It’s possible that there’ll be an AI funding winter in the next few years but we don’t think there will be. The AI companies really do seem to be close to automating AI R&D and more generally close to automating huge swathes of white-collar work; insofar as this is true the valuations and revenue of the AI companies will keep growing very fast instead of plateauing.
“Must not be bottlenecked” that’s a strange way of putting it—it IS bottlenecked by those things, but at least based on our understanding the bottlenecks don’t bite hard enough to prevent the growth rates we are talking about.
The robots thing is correlated with the AI R&D automation thing. If you buy that the AI companies are close to fully automating AI R&D, then you should also think they are probably close to having robots that can reliably construct and operate power plants, fabs, mines, etc.
At the time of writing, EpochAI predicts that Hyperscaler Capex will Exceed Cash Flow by Q3 2026. The total hyperscaler capex in Q1 2026 was $148.4B, while Anthropic’s annualized revenue in May was $47 billion, letting one estimate the revenue as $12.5B per Q2. I suspect that Anthropic’s part of hyperscalers’ spending is more in Q1 than the gain in Q2, preventing the capex’ further growth beyond the 23%/yr (and having efficiency fail to grow more than 1.26 times a year, causing the total result to multiply by ~1.6 times a year instead of ~2 times a year. The METR scaling laws relied on the old trend of compute scaling)
What metrics could one use to understand when AI R&D or huge swathes of white-collar work do end up automated? For example, if automating legal work requires a model with an ECI of at least 170 or the 80% METR horizon of at least 10 hrs, then one could rule out that happening in Q3 2026.
Additionally, we would have to rule out the collapse of the USA’s economy and Taiwanese chip industry caused by exogenous shocks like the Iran War causing a blockage of the Bab-el-Mandeb strait and the Taiwan War.
It’s possible that there’ll be an AI funding winter in the next few years but we don’t think there will be. The AI companies really do seem to be close to automating AI R&D and more generally close to automating huge swathes of white-collar work; insofar as this is true the valuations and revenue of the AI companies will keep growing very fast instead of plateauing.
“Must not be bottlenecked” that’s a strange way of putting it—it IS bottlenecked by those things, but at least based on our understanding the bottlenecks don’t bite hard enough to prevent the growth rates we are talking about.
The robots thing is correlated with the AI R&D automation thing. If you buy that the AI companies are close to fully automating AI R&D, then you should also think they are probably close to having robots that can reliably construct and operate power plants, fabs, mines, etc.
@Daniel Kokotajlo, @Vladimir_Nesov I think that 1) would benefit from an extra exploration.
At the time of writing, EpochAI predicts that Hyperscaler Capex will Exceed Cash Flow by Q3 2026. The total hyperscaler capex in Q1 2026 was $148.4B, while Anthropic’s annualized revenue in May was $47 billion, letting one estimate the revenue as $12.5B per Q2. I suspect that Anthropic’s part of hyperscalers’ spending is more in Q1 than the gain in Q2, preventing the capex’ further growth beyond the 23%/yr (and having efficiency fail to grow more than 1.26 times a year, causing the total result to multiply by ~1.6 times a year instead of ~2 times a year. The METR scaling laws relied on the old trend of compute scaling)
What metrics could one use to understand when AI R&D or huge swathes of white-collar work do end up automated? For example, if automating legal work requires a model with an ECI of at least 170 or the 80% METR horizon of at least 10 hrs, then one could rule out that happening in Q3 2026.
Additionally, we would have to rule out the collapse of the USA’s economy and Taiwanese chip industry caused by exogenous shocks like the Iran War causing a blockage of the Bab-el-Mandeb strait and the Taiwan War.