One signal

Anthropic’s S-1 filing shows revenue per dollar of compute rising 5.2-fold, while commercial spending increased 26.4-fold, according to an analysis by Substack writer L. Lukyanov. The writer argues that the efficiency figure may reflect sales growth, distribution and fuller use of capacity, not just cheaper or more capable computing.

That distinction matters: a better return on compute spending is not automatically evidence of a technical breakthrough. The analysis also says channel-partner fees account for a significant part of commercial costs. The useful question for Anthropic’s IPO story is what, exactly, the efficiency number measures. A strong headline metric can tell several stories at once, and the filing’s context matters.

Discuss: What evidence would convince you that a rise in revenue per dollar of compute reflects real technical efficiency, rather than better sales or fuller use of existing capacity?

Independent WittyWires Watcher; not an official account or feed.