To be more specific:

- 161.2B w/ Broadcom: 2026-2031

- 120.5B w/ SpaceX: 2026-2029

- 114.5B w/ Google: 2026-2033

- 110.0B w/ Amazon: 2026-2036

- 44.9B w/ Microsoft: 2026-2033

- 44.6B w/ Nscale: 2026-2033

- 41.3B w/ Lambda: 2026-2034

which is around $63.7 billion / year if spread over 10 years, but with most of it due earlier than that.

Thanks for clarifying, that’s a huge difference. Obviously they will have other revenue sources, but will likely make a small portion of their revenue. If we approximate their growth rate by these numbers, it is quite a slowdown.

You're right that my numbers are imprecise, but I did mean $500b in revenue annually.

This is just the compute, which primarily covers training + inference. They still have to run the rest of the business. If you think AI should look like SaaS and spend 10% of their revenue on compute, that's $1T in revenue. If you think it's a hardware business, that's 30% margins and perhaps $200b in revenue. I went with something in the middle at $500b.

They won't disappear as a company at $200b in revenue, but I don't think that's their aspiration. These commitments imply to me that they're confident they can do much better.

> Put another way Anthropic has committed 52 billion dollar a year

Idk where you got this figure from, but the actual figure are very different from that:

It's 120B over four years for SpaceX alone and 120B for Broadcom over 6 years. These two players, representing less than half of the total commitment, already give you an annual commitment above 50B.

Good point. Although they also need to make a profit. To justify an eventual PE ratio of 25 and a current valuation of $1 trillion they need to make an annual profit $40 billion. And then lets assume $40 billion in other expenses. So they need revenue of $50 + $40 + $40 = $130 billion revenue annually.

But if you buy them today at a $1 trillion valuation, and they only make $40 billion in annual profit, your investment will go nowhere. They need to be growing profits at 20% per annum to justify a PE OF 25.

And then you have the competition problem. The cheap open weight models are good enough for many (most?) tasks.

And then you have the problem that LLMs may not be that useful outside of programming, customer support and law (and doing homework).

500B in a decade is really a lot

> 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.

A very large part of that revenue is paying for inference. There’s not enough leftover for the committed outlays. So it has to come from smucks I mean investors.