At a point when scale should be driving down costs, Anthropic's compute economics are moving in the wrong direction.
Anthropic is still struggling to make its compute costs work in its favor, with margins worsening rather than improving even as the company scales usage and revenue. That runs counter to the assumption that bigger deployment automatically means better unit economics in AI.
The pressure points to deeper structural issues in how frontier labs price and source compute, especially as demand for inference and training capacity continues to outstrip available power and chips.
If a leading lab like Anthropic can't bend its compute cost curve downward, that's a warning for every company building a business on top of frontier models. Pricing, margins, and competitive positioning across the AI stack may stay squeezed longer than markets are pricing in.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →