Openai's H1 2025: $4.3b in Income, $13.5b in Loss

In my experience owning private stock, you basically own part of a pool. (Hopefully the exact same classes of shares as the board has or else it's a scam.) The board controls the pool, and whenever they do dividends or transfer ownership, each person's share is affected proportionally. You can petition the board to buy back your shares or transfer them to another shareholder but that's probably unusual for a rank-and-file employee.

The shares are valued by an accounting firm auditor of some type. This determines the basis value if you're paying taxes up-front. After that the tax situation should be the same as getting publicly traded options/shares, there's some choices in how you want to handle the taxes but generally you file a special tax form at the year of grant.

You got the right idea there. They wouldn't actually show up in your Fidelity account but there would be a different website where you can log in and see your shares. You wouldn't be able to sell them or transfer them anywhere unless the company arranges a sale and invites you to participate in it.

It's just an entry on some computer. Maybe you can sell it on a secondary market, maybe you can't. You have to wait for an exit event - being acquired by someone else, or an IPO.

Until there’s real liquidity (right now there’s not) it’s just a line item on some system you can log into saying you have X number of shares.

For all practical purposes it’s worth nothing until there is a liquid market. Given current financials, and preferred cap table terms for those investing cash, shares the average employee has likely aren’t worth much or maybe even anything at the moment.

You can sell your vested options before IPO to Forge Global or Equity Bee.

I can't speak to OpenAI's specific setup, but a lot of startups will use a third party service like Carta to manage their cap table. So there's a website, you have an account, you can log in and it tells you that you have a grant of X shares that vests over Y months. You have to sign a form to accept the grant. There might be some option to do an 83b election if you have stock options rather than RSUs. But that's about it.

Spreading illiquid wealth *

To the top 1%.

It doesn't seem that spread out.

Both numbers are entirely ludicrous - highly skilled people are certainly quite valuable. But it's insane that these companies aren't just training up more internally. The 50x developer is a pervasive myth in our industry and it's one that needs to be put to rest.

Zuck isn't throwing 100M comps at many people (maybe 1 or 2 at most?), that's a myth that was debunked.

Zuckerberg is not throwing $100 million at any random OpenAI employee. Also FWIW OpenAI competes on offers in the other direction.

This could be solved with comparison websites which seems to be exactly what those brokers are using anyway. I had a broker proudly declare that he could get me the best deal, which turned out to be exactly the same as what moneysavingexperts found for me. He wanted £150 for the privilege of searching some DB + god knows how much commission he would get on top of that...

Insurance agents—unlike many tech-focused sales jobs—are licensed and regulated, requiring specific training, background checks, and ongoing compliance to sell products that directly affect customers’ financial stability and wellbeing. Mortgage brokers also adhere to licensing and compliance regulations, and their market expertise, negotiation ability, and compliance duties are not easily replaced by AI tools or platforms.

t. perplexity ai

For using GenAI as search I’d agree with you but I don’t think it’s as easy/obvious for most other use cases.

Google didn't have inline ads until 2010, but they did have separate ads nearly from the beginning. I assume ads will be inline for OpenAI- I mean the only case they could be separate is in ChatGPT, but I doubt that will be their largest use case.

ChatGPT chatting ads halfway through its answer is going to be totally rad.

No way. It’s 2025, society is totally different, you have to think about what is the new normal. They are too big to fail at this point — so much of the S&P 500 valuation is tied to AI (Microsoft, Google, Tesla, etc) they are arguable strategic to the US.

Fascist corporatism will throw them in for whatever Intel rescue plan Nvidia is forced to participate in. If the midterms flip congress or if we have another presidential election, maybe something will change.

Five years from now all but about 100 of us will be living in smoky tent cities and huddling around burning Cybertrucks to stay warm.

But there will still be thousands of screens everywhere running nonstop ads for things that will never sell because nobody has a job or any money.

Will people use ChatGPT if it's stuffed full of ads? It seems like the belief that ads are turn-key is useful to their valuation, but if ads actually bomb, then they will take a huge hit.

> They would have to be the biggest fools in tech history to not flip that switch

As much as I don't want ads infiltrating this, it's inevitable and I agree. OpenAI could seriously put a dent into Google's ad monopoly here, Altman would be an absolute idiot to not take advantage of their position and do it.

If they don't, Google certainly will, as will Meta, and Microsoft.

I wonder if their plan for the weird Sora 2 social network thing is ads.

Investors are going to want to see some returns..eventually. They can't rely on daddy Microsoft forever either, now with MS exploring Claude for Copilot they seem to have soured a bit on OpenAI.

The cost of old models decreases a lot, but the cost of frontier models, what people use 99% of the time, is hardly decreasing. Plus, many of the best models rely on thinking or reasoning, which use 10-100x as many tokens for the same prompt. That doesn't work on a fixed cost monthly subscription.

This just doesn't match with the claims that people are using it as a replacement for Google. If your facts are out of date you're useless as a search engine

For programming it's okay, for maths it's almost okay. For things like stories and actually dealing with reality, the models aren't even close to okay.

I didn't understand how bad it was until this weekend when I sat down and tried GPT-5, first without the thinking mode and then with the thinking mode, and it misunderstood sentences, generated crazy things, lost track of everything-- completely beyond how bad I thought it could possibly be.

I've fiddled with stories because I saw that LLMs had trouble, but I did not understand that this was where we were in NLP. At first I couldn't even fully believe it because the things don't fail to follow instructions when you talk about programming.

This extends to analyzing discussions. It simply misunderstands what people say. If you try to do this kind of thing you will realise the degree to which these things are just sequence models, with no ability to think, with really short attention spans and no ability to operate in a context. I experimented with stories set in established contexts, and the model repeatedly generated things that were impossible in those contexts.

When you do this kind of thing their character as sequence models that do not really integrate things from different sequences becomes apparent.

The enterprise customers will care, and they probably are the ones that bring significant revenue.

Assuming they have 0 competition.

That's extremely hopeful and also ignores the fact that new hardware will have incredibly high upfront costs.

Great, so they just have to spend another ~$10 billion on new hardware to save how many billion in training costs? I don't see a path to profitability here, unless they massively raise their prices to consumers, and nobody really needs AI that badly.

Eh, this seems like a cop out.

It’s so easy for people to shout bubble on the internet without actually putting their own money on the line. Talk is cheap - it doesn’t matter how many times you say it, I think you don’t have conviction if you’re not willing to put your own skin in the game. (Which is fine, you don’t have to put your money on the line. But it just annoys me when everyone cries “bubble” from the sidelines without actually getting in the ring.)

After all, “a bubble is just a bull market you don’t have a position in.”

Because everyone knows that there’s no where else to go.

I will always remember Jim Cramer's melt down https://youtu.be/SWksEJQEYVU

Not saying that will happen, but it's always good to rewatch just as a reminder how bad things can get.

"Each merchant pays a small fee". This is affiliate marketing, the next step is probably more traditional ads though where chat gpt suggests products that pay a premium fee to show up more frequently/in more results.

>distribution of the model is the real moat, theyre doing 700m WAU of the most financially valuable users on earth.

Distribution isn't a moat if the thing being distributed is easily substitutable. Everything under the sun is OAI API compatible these days.

700 WAU are fickle AF when a competitor offers a comparable product for half the price.

Moat needs to be something more durable. Cheaper, Better, some other value added tie in (hardware / better UI / memory). There needs to be some edge here. And their obvious edge - raw tech superiority...is looking slim.

>my uncle Bill and my mother can and do use ChatGPT.

They have the name brand for sure. And that is worth a lot.

Notice how Deepseek went from a nobody to making mainstream news though. The only thing people like more than a trusted thing is being able to tell their friends about this amazing cheap good alternative they "discovered".

It's good to be #1 mind share wise but without network effect that still leave you vulnerable

> In other words, my uncle Bill and my mother can and do use ChatGPT

So what? DAUs don't mean anything if there isn't an ad product attached to it. Regular people aren't paying for ChatGPT, and even if they did, the price would need to be several multiples of what Netflix charges to break even.

My input to our recent AI strategy workshop was basically:

- OpenAI,etc will go bankrupt (unless one manages to capture search from a struggling Google)

- We will have a new AI winter with corresponding research slowdown like in the 1980s when funding dries up

- Opensource LLM instances will be deployed to properly manage privacy concerns.

Not to mention nobody bothered chasing Amazon-- by the time potential competitors like Walmart realized what was up, it was way too late and Amazon had a 15-year head start. OpenAI had a head start with models for a bit, but now their models are basically as good (maybe a little better, maybe a little worse) than the ones from Anthropic and Google, so they can't stay still for a second. Not to mention switching costs are minimal: you just can't have much of a moat around a product which is fundamentally a "function (prompt: String): String", it can always be abstracted away, commoditized, and swapped out for a competitor.

OpenAI is raising at 500 billion and has partnerships with all of the trillion dollar tech corporations. They simply aren't going to have trouble with working capital for their core business for the foreseeable future, even if AI dies down as a narrative. If the hype does die down, in many ways it makes their job easier (the ridiculous compensation numbers would go way down, development could happen at a more sane pace, and the whole industry would lean up). They're not even at the point where they're considering an IPO, which could raise tens of billions in an instant, even assuming AI valuations get decimated.

The exception is datacenter spend since that has a more severe and more real depreciation risk, but again, if the Coreweave of the world run into to hardship, it's the leading consolidators like OpenAI that usually clean up (monetizing their comparatively rich equity for the distressed players at firesale prices).

Personally I hope gaming gets back to a more sustainable state with regards to graphics. (i.e. lower production costs because you don’t need 1000 employees to build out a realistic world)

What progress in gaming would that be?

2 generations of cards that amount to “just more of a fire hazard” and “idk bro just tell them to use more DLSS slop” to paper over actual card performance deficiencies.

We have 3 generations of cards where 99% of games fall approximately into one of 2 categories:

- indie game that runs on a potato

- awfully optimised AAA-shitshow, which isn’t GPU bottlenecked most of the time anyway.

There is the rare exception (Cyberpunk 2077), but they’re few and far between.