Andrew Freedman of Hedgeye asks whether slowing the pace of frontier model development means slowing revenue…
Follow-up to Pacing the frontier means more compute, not less — at the labs' margins' expense — incremental development in the same debate.
People think that slowing down the rate of new model development means slowing down revenue and pulling back on capex spending?
If it means anything it means the revenue from selling tokens will flatline or fall as existing models do not get replaced with better ones. That means the token maker will have less ability to buy compute without third party lending which will slow revenue and pull back capex
Just seems like there is larger opportunity in scaling and driving adoption of what exists… and that people's baseline is off, with RSI on the table that rate of model improvement actually set to potentially accelerate further. So by slowing the rate of improvement doesn't necessarily mean that revenue and capex expectations are too high. But just my two cents…
> People think that slowing down the rate of new model development means slowing down revenue and pulling back on capex spending? fwiw, the Bloomberg inference excerpted below is significant poetic license from what Dario actually wrote - but also Dario's careful word choice having its desired effect (something everyone can read as confirming their priors given enough motivated reasoning)… @business : "Amodei insisted in his essay that 'pacing' the advance of AI capabilities won't necessarily translate into reduced spending or growth." (stratechery.com/2026/pacing-th) 👇 👆 @DarioAmodei : "To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this." (darioamodei.com/post/we-must-p)