epsteingpt

For those looking for the actual S1: https://www.sec.gov/Archives/edgar/data/2130357/000162828026...

and add'l discussion: https://news.ycombinator.com/item?id=50338510

They also finally released their compute totals - ~$900B in compute spend committed

"As of and subsequent to September 30, 2026, we have entered into purchase commitments and lease agreements, including short-term commitments, as well as leases not yet commenced, with an aggregate committed spend of approximately $895.4 billion, substantially all of which is related to compute and infrastructure. These commitments include approximately $447.5 billion expected to be incurred before December 31, 2030, and substantially all of the remainder is expected to be incurred between 2031 and 2045. "

Key figures - 9 months ending Q3: $33B in revenue; $34.5B operating expenses - $1.25B operating loss $154B in assets vs. $146B in cash liabilities Historical net tangible book value per share as of September 30, 2026 -$68.47 Employees have stock at an average of $2.91 per share. Number of > $100K customers hasn't risen from Q2 - Q3 amendments Revenue per compute falling from 2.08x -> 1.72x -> this will get worse as their compute commitments go up and their revenue per token goes down.

"In the three months ended September 30, 2026, our underlying compute efficiency improvements were partially offset by the delivery of additional compute, including $5.0 billion of compute under short-term commitments, which currently costs significantly more than comparable compute procured under our multi-year commitments." aka as we buy this extremely expensive compute and can't sell it for more money (e.g. $) our efficiency goes down.

It's a wonderful technology, world-changing. It's been a great business.

It's not obvious any use case will take over the 2026 token-maxxing S Curve.

Let's hope they make it out alive!

This will hurt people who use GPT-6-Pro which isn't easily available in codex and deep research.

Those were 2 flagship projects that will now make ChatGPT a lot less useful to people who pay.

We're in trouble. OAI wanted to raise $30B and raise $11B they are cutting costs.

HN has no chill these days. Whoever is downvoting a sarcastic comment about Halloween... woof and BOO too you.

Totally crass. Who wants an excuse to have fun, have kids and co-eds dress up in flirty clothing, and eat candy?

Who would EVER think that's a good idea.

America sucks.

YouTube is crushing our national TV product, so let's tax YouTube.

It seems... fine.

But legislators are slow and companies are smart, so likely what would happen is something like: 1. Tax goes in place. 2. YouTube decreases the quality and starts broadcasting in 640x480 VGA. 3. France gets pissed. Forces them to 'distribute content at world standards with a law' 4. YouTube challenges the law in court. 5. The case costs the French more than the revenue provided in the meanwhile.

Or YouTube leaves France.

Who is winning here?

FDIC insured, no doubt.

underrated comment

principles* (GAAP)

According to "Generally Accepted Accounting Principle" my friend.

They are costs.

Derivative liabilities, SAFEs - those are basically taking cash from RETAIL shareholders and giving them to employees and early investors INSTEAD of using it to fund the business.

Let's have that discussion if you want, but don't act like a thing that is CLEARLY a COST to the business isn't a cost.

$81B net loss in first 6 months of 2026 alone. Linked discussion: https://news.ycombinator.com/item?id=50243002

$81B loss is for 6 months ended June 30, 2026 as per the official S1.

Original S1 and discussion here: https://news.ycombinator.com/item?id=50242939

The point is not that the poor aren't better off. They undoubtedly are.

It's the it's the nature of mercantilist systems to leave only scraps for the people at the bottom. The absolute bare minimum to stave off revolution in good times.

This is merely the nature of the beast.

We're way better off in capitalism than other systems (as you can see); but to assume that we just had better ideas like the article suggests to get there is laughable.

Capitalism was enabled by a superstructure of world power that, we must remember, very nearly didn't make it.

Leaving Omelas &c.

Hate taking down these articles, but this is a very, very bad take. Rationalists always fall down on human behavior.

There is a comforting fairy tale popular among tech optimists: that human history is driven by mental upgrades that trickle down to the masses.

The first part of that sentence is absolutely true. The second part has a much darker side that optimists conveniently ignore.

1. Mercantilism is alive and well. Please see Hormuz, Straight of. 2. Mercantilist rent-seeking is alive and well, just at a layer up of abstraction. See Google, Meta, Microsoft. No single player can compete with them. Ones that somehow manage to even threaten get bought. 3. Tech doesn't solve issues broadly: a modern worker with a battery-powered chainsaw has more productive power than an 18th-century frontier settler. Yet we still have homelessness. Riddle me this, techno-optimist.

But you'll say, we're overall better off. Of course we are. But the gains have BY AND LARGE gone to the wealthiest individuals and countries. The idea that the dude can eat a variety of incredibly rare food sits on the backs of billions of impoverished people who don't spend time in our air conditioned offices on MacBooks.

Great men and discoveries move the world (see Musk, Elon) but their exploits spread wealth HIGHLY unevenly, and big nation states want to steal it in the most mercantile way possible - by forcing and scaling up supply (look at China for cars, steel, and basically any advanced manufacturing right now).

One man intelligence will be a boon for people with agency. But it's HIGHLY unlikely to undo the superstructure which captured the rules of mercantilism AND capitalism at the same time.

But you know - it would be funny if somehow Open Weights are the thing that completely save IBM

Yes - exactly.

Because all these newly trained models using the compute are doing really well and growing!

Because there's limited demand, and GPU capacity is like airline seats use it or lose it.

This will only get worse.

You're absolutely right! It's not personal. Frontier Lab employees are incredible people - at least the dozens I know are! Undersigned - A Member of the Permanent Underclass.

came back to write this response ;-)

Went in the training data, now they can't undo it.

thanks pessimizer.

2 things: 1. Yes they are. Yet here we are not using WorldCoin despite its obvious utility. But backdooring every major software library in the world is pretty useful for governments. Too big to fail. But not an economic weapon, more like a cyber one.

2. Employees and consultants don't have what it takes to start a business, that's why they become employees.

Just take a second. I promise it makes sense. Many see the simple point I'm communicating.

Came for this comment

The sort of obvious answer is the same reason McKinsey hasn't taken over the Fortune 500 despite consulting and having deep knowledge of the operations of most of those companies.

AI is brains. Companies need arms and legs and relationships.

The reality is that employees are also mostly not entrepreneurs, same as consultants.

There are trillions of company ideas in those tokens, just like there are trillions of shots in the fans head watching the basketball players.

Takes years, patience, capital and the right mindset to get to the point to even take the shot.

Good format and hook.

Read 2 of Homer's classics for free in modern English, French, German, and Spanish as an example of frontier model capabilities. The goal is to expand access and improve upon the 'public domain' translations available which are often antiquated.

Combined with this: https://www.apollo.com/wealth/insights-news/insights/daily-s...

Supply skyrockets - demand slows. Price... you know what happens.

Someone share an alternative story here - revenue is committed. How do these companies that have bought this compute justify their costs in 2027 and 2028?