Post

caught your tech radar piece on youtube. great interview! watching all of this rapt with interest

Saw your shout out last week for finance/accounting help. Thought something was up. Well handled and congratulations.

Can't wait for this to somehow boost OpenAI's stock price

You just got cited in the FT 33 minutes ago on this very subject.

... either never materialised (other than a few FOSS models) or were never a threat in the first place. Like, these new "models" are good at finding vulnerabilities? Or where the code bases done by lazy, exploited or underfunded, outsourced programmers in the first place? Even before AI rot was...

I think there is a big 'market' for LLMs, actually, albeit not in socially productive uses: slop art/writing, spam, plagiarism, fraud, and what Grok is famous for. But by definition, the kinds of people who always pursue the shortcut or grift will rarely pay hundreds per month. Oh, and some coding.

So...not good?

itshappening.gif

I endorse this soundtrack.

Whats your take on the FT reporting that the 2025 increase is mostly one-time costs due to the for-profit conversion? Seems unlikely to me given the intercept reported expected 2026 losses of ~$36B, which is oddly in line.

Cc: @carlquintanilla.bsky.social

One morbillion dollars more for is- I mean, anthropenxai

But for goodness sake, don't buy an espresso drink at a coffee shop, and don't buy avocado toast.

I mean, we all knew none of these companies are actually making any money. Investors just keep wanting to burn everyone's money on 'vibes' instead of actually, you know, doing legally obligated due diligence. They'll be sued into the ground and sent to jail when the house of cards falls.

Big loses but yet still OpenAI has no "real" product, nothing that stands out from the rest anyway.

Reading the article, does that mean that OpenAI has a gross margin on inference of around 40%? I know this doesn’t say *much* as there are absolutely bonkers costs elsewhere and opportunities for financial engineering, but doesn’t it go against your message of inference not being profitable?

Is that bad

Should data centers have a kill switch? Just in case

Wow! Even worse than I expected!

"At the end of the year, OpenAI had just over $50 billion in assets, with almost half of that in cash." It is a profound failure of our legal system and regulations that that 25 billion in cash is not immediately seized and redistributed to everyone OpenAI has stolen from.

Show this to anyone worried about AI

I believe it has told people to kill those around them at least. Simple search:

yes! but that's also why they are all rushing for the cash out, i'm sure they know this

I literally had someone pull a "but Amazon wasn't profitable for years" on me just last week (followed by a reference to Pets.com, so I guess that's some kind of booster bingo?). 🙄

Is this the info that would have to go on an S-1?

There’s an old saying that goes, “The game isn’t worth the candle.”

That seems high.

The inviability of general purpose LLMs hosted by mega corps is obvious, but I'm curious about your thoughts on locally run specialized open source models. I have my existing ethical qualms with LLMs, but do you think this sort of thing could be economically feasible for a company of 1-2 people?

I'm not a doctor; is this good?

So you're saying it's going great

You know that if you keep this up you're going to have to practice for testifying to Congress, right? I feel like once the fit hits the shan you'll be in high demand at the legislative level.

Part of their original investment was an agreement to use OAI's products (probably at a discount, but who knows, Masayoshi Son is bonkers).

Oooofff…

$38.5 billion in losses... To put that into perspective, Serbia's 2024 government budget was $38.29 billion (IMF estimate). So they lost more money than the Serbian government spent to run their country.

Ed, you absolute madman :-D \m/

Not trying to be a conspiracy theorist here, but why is there basically zero “mainstream news” coverage of this? Double-checking myself, I found this MSN article that quotes @edzitron.com and seems to conclude “meh”. www.msn.com/en-us/news/t...

With the amount of bleeding they're doing, I'd say it's Antietam. Not the high water mark, not quite yet. The upcoming IPOs may be Chancellorsville and Fredericksburg though...

Also open ai and anthropic combined have committed to iirc over a trillion in compute commitments, so even if they did quadruple revenue without raising expenses at all, they'd still be crushed by debt if the data centers get built on time

Oops, I misunderstood that.

youtu.be/PyM_6r4XJ5A?...

I don't have enough cash but I'd love to short/bet against all of them

So what do they have to charge per token to break even?

That is a good question.

They spent $5 on sales and marketing for each $13 of revenue. I don't know much about sales and marketing, but that seems bad.

Reddit.

The frontier models are all close enough that which you pick doesn't *really* matter for the vast majority of use-cases. Which is why it's crazy to invest a small-countries GDP into R&D every year to try and keep up. There's no moat.

who the fuck is spending money on monopoly go

NYPD has to do SOMETHING while on overtime

God, that’s a depressing list. And here I am still playing Threes and Good Sudoku every day…

Just finished reading. This is utter insanity.

Do they capitalize/amortize any of their R&D or are those expenses all taken in the same period?

If revenue is $13B and cost of revenue is $7.5B are they profitable on a per token basis? But if their model training is covered under R&D they can't exactly stop doing that because they always need to train models or they get out of date right? So then shouldn't that be included in cost of revenue?

I was reading the numbers, and their Marketing costs seemed astronomically inflated. Then I remembered something.. 2024 and especially 2025 saw a bunch of stuff like this.. "strategic partnerships" there was one with Wikipedia too if I recall.. (Buying out massmedia outlets)

PONZI

If costs are increasing, now is the perfect time to cut prices 🤪

60 billion. Think about the good that could have been done with that.

❤️

The devil is in the details, isn’t it?

I got slated on here earlier for talking about Reddit vibes, but I can tell you now, casual users are flocking to Chinese models or open source models.

it’s on the front page of the financial times

That fascist syndicate got everything they wanted. It's also the same group, basically.

We must uphold Smiling Manism and Ed Zitron Thought

congratulations, Ed!

Hey Ed, is the Age of Smiles upon us yet?

"Soundtrack: In Flames - Colony" Based

Is there any chance these $17.87 billion are revenue backstop payments, accounted in this fucky manner because it would completely destroy the "insatiable AI demand" myth?

This does not seem a like a sign of good things to come.

I suspect they are misclassifying some opex as R&D. Makes the gross margins look better and it might also trigger the R&D tax credit although it's gonna take forever to work off the carryforward if ever.

Are political *ahem* donations considered "marketing"?

I'm assuming they are using it one way or another to subsidise

Ed you have a very cool taste in music :))

All I know is whenever I call and get an AI Babe in customer service I either never get an answer or it takes so goddamn long to get to someone who does have an answer! Or I get told by AI Babe saying that they have very few customer people working so "email us" for an answer in 24 to 48 hours. 🤬

That's just a data center with a fancy generated graphic that has a consumer-facing business.

> Due to the seriousness of this story, I am not going to do very much editorializing Well, where's the fun in that?

yeag

Can it just happen already?

Those numbers, if used to try and get a mortgage loan (the deficit income) would get a reply of "no way in hell". Why anyone thinks there isn't an AI/DataCenter Bubble about to burst doesn't read any damned financial history.

at this point my question is like, do i pull my retirement out early before the bubble pops? I'm not enough of a finance-knower to know if the bubble is going to ruin all my my retirement and if it's just worth it to pull it early and wait that out. money is dumb

And people are actually willing to invest in this?

This is THE story?