[nl]

Revenue is higher than cost of revenue and revenue is growing faster than cost of revenue.

We know OpenAI is forecasting $25-30B revenue for 2026. They will be very close to breaking even at those number.

Given Anthropic has forecast more revenue than OpenAI and we know has spent less on R&D (cite their desperate scramble for compute capacity!) the rumours of them being profitable this year seem very credible.

[rvz]

Imagine still saying this after open weight models being released everywhere and the likes of DeepSeek keeping their prices at the lowest possible for more than good enough intelligence that even Microsoft is migrating to deploy DeepSeek's models.

[nl]

I use DeepSeek. I'm very familiar with open models. I run a popular benchmark to check their progress.

DeepSeek is somewhere between Sonnet and Opus in capability, for much lower price.

Their $0.87 per million output tokens is one of the few API offerings that is probably subsidized (the break-even price is probably around $2 judging by[1])

I think Anthropic and OpenAI's margins will erode some over time. But I think they are very profitable now on the API prices they are charging, and their margins will remain healthy.

Put it like this: AWS is a very expensive way of getting compute, and there are numerous competitors that are much cheaper.And yet AWS is very profitable.

[1] https://openrouter.ai/deepseek/deepseek-v4-pro#providers

[xela79]

I find that Deepseek Pro to be below Sonnet 4.6 for much of the coding/review/executing tasks from plan. The only good thing it has going is the price. It will is cheaper for it to crunch through your well worked out implementation plans, it's slower, makes more errors, needs more tokens to fix those. And even at that increased spent it does deliver in the end. More tokens, lower cost, slower.

I use it as something that sits right below Sonnet for daily tasks that can be run in loops with validation checks, perfect. Anything more advanced...it just doesn't cut it.

Same experience with z-ai glm 5.1 btw... don't know what system prompt magic antrophic set in front of their Sonnet 4.6, but it does execute tasks efficiently.

claiming any of these deepseek models are close to opus is a bit optimistic

[surgical_fire]

I use Claude at work, DeepSeek at home.

For software development, I notice no difference. For all the work I did using AI, DeepSeek is as capable as claude.

[Alphanymous]

Have you tried other tools like windsurf? Or other free open source Ai?

[surgical_fire]

I never tried Windsurf, no.

I tried DeepSeek and MiMo through both Claude Code and Pi.dev. I found Claude Code a slightly better harness, but I still have both installed and switch around frequently.

As for "open source AI", what exactly are you talking about? Is it open models?

I tried some through Ollama, but I don't have the hardware to run decent models.

[scottyeager]

Did you mean Sonnet and Opus?

[nl]

Yes, edited, thanks.

[scottyeager]

Sure!

[NetOpWibby]

Microsoft has been dead last at everything for decades, so that point isn't surprising.

[Quothling]

Microsoft quietly won the AI race in enterprise with the addition of cowork to their copilot app. Being in the energy sector in Europe we're quite limited in what we can do because of things like NIS2 compliance. We have access to corporate AI tool though the equity fund which owns part of us, and while it allows you to create personalized agents that can run sub agents and use "skills" it's all done without any form of filesystem access. Being married to Microsoft because our IT loves that sort of thing, and having spend a decade in the public sector I sort of get why from an enterprise perspective, anyway, we have always had access to their Copilot app. Which has been so bad that it's actively turned people away from AI. Then last month we get cowork frontier, and now I'm in the process of helping everyone adopt it. Not only does it play directly into our licenses, it also has access to all the Microsoft 365 stuff, so that our HR can use it to sort applications into the categories they belong in and what not. Sure they could've had a better on-boarding system, but they don't so someone has to go through the emails and sort "financial controller" from "sheep shepherd" (yes, we have sheep on our solar plants).

Anyway I'm rambling, what Microsoft is doing with AI in enterprise is basically what they did with Teams and similar systems. They provide a platform for it which is good enough that your organisation is going to want it rather than deal with multiple vendors. Not for tech organisations, but for every other enterprise organisation it'll be so much easier to just go this way. I imagine that Anthropic is getting some sort of payment from Microsoft for Cowork, but what Cowork shows is that Microsoft can be completely model agnostic and still sell "top" AI. Especially because they've set cost on a fixed rate that I'm sure they'll increase by 25% every year.

Or do things like the fact that you need some sort of special Agent 365 license for your sysadmins to manage the admin.microsoft part of Copilot which has to do with security policies... Ask me if it was fun doing that agent by agent... It's frankly the most Microsoft thing I've ever seen.

[nullbio]

I think OpenAI will still maintain the lead for at least another few years.

The cost of hardware still needs to dramatically drop for open-weight models to be viable for local usage. Even with the release of things like Nvidia DGX Spark and Ryzen AI Halo, you'd likely want a few of them to run agents in parallel.

Sure, you can use cloud hosted variants of models like DeepSeek etc at API rates, but subscriptions still come out on top for bulk usage. GPT is already tightly integrated into peoples workflows, has wide adoption, has good tooling for developers, etc.

Plus there's nothing stopping them from competing on a price level if they really feel the need. It just means they might burn more cash in the short term.

[zozbot234]

> The cost of hardware still needs to dramatically drop for open-weight models to be viable for local usage. Even with the release of things like Nvidia DGX Spark and Ryzen AI Halo, you'd likely want a few of them to run agents in parallel.

It's more efficient to do the opposite on a constrained platform. Run agents in parallel using a single model, then round-robin among models for cross-checking purposes. (The makers of local inference engines are dropping the ball by not making batched inference a first-class citizen of that workflow. It's not just useful for vLLM and SGlang.)

[LeFantome]

Looking forward to GLM-5.2 in OpenCode.

But a lot of other people are going to happily shell out for Opus 4.8 at way higher prices.

[SyneRyder]

I think it's out now - or do you mean as part of the Zen subscription?

The API GLM 5.2 launched last night with several providers on OpenRouter. I had a short conversation, didn't get to test much, but initial impressions "the vibes were good".

[psv2522]

I think op is on the go sub, i m on the same we have glm5.1 but dont see glm5.2 yet

[Daviey]

Been using GLM-5.2 in OpenCode since Saturday, what are you waiting for?

[w29UiIm2Xz]

What's the adage, it is difficult to get a man to understand something, when his salary depends upon his not understanding it

[nl]

My salary is in no way dependent on OpenAI or Anthropic (in fact I could probably get more money if we switched to using open models and they were just as capable).

[surgical_fire]

> Revenue is higher than cost of revenue and revenue is growing faster than cost of revenue.

That's not necessarily true.

It depends on two things that are impossible to derive from the leaked numbers:

1 - How much of their compute costs are subsidized.

2 - How much of that R&D chunk can actually be reduced for the company to continue working. What goes into "R&D"? Is that only things such as training for new models? This is impossible to determine from those numbers alone.

If those numbers are true, they would not be "close to profitability". They would be profitable, period.

This does not explain why they need to raise money like crazy. It would be possible to train new models at a more sustainable pace without the need to jack up prices at all, only with the difference in between "Revenue minus Cost of Revenue".

They also wouldn't need to be on a mad race to IPO and dump this into the public market.

To be frank, the more I look at those numbers, the more I think this is completely unsustainable.

[vkou]

Sales and marketing have to be added to the COR, which makes COR slightly > Revenue.

[danielmarkbruce]

No, no they do not. Those are operating expenses. This is accounting 101.

[hparadiz]

It's wild how basic math goes poof as soon as someone has an agenda to push in internet comment threads.

[danielmarkbruce]

The words have meaning in accounting. The basic idea of accounting is to communicate what is going on in the business in as consistent a way as possible. It's not a perfect system, but this stuff really is 101.

[mrcwinn]

That’s not how that works.

[gamblor956]

That's only true once they become public. As a private entity they can calculate COR any reasonable way that their investors are okay with.

(For tax purposes for most businesses it doesn't matter whether an expense is CORS or not.)

[nl]

To expand on this slightly - startups often argue that marketing expenses should not be reported as part of cost of revenue because they are temporary and will change as they grow and the market becomes more aware of them.

There are arguments for and against this - awareness is an issue for startups, but most large companies continue to market. It is true that it is fairly easy to change the amount a company spends on marketing.

Either way, as the parent says - provided the investors understand it there is nothing weird about doing it this way.

[danielmarkbruce]

No, they argue they shouldn't be counted as operating expenses because they are basically capex dollars. It's a completely reasonable argument. If I spend $100 million on s&m to get a bunch of customers to my saas, it's not different economically to drilling 20 oil wells at $5 million a pop.

[nl]

Yes that capex is a fair point too.

(To explain: It's basically saying they are spending $X now to get them as a customer which will yield a lifetime value which is some multiple of $X. That's a very valid thing to do - see the whole field of cohort analysis for SAAS)

[vkou]

An oil well doesn't disappear once you put it in the ground (The oil might).

If the marketing and promotions and discounts stopped tomorrow, that revenue will drop.

[nl]

Page 6 of Google's 10K statement to the SEC[1] they are separate items. The line items are:

Cost of revenues Research and development Sales and marketing General and administrative

Take it up with the SEC I guess?

[1] https://s206.q4cdn.com/479360582/files/doc_financials/2026/q...

[danielmarkbruce]

If you drill a dry well, under successful efforts accounting, you expense it. You also make zero dollars.

On top of that, oil wells decline, slowly but surely, just like... customers churn.

If you spend $100 mill on sales and marketing and get zero customers, I'd agree it should expensed. If you get a bunch of customers, it's hard to argue against the capex treatment idea.

[otterley]

Brand equity lasts long after advertising has ceased. Coca-Cola could stop advertising for a year tomorrow and people would still be aware of it and buy it.

[claw-el]

For publicly listed company, isn’t marketing expense in the ‘Selling & Marketing Expense’ part under ‘Operating Expense’ and not part of ‘Cost of Revenue’ as well?

[danielmarkbruce]

No, public companies do not include sales and marketing as cost of revenue.

[HWR_14]

COR is a GAAP concept.

[reinitctxoffset]

Anthropic has a durable advantage that OpenAI does not, as much as they choose to squander it to pump the numbers.

They were first on a few things but the tell is the coherence of the thinking traces of Claude. You have to put a loss on that. GPT 5 series thinking traces are creepy, Gemini thinking traces are disturbing. They both represent forced discontinuities on the policy gradient.

Claude is good at tool use because it's gigantic and well-labeled, but the reason you pay the premium is for a thinking partner, not a tool user.

Claude Code is the cancer that will kill the patient, Boris is the the Kardashian version of Karpathy, with less business sense.

[locusofself]

You lost me in the last sentence there. What do you mean?

[reinitctxoffset]

You and another commenter asked the same question at the same time, it's more honest to link the one reply than to paste it: https://news.ycombinator.com/item?id=48566097

[ElFitz]

> Claude Code is the cancer that will kill the patient, Boris is the the Kardashian version of Karpathy, with less business sense.

I am not sure I understand.

[reinitctxoffset]

Using a multi-trillion parameter softmax attention transformer to parse nested delimiters is a perverse thing to do. It is hard to imagine a sillier way to boil the oceans than feeding JSON to an LLM, a task that a pushdown automata from the 1960s effortlessly did on a PDP-X.

The API business throws a massive model that by definition can't be inferred efficiently because nothing can across 4 different compute substates, at a problem that DSv4 nails at or near 100% while leaving most of the actual unique value of Claude on the table.

Claude should be in your house and car and your kid's classroom and shit.

Having it write tail -n5?

That's because Anthropic's A-Team is Meta's C-Team. Hell, I fired some of their stars myself.

[tobyhinloopen]

You know what’s worse (less efficient) at parsing and writing code than LLMs?

Humans.

[Bolwin]

None of them even expose thinking anymore. You're seeing a summary

[reinitctxoffset]

I am happy to show you a thinking trace from 4.8 without even opening the web tools.

The legibility of thinking traces has gone down, but mostly this nets out to spilling more XML because their website is a vibecoded heap

[didip]

Is this AI bot? So annoying that bots can join in HN conversations.

[reinitctxoffset]

No it isn't, and I'm happy to prove it any way you like. If you decline that offer please retract the claim.

[babelfish]

post your comment about Meta/Anthropic C/A listers on LinkedIn and link it here