These commitments aren't annual though, most of them amount to 10 billion a year commitments for 10 years. Put another way Anthropic has committed 52 billion dollar a year to these companies and their S-1 indicates they earn 67.2B a year in revenue.

I'm not going to predict whether Anthropic succeeds or fails, just that it's very misleading to claim that Anthropic needs to grow to $500b in revenue without making it clear that it's $500B in revenue per decade rather than $500B in revenue per year which is what most people reading such a number would assume.

Meanwhile the Chinese are going to undercut them every chance they get.

Today's prices are the ceiling. But AI capability is getting less of a scarce resource with every passing day and it would be a big mistake to plan that your current margins are going to be your future margins as well. Long terms AI will converge on energy costs + capex write-off + margin and the latter will be a race to the bottom.

Comments here are not considering what Anthropic is doing with Claude as a product. The models matter, yes. But they are building Claude into the only tool you need to get business done. In the past few weeks they have added slides and docs - a shot across the bow to Microsoft's dominance. If they can pull off building the best interface to get work done, $500b in revenue starts to seem not too crazy. It's more about selling a complete product than it is about selling tokens.

the market for their valuation is just not there imo. they need to significantly expand into non-coding products and outshine their coding offerings to make it work.

as in, if they capture every single professional software dev and extract $1000 from them every month, their yearly revenue will be like $500B. do your own back of the envelope on that with ongoing costs and consider how unrealistic a goal this actually is.

even worse, the very product they offer is self limiting. if it becomes so good as to actually replace software devs, they will have less customers to extract money from.

their valuation essentially relies on a small percentage of mega users having unheard of levels of spending.

the part about a small percentage giving them most of their money is already true (their rumored $100+ bil revenue as of mid 2026 already outshines all software devs having a $100 max subscription), but they are two orders of magnitude or so away in terms of the actual amounts to make it work.

If you owe the bank a million dollars you have a problem. If you owe the bank a billion dollars the bank has a problem.

The same logic applies here.

Google, Amazon, Microsoft, Meta, Broadcom, Nvidia, SpaceX and others are spending hundreds of billions a year each on fast-depreciating infrastructure to serve this potential AI demand. If the demand never materializes then they know they are in big trouble. So they have every incentive in the world to prop up Anthropic (and OpenAI) and keep the party going.

Can they keep the bubble inflated indefinitely? Probably not, but all these companies combined have heck of a lot more influence than Anthropic alone.

I have a hard time believing they are likely to get to $500bn in revenue. That’s another $450bn in revenue growth that needs to happen before demand tapers off.

But I see every reason to believe demand will taper off. It seems it’s already happening. Enterprises have been deciding that maybe corporate chatbot and coding agent budgets shouldn’t actually be infinite. AI startups are starting to successfully outcompete each other, effectively killing off members of Anthropic’s customer pool in the process. And consumer demand for AI products seems to be dropping. None of these are signs that suggest that Anthropic has thus far met only 10% of plausible long-term demand.

> Everything comes back to whether the demand is really there

And unit economics.

HY1 2026 revenue less compute and infrastructure turned positive (45% gross compute margin; $8.9bn C&I to turn $16bn revenue) for the first time; it was -176% in FY 2025 with $7.3bn C&I on $4.6bn revenue.

It looks like a lot of costs of revenue are in sales, marketing and partnerships, though. Backing those out shows H1 2026 margin at 26% ($4.2bn) after -72% for FY 2025. Still positive.

>Anthropic thinks it is.

Anthropic has no choice, just like you said. It's crush and burn right now or double down.

And they will keep doubling down, because stopping is simply not an option.

These were ~Q1-Q2 commitments based on what Anthropic THOUGHT token maxed compute demand was going to be over next few years. Then Late Q2, and Q3 PRC models dropped floor for token prices. It may very well be that Jevons demand which Anthropic though required 500B of compute now only requires 100B, and unless demand is 5x what they thought, those additional 400B non hardware cancellable commitment are going to be balance sheet concrete shoes.

No it doesn’t. Everything comes back to whether the fundamental business can ever be profitable.

So far not a single company has ever demonstrated that.

Even if demand is through the roof which it seems to be, it doesn’t matter if every dollar they sell costs them more than a dollar.

Now ask yourself what it takes to generate 500bn…

You people are nuts. The failure risk is off the charts.

Funny animated cartoon from the Soviet Union (has subtitles). You must see this:

“Robbery, Italian Style”

https://www.youtube.com/watch?v=vktTgOr4M1s

(it has a wiki page https://en.wikipedia.org/wiki/Ograblenie_po... )

Marcello Mastroianni is performing a bank heist, it goes well - but in the end he has to return all the money he borrowed. And that's where he is left with nothing...

The SpaceX thing interests me. They are basically paying SpaceX at a $10B per gigawatt equivalent rate right now, yeah? Because they only get 0.3GW for their $3B.

So what that means is that if the demand exists for any AI data centre, it's a solid business to run one — they cost maybe $50-60B to build out, they cost maybe $10B per year to run. You get your money back fast if the demand is there? And you can project being able to replace with better GPUs in the future without having to borrow for all of it.

But it is extremely difficult to believe that Anthropic is not projecting that all of its customers will ultimately use their service to the extent that their outlier customers do right now. They probably won't be projecting that their customers will find the need or ways to use them less.

And also, "no customer is more than 10% of our revenue" means that one customer is. If that customer is Microsoft (can it be? I honestly don't know) then things get tricky, because Microsoft clearly want an unmetered, on-device AI future.

Not just is the the man there? Will anthropic be able to capitalize on that demand? Or are we going to build so much excess capacity that tokens are going to be dirt cheap (like what happened with the fiber build out in the 90s).

Will anthropics models continue to be significantly better than openweight models? Even if their latest model is, Will the open weights gobble up all of the demand for lower models? (I.e. will anthropic be able to recoup training cost while their model is the best, will they make enough revenue on the long tail or will it get crushed by the open weights)

Crazy that SpaceX has similar commitments to AWS/GCP. Means it’s very bullish for SpaceX given they produce their own energy turbines and have the best ability to deal with supply chain constraints

You know how they say, don't keep all of your eggs in one basket? It's really concerning to me that they've got all their eggs on LLMs. They really should have branched out more long ago. OpenAI has a lot of branches they can aim towards.

For kicks - lets take the optimistic side here (because the whole thread will be reading pessimistic). What if they are under built and this growth continues to go parabolic. We then still have supply constraints and pent up demand for frontier model and you have a company thats currently at 1-1.5 T and levels up to 4-5 T.

I'm not saying I think thats what is going to happen, but it really isn't a stretch to think that this keeps running given that it hasn't really hit anywhere near as broad as we think it has (being HN community based).

I've seen the bills at my org, I think the demand is there for now.

Our claude bills are rivaling our datadog bills and might get to the size of our AWS bills. It's pretty crazy.

There's an unlimited demand for intelligence.

Even the smallest margin or error in the demand forecast will put them at a huge loss, I can't imagine sleeping well at night if I was on that management team.

> Anthropic thinks it is.

Does that have to be true? I thought IPOs were about selling the company.

Never mind the fact that companies do not think and this is like saying "My lemonade stand thinks it's going to have good sales this quarter."

It's not just about demand, it's about do you think valuation that prices them into serving 1000x that demand is worth paying for.

The obvious business is rewriting all of our software. That sounds flippant, but I'm entirely serious. I'm of the opinion that all software written by humans will get replaced and in pretty rapid order.

First to fall is the software that has basically been abandoned but industry leading and has sticky customers. We're replacing those right now. Our version is more accurate, ten times faster, and a hundred times cheaper.

The average human spends 100$ per month in food.

> If the party continues, they'll go from $50b in revenue to $500b and be fine.

You really have to be insane to think that an AI company will ever reach $500B in revenue. That's almost the same as (2025) Amazon, Walmart, etc., companies with a MUCH wider market.

It is MORE than Apple, it's 5 times more than NVidia even. There is absolutely no software seller that is even remotely close to $500B in revenue.

Absolutely ridiculous.

Broadcom isn't a name I'd have expected to see next to the other three.

There are not enough revenue in tokens, etc. The only way AI ever becomes profitable is to replace labor everywhere. That is the only market big enough to make all the spend in AI profitable.

Their entire goal is to get to AGI via RSI ASAP. If they do they win. But if not they’ll be a legendary story of a flying too high too fast.

That they are IPOing tells me that the massive amount of funding they need is no longer available privately and that this is their 1 opportunity to get the funds they think they need to get there.

Biggest risk is the growth stalls as everyone moves to open weight models. They certainly like telling stories.

So, puts?

If someone just way outcompetes them it's possible they sell their commitments to the winner and still do ok.

> But if they don't, it's going to hurt.

They literally said revenue growth rate "may not be sustainable" and "may decrease".

So get ready for the hurt.

It is a business losing $80B a half on $18B in revenue. This is a trash business.

[flagged]

[flagged]

> If the party continues, they'll go from $50b in revenue to $500b and be fine.

Heads I win, tails we all lose. GFC Part II.

I do not understand 500 billion revenue? I mean how much is that in terms of tokens burned?