bottleneckScore 30/100Add to thesis
Cost of AI inference is rising sharply—Claude code being cut due to prohibitive token expenses—creating pressure on model deployment economics
Mustafa Suleyman· Microsoft AI· AI· 2026-06-02· about Anthropic
“I want to ask you Mustafa about that question of cost and we talked about tokens and Uber for example today said that they were sort of cutting off Claude code because it's getting so expensive”
Why it matters
This reveals that even large companies (Uber) are abandoning AI features due to token costs. This is a material constraint on AI adoption and suggests inference costs are not yet sustainable at scale for many enterprise use cases.
Investment implication
Companies focused on inference optimization, cost reduction (quantization, distillation, edge deployment), or alternative model architectures with lower compute footprints will be differentiators. This also validates demand for more efficient model architectures and specialized inference hardware.