# AI Will Hit Trillions Before 2030 — Dario Amodei *Dwarkesh Clips* · https://www.youtube.com/watch?v=O18IcnyGEmw ## Summary Dario Amodei argues that AI could generate trillions of dollars in revenue before 2030, driven by a combination of technical progress and diffusion of the technology. He explains why leading AI firms currently lose money despite high gross margins: each model is profitable, but exponential compute scale-up for training the next model consumes those gains. He envisions an equilibrium where a 'country of geniuses in a data center' emerges around 2028 and accelerates revenue to trillions. ## Key points - Amodei finds it hard to see that there won't be trillions of dollars in AI revenue before 2030. - He sketches a plausible slow scenario where the 'country of geniuses in a data center' arrives in 2028, with revenue in the low hundreds of billions by then, accelerating to trillions by 2030. - He suspects that composing the technical exponential and the diffusion exponential will get to trillions before 2030. - Gross profit margins on inference are very high because inference is efficient, and models are differentiated despite some competition. - With a small number of firms (e.g., three), the Cournot equilibrium does not equilibrate to perfect competition with zero margins. - Companies currently lose money because they are in the exponential scale-up phase of compute: each model makes money, but spending on training the next model exceeds those gains. - He gives a stylized example: a model trained for $1 billion produces $4 billion of revenue with $1 billion inference cost, yielding 75% gross margins and $2 billion profit, while the company spends $10 billion training the next model. - The equilibrium he describes is one where the 'country of geniuses' exists and model training scale-up has leveled out or equilibrated. ## Notes ### Trillions in AI Revenue Before 2030 - Dario Amodei rejects the framing that we are 10 years away from generating trillions; he predicts trillions of dollars in revenue before 2030. - A plausible (slow-end) scenario: the "country of geniuses in a data center" arrives in 2028, with revenue already in the low hundreds of billions by then. - That "country of geniuses" then accelerates revenue to trillions, taking perhaps two years of diffusion — reaching trillions by 2030. - He suspects that combining the technical exponential with the diffusion exponential will get there before 2030. ### Why Firms Can Be Profitable - Amodei abstracts the industry into an economics-textbook model: a small number of firms, each investing a fraction in R&D, with some marginal cost to serve. - Gross profit margins on that marginal cost are very high because inference is efficient. - There is some competition, but models are also differentiated. - With a small number of players — e.g., three firms behaving rationally — the equilibrium (he references the Cournot equilibrium) does not collapse to perfect competition with zero margins. ### Why Leading Firms Lose Money Today - Gross margins right now are very positive, but two things are happening. - First, the industry is still in the exponential scale-up phase of compute. - Stylized example: a model trained for $1 billion last year produces $4 billion of revenue and costs $1 billion to inference from — 75% gross margins, so the model makes $2 billion. - But the company simultaneously spends $10 billion training the next model, so each model makes money while the company loses money. ### The Equilibrium He Envisions - The equilibrium he describes is one with the "country of geniuses in a data center," where model training scale-up has equilibrated or leveled out. - Compute may still be scaling as firms try to predict demand, but it is more leveled out than today. ## Transcript [00:00] It seems like your worldview is compatible with somebody who says uh we're like 10 years away from a world in which like we're generating trillions. >> That's just that's just not my view. Yeah, that is that is not my view. Like I I so so I'll like I'll like make another prediction. It is hard for me to see that that there won't be trillions of dollars in revenue before 2030. Um like uh I can I can construct a plausible world. It takes maybe three years. So that that you know that would be the end of what I think it's plausible like in 2028 we get the the real country of geniuses in a data center. You know the revenue's been been go you know the revenue has been going into the maybe is is in the low hundreds [00:39] of billions by by by by 2028 and and and then the country of geniuses accelerates it to trillions you know and and we're basically we're basically on the slow end of diffusion. It takes two years to get to the trillions. that that that would that that that would be the world where it takes until that would be the world where it takes until 2030. I I I suspect even composing the technical exponential and diffusion exponential will get there before 2030. So you laid out a model where anthropic makes profit because it seems like fundamentally we're in a computed world and so it's like eventually we keep growing comput. No, I think I think the way the profit comes is again and and you know, let's [01:18] let's just abstract the whole industry here. Like we have a you know, let's just imagine we're we're in like an economics textbook. We have a small number of firms each can invest a limited amount in you or or or like each can invest some fra fraction in R&D. They have some marginal cost to serve. the margins on that the profit margin the gross profit margins on that marginal cost are like very high because because because inference is efficient there's some competition but the models are also differentiated there's some there's some um you know companies will compete to push their research budgets up but like because there's a small number of players you know we have the what is it called in the corno [01:59] equilibrium I think is what the what the small number of firm equilibrium is it the point is it it doesn't equilibrate to perfect competition with with with with with with zero margins. If there's like three firms, if there's three firms in the economy, all are kind of independently behaving behaving rationally, it doesn't equilibrate to zero. >> Um, help me understand that because right now we do have three leading firms and they're not making profit. Um, and so what what uh yeah, what what is changing? Yeah. So the the again the gross margins right now are very positive. What's happen what what's happening is a combination of two things. One is we're still in the exponential scale up phase of compute. [02:43] Um so what basically what that means is we're training like a model gets trained it costs you know let's say a model got trained that costs uh a billion last year um and then uh this year it produced uh $4 billion of revenue and cost $1 billion to to uh to to to inference from. Um so you know again I'm using stylized number here but you know that would be 75% you know gross gross gross margins and you know this this 25% tax. So that model as a whole makes $2 billion. Um but at the same time we're spending $10 billion to train the next model because there's an exponential scale up and so the company loses money. Each model makes money but the company loses money. The equilibrium I'm talking about is an equilibrium where we have [03:34] the country of geniuses. We have the country of geniuses in a data center, but that that um model training scale up has equilibrated more. Maybe maybe it's still it's still going up. We're still trying to predict the demand, but it's more it's more um leveled out. If you enjoyed this clip, you can watch the full episode here and subscribe for more clips. Thanks.