Enterprise token-cost ceiling: AT&T cuts AI spend 56%, Uber burns annual budget in 4 months

rohanpaul_ai · x · 2026-09-17

Enterprises are hitting a token-cost ceiling: AT&T cut AI task costs 56% with only 2% quality loss by routing to cheaper models, while Uber burned its entire annual AI budget in 4 months — pushing companies toward routing, flat-rate access, and open-weight models. A new 91-page Mozilla report adds that open-weight AI is now only 4 months behind the frontier: 8 of OpenRouter's top 10 models by August token volume were open-weight, 7 Chinese-built, with DeepSeek the first open model to lead weekly requests — yet open models handled only 20% of measured traffic while capturing disproportionately little revenue.

Original post →

More from Venture

Venture channel →