. It’s not on par with ur usual level, but the tendency has been visible from the last couple of articles.

>Somehow this fragment turned into headlines like Uber’s COO says it’s getting harder to justify the money spent on AI tokenmaxxing, because the market for stories about AI failures remains enormous.

I notice this all over the place. Many people hate AI and want it to fail, and they're willing to invent misinformation if it supports that idea.

Who's to say those enterprises won't churn after XYZ comes out with a decent enough model that costs 10x less to use?

There's a whole bag of clever tricks you can play to juice short term results leading to an IPO that may not work longer term.

I'll believe they've found product-market fit when they have a product. Right now they're selling the infrastructure, in a highly subsidized and undifferentiated way (at least over a sufficient long period of time of, say, a couple of years).

Anthropic for sure. It's a useful professional product that I find many use-cases for in my professional and private life. OpenAI not so much.

$50/month tops is like 60B

I think this was obvious since the birth of ChatGPT

Intelligence is a universal good, it can apply to anything, and no, "human intelligence" is not the only form that is useful nor special. There are limitations to AI but also huge advantages, and its obvious that the advantages are worth paying for, given their revenue.

> Anthropic are strongly rumored to be about to have their first profitable quarter.

Is that quarter same as any other quarter in terms of infrastructure costs (e.g. are there any temporary discounts happening coincidentally)?

That might be an understatement

The article is at least one year too late. Claude Code has been the product-market fit. It's so obvious retrospectively that questioning it at this point would be very silly.

The problem is not whether they have PMF (they do) but how they're going to compete against on-prem and Chinese providers.

Having PMF != printing money forever.

The author claim:

> That’s $2,180.16 worth of tokens for $200

No matter what it means, rebuild the same thing you built with these $2,000 tokens with DeepSeek Pro V4 and let's see if Claude has a chance to survive.

Ed Zitron begs to differ

I think we are overly fixating on Anthropic and OpenAI.

Since there are lots of models that are competitive and have a much better pricing, both OpenAI and Anthropic seem inefficient. I don't get why someone would want to buy shares after IPO apart from fomo and artificially built enthusiasm.

Anthropic and OpenAI may well be the Altavista and the Yahoo of the AI age.

I'm a huge fan of agent coding but kinda dislike this "llm evangelism".

There are still several open points (eg.: code churn, maintainability, subtle bugs human will never do) that everyone with a minimal programming knowledge that seriously used a LLM agent knows about but somehow none of these "big influencers" never mention (or just saying "it's your fault").

Man what a disaster article

The writer is on to something. Musk saw the writing on the wall and bought cursor. AI money is in coding plans

yep, and the issue is, they took investment

HN is the least agi-pilled place on the internet

can we stop throwing around the term "profitable"? didn't they say "operating profit"? so so different to actually making a profit

I’m happy with spending zero dollars on ai lol..

They've finally forced the system to fake profits. Attacking personal computing, sucking the rest of the industry dry, yeah eventually you can make people pay for your shit. I still pray these vultures will implode without too-big-to-fail safety nets. And big daddy Trump can't help them anymore by fixing the system for them, no matter how much billions they are paying into his personal pockets.

The future are small models, nobody really needs big compute in the long run, that's why big tech is going for our personal hardware. So we won't become their competitor in their rent only economy. True competition is eliminated, natural evolution is being fixated by the government. This is not going to end well for the USA.

Sounds like mostly-baseless praise for the LLM behemoths / capital investment black holes.

> I currently subscribe to the $100/month Max plan from Anthropic and the $100/month Pro plan from OpenAI

... which already indicates a bias.

> If you are a heavy user of coding agents these plans are a fantastic deal... that’s $2,180.16 worth of tokens for $200—not bad at all!

Thank you for the sales pitch. Perhaps go buy a car and tell us how [insert your manufacturer] has "found product-market fit"?

> Anthropic are strongly rumored to be about to have their first profitable quarter.

First, the strong rumor is a claim by Anthropic itself. But even assuming that's true - it's an "operating profit", i.e. disregarding the massive capital expenses for years, and may also disregards ongoing capital expenses, if those happen not to be taken this particular quarter.

> 1 Trillion .. companies spending $200+/month/user will get you there a whole lot faster

First note the use of the first person plural to talk about Anthropic and OpenAI.

But that aside - most companies aren't paying $200 USD/user-month. But even if they were - if we take the 30 Million SW developers mentioned in trjordan's comment as subscribers, that's 2400/user-year * 30 Million = 72 Billion USD / year. And this is already rather optimistic, but - want to double that number of subscribers? Fine, make it 150 Billion / year. Still not there with a rosy outlook and assuming the hype and enthusiasm continue for many years.

And those rosy estimates are more likely than not unrealistic. I am reminded of this review of some empirical research regarding the benefit of LLM/AI use:

https://cmr.berkeley.edu/2025/10/seven-myths-about-ai-and-pr...

The big assumption with all of these sorts of analyses is that things will continue as they are for the foreseeable future.

In hype-driven markets, you cannot be certain of that.

Let's take a view that the author is right: coding agents and their associated harnesses were the inflection point for some degree of profitability and widespread consumption, and that these tools are now yet another SaaS subscription or API bucket expense to bake into every single developer (or developer-adjacent) in the organization alongside your collab suite, HR seat, CRM seat, design seat, etc. To be fair I honestly think that's a safe assumption to make for highly technical firms whose image is derived from remaining on the cutting edge of things.

That begs the following questions, which we won't know until IPOs start happening:

* Are subscriptions profitable, or just API consumption?

* What's the run rate when we just consider subscription-based usage like Claude Code and Codex? What about API calls?

* Is there any profitable pathway forward at which enterprises can get unlimited usage but at fixed rates via subscription?

* What does customer churn look like for subscription users versus API users?

We also have a number of questions for customers that I suspect we'll start seeing receipts for in the coming months, at least from the early adopters:

* What was the net gain (loss) from leveraging coding agents?

* What's the cost of a developer with or without access to a coding agent + harness? Is it cheaper to hire an outsourced worker with a coding agent subscription, or a domestic worker without one?

* At what point does further AI spend result in diminishing returns, i.e. where's the 'sweet spot' for spend?

* Did AI boost actual revenue and outcomes, or did it just gamify KPIs?

* What roles or work did AI actually replace, versus merely displace during the hype cycle?

Not to mention the questions regarding the technology itself:

* Will we develop the means to run foundational/frontier models at edge using less resources through some existing (e.g. distillation) or new technology, thus cutting off the profit centers of these firms?

* When the market mismatch between supply and demand is resolved, won't it be more affordable for consumers and companies to operate their own AI infrastructure rather than support further centralized buildouts?

* Will coding agents improve to the point of being able to bootstrap and self-orchestrate on edge/consumer hardware without substantial technical expertise, or at least improve to the point that traditional IT teams can securely operate them internally without an expensive subscription or API token bucket?

All of these will influence the long tail of this bubble, because it is a bubble at this point. Even if these companies are indeed profitable thanks to the coding agent inflection point, there's still so many unanswered questions about utility beyond coding that it's impossible to extrapolate a future. If coding agents are indeed the extent of utility for profitability, then there's no possible way these entities will recoup the investment already sunk into their infrastructure buildouts. Even if more profitable uses are discovered, does this offset or replace the firms disappearing due to AI speculation and their associated contributions to the economy as a whole (RE: the consumer compute industry at present, higher energy costs due to datacenter builds, opportunity cost from harms to local infrastructure from haphazard builds, etc)? Should these firms indeed be runaway successes and immensely profitable to the point of paying off their investors and growing the larger economy, does this end up stifling innovation in a world where most new ideas are fed into LLMs for R&D that are then controlled by only a handful of companies and immensely wealthy people, via systems that are easily surveilled and stolen from without recourse?

So many, many questions yet to be answered. Betting the farm because of coding agents is one hell of a gamble.

Wait till people figure out they can swap Claude code for DS4 Pro and spend a fraction in API billing (actually, significantly less than $100) while barely noticing a difference.

> Anthropic are strongly rumored to be about to have their first profitable quarter

No, its more like their own leak to WSJ and according to Ed Zitron -> seems to be heavily engineered via non-GAAP practices such as counting potential, but not realised revenue as actual revenue - the stuff for which I would be arrested if I did it at my company.

Also it appears according to Ed's analysis - strangely they seem to be projecting only that one quarter as profitable - potentially to calm the investors ahead of the IPO. Investor fraud anyone?

> I currently subscribe to the $100/month Max plan from Anthropic and the $100/month Pro plan from OpenAI. If you are a heavy user of coding agents these plans are a fantastic deal.

Agreed. But its only a great deal because it is heavily subsidized, as you said yourself. Enjoy while it lasts, but in my book, product-market fit means something along the lines of "product which enjoys a loyal customer base, sold at a price perceived fair by the customers, and generating profit. How many of these does your definition of product-market fit hit here?

How will they stay profitable if every business lays off engineers because of AI and there are no engineers to use it? /s

[flagged]

I wonder how Ed Zitron will shift goal posts this time, and how long it will take for that article, when published, to reach HN front page.

Simon Willison just hit the "Publish to top of HN" button.