Uber is reconsidering its massive investment in Artificial Intelligence as top executives report that the high costs are no longer matching the utility delivered to the platform. Senior engineering leaders have observed that a surge in token consumption—the metric used to measure generative AI usage—has failed to translate into a proportional increase in valuable features or efficiency for end-users. This gap between expenditure and actual output has sparked internal debates about the long-term viability of current AI spending models.
While many tech giants are doubling down on large language models, Uber's skepticism centers on a lack of ROI (Return on Investment) for complex tasks. Key challenges identified include:
- High operational costs associated with LLM API calls and infrastructure.
- Diminishing marginal utility where more data processing does not yield better user experiences.
- Difficulties in justifying expensive inference costs for features that don't directly drive revenue.
