# Avoiding Bankruptcy Before AGI – Dario Amodei

*Dwarkesh Clips* · https://www.youtube.com/watch?v=y8ZEJe2Ss9k

## Summary

Dario Amodei explains why Anthropic cannot simply buy compute as if its 10x annual revenue growth will continue indefinitely. Even if powerful AI arrives around 2026 or 2027 as he suspects, economic value like curing diseases takes years to diffuse, and overbuying data centers by even a year or two could bankrupt the company. He contrasts Anthropic's thoughtful approach with what he sees as other companies spending recklessly.

## Key points

- Dario Amodei estimates powerful AI, a 'country of genius in the data center', could arrive in 2026 or 2027, though he could be off by a year or two.
- Even after powerful AI exists, curing diseases takes time due to manufacturing, regulatory processes, and access, as shown by the year and a half it took to distribute COVID vaccines.
- Anthropic's revenue has grown about 10x a year, reaching a 10 billion dollar annualized rate at the beginning of the year.
- If Anthropic bought compute assuming revenue of 1 trillion dollars and actual revenue came in at 800 billion, no hedge could prevent bankruptcy.
- Amodei says Anthropic supports hundreds of billions in spending, not trillions, accepting risk on both demand and slower growth.
- He criticizes other companies for spending without understanding their risk, saying they seem to be 'yoloing' rather than being thoughtful.

## Notes

### Powerful AI timing
- Dario Amodei's hunch, from his essay "Machines of Loving Grace", is that powerful AI — a "country of genius in the data center" — could arrive in 2026 or 2027.
- He acknowledges he could be off by a year or two.

### Diffusion takes time even after the breakthrough
- The arrival of powerful AI is a "starting gun", not the finish line.
- Curing diseases requires biological discovery, drug manufacturing, and regulatory approval — the COVID vaccine took a year and a half to reach everyone.
- Polio vaccine has existed for 50 years, yet eradication in remote corners of Africa is still incomplete, though Amodei calls this the most difficult case.
- His conclusion: diffusion will be faster than anything we've seen, but still has limits.

### The data center bankruptcy dilemma
- Anthropic has seen roughly 10x revenue growth per year, at a 10 billion dollar annualized rate at the beginning of the year.
- Data centers must be bought a year or two in advance.
- If he assumed 10x growth continued, revenue would be 100 billion at end of 2026 and 1 trillion at end of 2027, justifying buying roughly 5 trillion dollars of compute (a trillion a year for five years).
- But if revenue came in at 800 billion instead of 1 trillion, "there's no hedge on earth" that could stop bankruptcy.
- Being off by a year, or growing at 5x instead of 10x, would also mean bankruptcy.

### Behaving responsibly
- Anthropic ends up supporting hundreds of billions, not trillions, accepting risk both of excess demand and of slower growth.
- Amodei distinguishes responsible behavior from absolute spending amounts — the key is being thoughtful rather than "yoloing".
- He suspects some other companies haven't written down the spreadsheet and don't understand the risk they're taking, doing things because they "sound cool".

## Transcript

[00:00] Even if the technology goes as fast as I suspect that it will, we we don't know exactly how fast it's going to drive revenue. We we know it's coming, but with the way you buy these data centers, if you're off by a couple years, that can be ruinous. It is just like how I wrote, you know, in machines of loving grace, I said, look, I think we might get this powerful AI, this country of genius in the data center, that description you gave comes from the machines of loving grace. I said, "We'll get that 2026, maybe 2027." Again, that is that is my hunch. Wouldn't be surprised if I'm off by a year or two, but like that is my hunch. Let's say that happens. That's the starting gun. How long does it take to cure all the

[00:39] diseases, right? That's that's one of the ways that like drives a huge amount of of of of economic value, right? Like you cure you cure every disease. You know there's a question of how much of that goes to the pharmaceutical company to the AI company but there's an enormous consumer surplus because everyone you know assuming we can get access for everyone which I care about greatly we you know we we cure all of these diseases how long does it take you have to do the biological discovery you have to you know go you have to you know manufacture the new drug you have to you know go through the regulatory process I mean we saw this with like vaccines and co right like it there's just this we we got the vaccine out to everyone but it

[01:17] took a year and a half, right? And and so my question is, how long does it take to get the cure for everything, which AI is the genius that can in theory invent out to everyone? How long from when that AI first exists in the lab to when diseases have actually been cured for everyone, right? In in you know, we've had a polio vaccine for 50 years. We're still trying to eradicate it in the most remote corners of Africa. And you know, the Gates Foundation is trying as hard as they can. others are trying as hard as they can but you know that's difficult again I you know I don't expect most of the economic diffusion to be as difficult as that right that's like the most difficult case but but there's a there's a real dilemma here

[01:59] and and where I've settled on it is it will be it will be it will be faster than anything we've seen in the world but it still has its limits and and so then when we go to buying data centers you know you again again the curve I'm looking at is okay. We you know we've had a 10x a year increase every year. So beginning of this year we're looking at 10 billion in in in annual in you know rate of annualized revenue. At the beginning of the year we have to decide how much compute to buy. Um and you know it takes a year or two to actually build out the data centers to reserve the data centers. So basically I'm saying like in uh 2027 how much compute do I get? Well, I could assume um uh that uh the uh revenue will

[02:52] continue growing 10x a year. So it'll be uh one uh one 100 billion at the end of 2026 and 1 trillion at the end of 2027. And so I could buy a trillion dollars. Actually, it would be like5 trillion dollars of compute because it would be a trillion dollar a year for for five years, right? I could buy a trillion dollars of compute that starts at the end of 2027. And if my if my revenue is not a trillion dollars, if it's even 800 billion, there's no force on earth. There's there's no hedge on earth that could stop me from going bankrupt if I if I buy that much compute. And and so even though a part of my brain wonders if it's going to keep growing 10x, I can't buy a trillion dollars a year of compute in in in in in in in in

[03:40] 2027. Uh if I'm just off by a year in that rate of growth, or if the the growth rate is 5x a year instead of 10x a year, then then you know you go bankrupt. Um and and and so you end up in a world where you know you're supporting hundreds of billions not trillions and you accept you accept some risk that there's so much demand that you can't support the revenue and you accept still some risk that you know you got it wrong and it's still slow and so when I talked about behaving responsibly what I meant actually was not the absolute amount that that actually was not um you know I think it is true we're spending somewhat less than some of the other players. It's actually the other things like have we been thoughtful

[04:26] about it or are we yoloing and saying, "Oh, we're going to do hundred billion here or hundred billion dollars there." I kind of get the impression that, you know, some of the other companies have not written down the spreadsheet that they don't really understand the risk they're taking. They're just kind of doing stuff because it sounds cool. If you enjoyed this clip, you can watch the full episode here and subscribe for more clips. Thanks.